← All companies

Harvia

HARVIA.HEHelsinkiEURSauna & wellness equipmentReviewed Nov 15, 2026

Organic growth and reinvestment

8 of 11 research checks complete. Still to complete: Security and evidence base; Opportunities and funded scenarios; Variant view and catalysts · TTM to September 2026; September financial position plus known October dividend; fixed October 2026 forecast vintage

Replacement demand, distribution gains and high returns on organic investment drive compounding. Growth must fund working capital and capacity before cash reaches shareholders.

  • Organic growth by geography
  • Margin recovery and capacity utilization
  • Maintenance, leases and working capital
Inspect live quotes, sensitivity grid & cash-flow bridge

Scenario assumptions

Organic growth and reinvestment · TTM to September 2026; September financial position plus known October dividend; fixed October 2026 forecast vintage

DriverDownsideBaseUpside
Near-term organic growth0.0%10.0%17.0%
Long-run organic growth5.0%9.0%9.0%
Normalized EBITDA margin22.7%25.7%29.2%
Return on incremental organic capital26.1%38.4%45.0%
Annual dividend growth5.0%5.0%5.0%
Exit EV / steady-state cash NOPAT11.89×20.00×20.00×
Additional annual share dilution0.0%0.0%0.0%
Revenue growth
Year 1: 0.0%
Year 2: 0.6%
Year 3: 1.1%
Year 4: 1.7%
Year 5: 2.2%
Year 1: 10.0%
Year 2: 9.9%
Year 3: 9.8%
Year 4: 9.7%
Year 5: 9.6%
Year 1: 17.0%
Year 2: 16.1%
Year 3: 15.2%
Year 4: 14.3%
Year 5: 13.4%
Margin
Years 1–5: 22.7%
Year 1: 23.3%
Year 2: 23.9%
Year 3: 24.5%
Year 4: 25.1%
Year 5: 25.7%
Year 1: 24.0%
Year 2: 25.3%
Year 3: 26.6%
Year 4: 27.9%
Year 5: 29.2%
  • Downside case. Adverse adoption/destocking with 0% initial growth recovering toward 5%, no recovery from current 22.68% EBITDA margin, and unchanged measured capital per sales euro. Terminal11.89x is a finite 15-year/3%nominal growth/10%required-return judgment, followed by no-growth value; no claim that this is a loss bound.
  • Base case. Regional demand bridge supports initial 10% growth fading to 9%, about EUR349m in year 5: current TTM sales allocated by FY2025 regional weights, North America15%, Europe5%, APAC/MEA10%. Steam is included in the regions. Retain the 25.7% cycle-margin recovery and measured capital intensity; growth above this needs additional identified opportunities.
  • Upside case. Faster near-term adoption begins at 17% and fades to the same 9% long-run rate. Keep 29.2% normalized EBITDA, 45% ROIIC and 20x cash NOPAT. Favorable demand does not itself establish a higher terminal multiple or a permanently longer growth runway. No M&A uplift is assumed.

Research notebook

Reviewed Nov 15, 2026

8 of 11 research checks complete. Still to complete: Security and evidence base; Opportunities and funded scenarios; Variant view and catalysts
Business model

Harvia makes sauna heaters, control units, heater stones, spare parts, sauna rooms, Scandinavian hot tubs and steam equipment. It sells mostly to retailers, distributors, wholesalers, sauna builders and DIY chains in nearly 100 countries, and directly to consumers in North America and, from 2026, Germany and Austria (harvia-annual-report-2025, pp.8 and 141). Factories: Muurame, Finland (heaters, the largest); Lewisburg, West Virginia (Almost Heaven rooms, expansion completing Q4 2026); Driedorf, Germany (EOS premium heaters and controls); Guangzhou, China; Romania; Italy (Phoenix controls); Sastamala, Finland (Kirami tubs). Money is EUR millions unless stated.

Evidence plan

Primary: financial statements bulletins FY2019–FY2025 (harvia-fsb-2019 … harvia-fsb-2025), the 2025 annual report (harvia-annual-report-2025), the Q3 2026 interim report and presentation (harvia-q3-2026, harvia-q3-2026-presentation), and the acquisition releases. Outside observations: Masco's price for Sauna360 (masco-10k-2023-sauna360), Tulikivi results (tulikivi-fy2025). No fieldwork, dealer survey or customer interview has been done.

Value-creation mechanism

Harvia turns capital into owner cash by manufacturing a small, high-margin component — the electric heater and its control unit — that goes into almost every new or renovated sauna, whoever builds the room, and then selling the room, stones, spares and accessories around that installed base through a dealer network it already owns. The step where most value is created is the heater/control sale into a fragmented, specified market: the product costs little capital to make (group materials and services were 35.8% of FY2025 revenue, so gross margin before labour was 64.2%), the installed base of over 18 million saunas generates replacement demand, and Harvia's estimated "clearly over 20%" share of heaters and components (harvia-q3-2026 p.4) lets one factory serve many channels. Growth capital (capacity and inventory) is needed, but each organic sales euro has needed about EUR0.49–0.63 of growth capital through the cycle (07-assumption-register.md, 3 November). The weak step is overhead: since 2021 the profit gap has opened in personnel and other operating costs, not in product margin.

Business-model profile

Comparable firms are chosen by these traits, not by the industry label.

Growth vehicle: Organic. Revenue moved from 179.1 in 2021 to 198.9 in 2025 through the post-pandemic downturn and recovery. Acquisitions are occasional bolt-ons for a product or region (Almost Heaven 2018, Kirami 2021, ThermaSol 2024), not a recurring capital program: acquisition cash was 5.5% of revenue over 2023–2025, almost all ThermaSol. Growth comes from capacity, distribution and price, with growth capital earning about 34% after tax across 2019–2025. The model forecasts organic growth only; ThermaSol is in the starting base and further deals are not forecast, because their timing and price cannot be estimated from the record. That omission is upside, not a base-case assumption.

Capital intensity: 4.6% of sales, asset-light. Capital spending is about 4.6% of revenue over 2023–2025, raised by the Lewisburg factory expansion (completing Q4 2026) on top of maintaining factories in Finland, Germany, China, Romania and Italy. Growth needs plant capacity and inventory (tangible capital 86.7 in 2025), which earns about 35% after tax. The share sits just under the 5% boundary, so the asset-light label is marginal.

Comparables. A comparable firm makes branded durable consumer products sold through retailers, builders and dealers, grows organically through capacity, distribution and price, buys only occasional bolt-ons, and carries moderate plant and inventory. Its revenue follows housing, renovation and discretionary-spending cycles. Not the label: "Leisure" or "building products" groups Harvia with commodity fixtures; its heaters and controls are a niche brand franchise earning about a 20% EBIT margin.

Declared in valuation/model.json (businessProfile). npm run check measures deal and capital intensity from the reported facts, and the outside view grades the forecast only against firms that share them.

Economic snapshot

Definitions. Adjusted EBIT as reported (items affecting comparability were 0.8 in 2025, 1.6 in 2024). Tax at the model's 23% normalized rate (FY2025 effective rate was 20.4%: 6.776 / 33.182). Tangible capital = PP&E + right-of-use assets + net working capital (inventories + trade and other receivables − trade and other payables); excess cash is excluded. Return on tangible capital (ROTC) capitalizes leases; return on capital employed (ROCE) here includes goodwill and acquired intangibles, which is how a buyer of the whole company pays. Sources: balance sheet and income statement harvia-annual-report-2025 pp.134–135; bulletins for earlier years.

FY2025 capital base (31 December 2025):

ItemEUR m
PP&E31.6
Right-of-use assets7.3
Net working capital (inventory 52.1 + receivables 24.6 − payables 28.6)47.9
Tangible capital86.7
Intangible assets (acquired brands, customer relationships, software)19.2
Goodwill89.3
Loans from credit institutions95.4
Lease liabilities7.5
Cash45.2 (none treated as excess while net debt is 57.7)

Returns, FY2025: adjusted EBIT 39.1 → NOPAT 30.1. ROTC on year-end tangible capital 34.7%; on average tangible capital (84.0) 35.8%. ROCE including goodwill and intangibles (195.4) 15.4%. The issuer's adjusted ROCE (pre-tax, excluding goodwill but including intangibles) was 41.3% (harvia-annual-report-2025 p.55); at 23% tax that is about 32%, consistent with our ROTC once the 19.2 of intangibles is included.

Seven-year history (PP&E + NWC as tangible capital, leases excluded before 2024 for comparability):

FYRevenueGross margin*Opex ratio**Adj. EBIT marginInventory days***Tangible capitalAfter-tax ROTCROCE incl. goodwillIncremental EBIT margin
201974.160.3%38.7%18.7%17131.434%11.7%—
2020109.161.5%37.3%22.4%18034.954%17.8%30%
2021179.160.9%31.9%26.4%24069.952%25.3%33%
2022172.459.3%35.3%21.1%23672.439%19.2%n/m (sales fell)
2023150.562.7%37.2%22.4%23163.041%19.0%n/m (sales fell)
2024175.264.3%40.6%21.2%28773.139%17.4%14%
2025198.964.2%42.0%19.6%26779.438%17.8%8%

* 1 − materials and services / revenue. ** Employee benefit expenses + other operating expenses, / revenue. *** Year-end inventory / materials and services × 365. Sources: income statements and balance sheets in harvia-fsb-2019 … harvia-fsb-2025.

What the snapshot says:

  • Product economics improved: gross margin rose from about 60% to 64% between 2019–2022 and 2024–2025, even through the 2025 tariffs and a weak dollar.
  • Overhead absorbed all of it and more. The opex ratio rose ten points from 2021 to 2025. Employee costs grew from 30.6 to 42.0 and other operating expenses from 26.5 to 41.4, while revenue grew 19.8. Q3 2026 repeated it: personnel and other costs grew faster than revenue, headcount 816 vs 728 (harvia-q3-2026 pp.1–3, 8).
  • Inventory is heavy: 230–290 days of materials cost since 2021, versus 170–180 before the pandemic. Working capital, not machinery, is the main growth capital.
  • Incremental margins fell from about 30% (2020–2021, existing overhead) to 8–14% (2024–2025, while overhead was being built). Normalizing for the unconsolidated part of ThermaSol (consolidated from 31 July 2024) moves 2024–2025 by under a point and doesn't change this.
  • Current price: at EUR39.70 (3 November 2026, harvia-q3-2026 share count) market capitalization is about 742 and EV about 803. Trailing September 2026 cash NOPAT on the engine definition ((49.6 − 1.26% × 218.9) × 0.77) is 36.1, so EV is 22× today's earning power and about 17× earning power at a 29.2% EBITDA margin.

Economic segmentation

Harvia reports revenue by region and product group but no segment profit. Profit has to be inferred from which units Harvia has bought and at what margins, and from what moved the group margin.

Economic unitFY2025 revenue (share)Demand driverEstimated EBITDA marginEstimated share of FY2025 EBITDA (46.5)Trend
Heaters, controls, IR components107.9 (54%)New saunas and replacement of the installed base; specified by builders and dealersHigh: residual after the units below, about 27%about 63%Growing; +14.6% in 9M 2026
Accessories, stones, spare parts and services26.2 (13%)Installed base usage and replacement; no new room neededHigh, similar to heatersabout 15%Growing 16–17% in 9M 2026
Saunas (rooms, cabins) and Scandinavian hot tubs47.5 (24%)New-room purchases, mostly North America (Almost Heaven, D2C and retail) and Finland (Kirami); discretionaryEstimate 15% (Kirami was bought at a 15.0% adjusted EBITDA margin, harvia-kirami-2021)about 15%Growing fastest in NA, +18.8% in 9M 2026
Steam products17.2 (9%)US residential construction and renovation; commercial projectsEstimate 15–17% (ThermaSol was bought at 17.2%)about 6%Declining: −8.2% in 9M 2026

Method: allocate the room/tub and steam units the margins at which Harvia bought comparable businesses, and attribute the rest of group EBITDA to heaters, controls and aftermarket. Heaters, controls and the aftermarket are then about two thirds of revenue and about 78% of EBITDA (36.5 on 134.1 of revenue, a 27% margin). Adverse allocation: give rooms and steam the full group margin of 23.4%; heaters and aftermarket still produce about two thirds of EBITDA. Either way, revenue share overstates the profit share of rooms and steam. These margins are estimates from acquisition disclosures, not reported segment data.

Regional view (FY2025 revenue / 9M 2026 growth): North America 75.8 (38%) / +21.9%; Continental Europe 55.6 (28%) / +5.1%; Northern Europe 46.6 (23%) / +12.8%; APAC & MEA 21.0 (11%) / +9.6% (harvia-fsb-2025 p.5, harvia-q3-2026 p.5). Growth comes from North America, where the mix is richer in rooms and steam. Northern and Continental Europe are mostly heater deliveries from Muurame, which is why the Q2 2026 IT stoppage hit those regions (harvia-q3-2026 p.5). Group growth led by North America therefore tends to dilute the margin even if each product earns what it did before.

Unit economics

Unit: one euro of additional organic sales, which is how Harvia adds value. It sells through existing dealers and factories, and one more sauna heater shipped from Muurame or room shipped from Lewisburg needs inventory, receivables and occasionally a capacity step. The P&L and capital per EUR100 of incremental sales, from the group accounts:

Per EUR100 of incremental sales2019→2021 (existing overhead)2023→2025 (building overhead)Source / basis
Materials and services3931Incremental: gross margin on total sales rose from 62.7% to 64.2% over 2023→2025
Gross profit6169
Incremental personnel and other operating cost2757ΔOpex / Δsales: 2019→2021 +28.4 / +105.0; 2023→2025 +27.4 / +48.4
Depreciation increment32ΔD&A / Δsales
Other operating income and comparability items0+1Residual to reported adjusted EBIT
Incremental operating profit3211Reported incremental adjusted EBIT margin
NOPAT at 23% tax249
Growth capital (working capital + capex above the recurring charge)4949Full-cycle measured EUR0.49 per organic sales euro, FY2019→FY2025, lagged (calculations.json capitalPerOrganicSale2019to2025)
of which working capitalabout 29about 29Working-capital investment / organic sales change, FY2019–FY2025
After-tax return on the unit's capitalabout 50%about 18%NOPAT / growth capital

Product economics got better between the two columns (incremental gross margin 69 against 61); overhead absorbed the gain and more. At the measured capital intensity, an incremental sales euro returns about 50% after tax if overhead is held, and under 20% while overhead grows at today's pace. The model's 45% organic ROIIC, using the 29.2% normalized margin as the steady state, sits close to the first column. It assumes the overhead now being built is a one-off step that later sales absorb.

The FY2021 build (capex 11.8, inventory +22.6) was put in place when organic sales rose 43%. Sales then fell 7.7% and 9.4% in 2022–2023 as dealers destocked, and the capacity looked like dead money: ROTC fell from 52% to 39–41%. Measured from FY2019 to FY2025 across that cycle, growth capital of 40.5 supported 82.6 of organic sales at the same start and end NOPAT margin (about 17%), a 34% after-tax return at actual margins (07-assumption-register.md, 3 November). The project still earned its cost of capital several times over through a full boom and bust.

Failure case: steam. Steam revenue then fell 8.2% in 9M 2026 because US residential construction and renovation, which drive steam showers, weakened (harvia-q3-2026 pp.3–5). A steam shower is part of a bathroom build. The buyer is a contractor responding to housing turnover, not a consumer adopting sauna, so the replacement-and-adoption lever behind heaters doesn't apply. The same fixed overhead earns less on it.

Independent checks

  1. Reported capital returns. Harvia's adjusted ROCE excluding goodwill was 41.3% in FY2025 and 39.9% on a trailing basis in September 2026 (pre-tax, average capital). Our after-tax ROTC of 35–36% grossed up for tax (about 46%) and including 19.2 of intangibles agrees with it. The unit model's 49% "existing overhead" return and the 34–44% measured cycle return bracket the group's tangible return, so the unit model reconciles with group accounts.
  2. Private-market price of the closest competitor. Masco paid about EUR124m net of cash for Sauna360 (Tylö, Helo, Finnleo, Kastor, Amerec) in 2023 and recorded USD60m of goodwill on USD136m (masco-10k-2023-sauna360). Sauna360's revenue isn't disclosed in the filing, so its margin can't be checked. The price shows that the second global heater platform changed hands for about one sixth of Harvia's current EV.
  3. Scale check against a sub-scale peer. Tulikivi, a listed Finnish stove and fireplace maker with a sauna-heater line, had EUR29.5m of 2025 sales and a comparable operating loss of EUR0.1m (tulikivi-fy2025). A small heater and stove maker without Harvia's dealer reach and volume doesn't earn Harvia's margins. That supports the view that Harvia's returns come from scale and distribution rather than from the product category itself. This is one observation, not a peer set.
  4. Market-share arithmetic. The issuer estimates a EUR4bn global sauna market and puts its share above 5% (harvia-q3-2026 p.4). Harvia's 2025 revenue of 198.9 is 5.0% of 4bn, consistent with the estimate; its heater and component share is "clearly over 20%". The estimate is management's, not independent.

Adverse allocation: assign all of the 2023–2025 opex increase to the growth program and none to inflation; incremental NOPAT is then 9 per 100 against 49 of capital, an 18% return. That still beats the cost of capital, but it is less than half the model's 45%. The conclusion that growth creates value survives. The conclusion that it creates value at 45% depends on overhead being absorbed.

Where the consensus differs. The issuer's own long-term target is an adjusted EBIT margin "exceeding 20%" (harvia-q3-2026 p.2), about 24% EBITDA at today's D&A. The model's 29.2% EBITDA is about 25.5% adjusted EBIT, achieved only in 2021. The observation that would settle it is the FY2026 and FY2027 opex ratio: if personnel and other operating costs grow more slowly than revenue once Lewisburg and the IT platform are complete, the 29.2% path is plausible. If the ratio stays near 42%, the unit return is about 20%.

History, market & cycle

Harvia was a Finnish heater maker (founded 1950, Muurame), listed on Nasdaq Helsinki in March 2018. It became an international sauna company through acquisitions: Sentiotec (Austria, 2016), Almost Heaven (US rooms, December 2018), EOS (German premium heaters, controls and steam, April 2020), Kirami and Sauna-Eurox (Finland, 2021), Phoenix El-Mec (Italian controls, 2023) and ThermaSol (US steam, July 2024). Russia was exited in 2022. Matias Järnefelt replaced Tapio Pajuharju as CEO in 2023. The record that follows separates structural adoption from the pandemic pull-forward and the dealer destocking that followed. Money is EUR millions.

Demand drivers and cycle

Who decides. The end user is a homeowner, or a hotel, spa or gym. The payer is the homeowner, either directly (D2C, retail, DIY chains) or through a builder or renovator. The heater brand is often chosen by the dealer, sauna builder or electrician who specifies it, and the consumer chooses the room. Harvia sells mostly to retailers, distributors and builders on contracts that are usually annual (harvia-annual-report-2025 p.141). That gives dealers timing power: they can destock without losing the brand.

Demand split (judgment from the product and regional data; the issuer doesn't split demand this way):

  • Replacement and aftermarket. A heater, its elements and stones wear out in an installed base the issuer puts at over 18 million saunas, more than half in Finland, Germany, Russia and the US (harvia-q3-2026 p.4). Spare parts, services, accessories and stones were 26.2 (13%) in 2025 and grew 16–17% in 9M 2026. Replacement is the least cyclical demand and needs no new room.
  • Growth (adoption). New saunas in North America and APAC, driven by health awareness rather than consumer confidence, according to management. North American revenue rose from 62.0 (2024) to 75.8 (2025), +21.9% in 9M 2026, while US consumer confidence and residential construction were weak (harvia-q3-2026 pp.3–4). This is most of the incremental demand and the most uncertain part.
  • Regulation-driven. Little demand is created by rules, but certification is a barrier to entry: a heater sold in the US needs ETL/UL listing. Tulikivi announced US ETL approval for its electric heaters in 2026 (tulikivi-us-etl-2026), which shows the barrier can be crossed. Product-safety recalls (Sauna360's 2026 infrared recall, sauna360-cpsc-2026) show that regulators can hit a competitor's channel.
  • Discretionary. Rooms, outdoor saunas, hot tubs and steam showers are big-ticket purchases tied to renovation and housing. Steam fell 8.2% in 9M 2026 with US residential construction (harvia-q3-2026 p.5).

The measured cycle (issuer organic growth at comparable currency; harvia-fsb-2020 … harvia-fsb-2025, harvia-q3-2026):

YearOrganic growthRevenueAdj. EBITAdj. EBIT marginInventory (year end)What happened
2020+28.0%109.124.422.4%20.7Pandemic home-improvement surge; EOS acquired
2021+43.3%179.147.326.4%46.1Peak; capacity and inventory build
2022−7.7%172.436.521.1%45.3Dealer destocking; Russia exit; energy price spike in Europe
2023−9.4%150.533.722.4%35.5Trough; inventory released 10.1 of cash
2024+12.9%175.237.121.2%49.2Recovery led by North America (+62.7% in Q4); ThermaSol
2025+14.4%198.939.119.6%52.1All regions grow; Northern Europe back to growth after two years of decline
9M 2026+15.9%165.231.018.7%—North America +21.9%; Q2 IT stoppage shifted about 4 of deliveries to Q3

Peak to trough (2021→2023), revenue fell 16% and adjusted EBIT 29%, but the margin never went below 21% and operating cash flow peaked in 2023 (39.1) as inventory was released. This was a two-year organic decline of about 16%, not a collapse. By 2025 organic sales were about 8% above the 2021 peak (1.000 × 0.923 × 0.906 × 1.129 × 1.144 ≈ 1.08), and 2026 added another 16% at comparable currency.

The market misperception to test is that the sauna boom was a pandemic fad. The organic index above refutes it for the group: the base exceeded its 2021 peak within four years. The right series to keep testing is North American organic growth against US residential construction. If NA sauna demand follows housing, the "health adoption" explanation is wrong. So far it has diverged: rooms and heaters grew while steam, the housing-linked product, fell.

Seasonality matters for reading quarters: Q3 is the weakest cash quarter because working capital is built before winter (operating free cash flow −0.1 in Q3 2026).

Competition

Where competition happens:

  • Premium and professional (Germany, spas, hotels). EOS competes with Klafs and others in commercial projects. These are project sales, which are lumpier (Middle East projects were postponed in 2026).
  • Rooms and outdoor saunas (North America). Almost Heaven competes with Finnleo (Sauna360), many infrared brands and imported barrel saunas. Competition here is on price, delivery and online reach; about 70% of what Harvia sells in the US is made there, which shields it from tariffs relative to importers (harvia-q3-2026 p.11). Campaign pricing is visible (almost-heaven-black-friday-2026), so this is the most price-competitive unit.
  • Steam (US). ThermaSol competes in a contractor channel tied to bathroom renovation; demand is weaker.

Trend in the number of competitors: consolidation at the top (Masco–Sauna360 in 2023; Harvia's six deals), fragmentation at the bottom (imported rooms; new US entrants in heaters).

Advantage in one sentence: Harvia makes the highest-volume heater range in the industry in its own factories, sells it through the widest dealer network, and builds most US product in the US, so it has the lowest unit cost and the fastest availability in the component dealers specify most. Evidence that it lasts: gross margin rose from about 60% (2019–2022) to 64% (2024–2025) through tariffs, the weak dollar and campaign pricing, so it isn't buying volume with price. Sub-scale Tulikivi lost money in 2025 (tulikivi-fy2025). The share estimates rose from 13% (Harvia alone, 2020) to over 20% (2025).

The competitor who could hurt: Masco, through Sauna360. It owns the second heater platform, a US plumbing and wellness channel (Watkins hot tubs), and a balance sheet many times Harvia's. It could cut heater prices in North America or bundle saunas with hot tubs through Watkins dealers. Why it hasn't (judgment): Sauna360 is a small bolt-on inside Masco's plumbing segment, bought to enter a category, and still being integrated into Watkins. The 2026 Sauna360 infrared recall points to execution problems, not a price campaign. Harvia's rising gross margin is the observation that would change first if Masco started a price war.

Counterparty incentives

CounterpartyWhat they wantWhat Harvia gives themTheir outside optionsWhat would change the arrangement
Largest customer (about 11% of 2025 net sales, 9% in 2024; not named, harvia-annual-report-2025 p.141)Margin per shelf or web listing, sell-through, campaign support, reliable US supplyA brand consumers search for, US-made product (tariff-safe), Black Friday campaign volumeSauna360/Finnleo, Huum, imports, private labelA private-label or alternative-supplier program; a demand for more promotional funding. Dependence runs both ways: one customer at 11% is a risk, but the concentration is also how Harvia reaches US consumers at scale
Dealers, distributors, sauna builders (many, annual contracts)Availability, margin, a brand that installers trust, low warranty returnsThe full range (heater, controls, room, stones, spares) from one supplier; delivery speedSwitch heater brand at the next contractA prolonged delivery failure (the Q2 2026 Muurame IT stoppage postponed about 4 of deliveries; recovered in Q3) or a better-margin rival. Dealers showed their power in the 2022–23 destocking
Consumers (D2C)Price, delivery, health experienceAlmost Heaven brand, direct web salesInfrared and imported roomsPrice competition in rooms; fading health narrative
Suppliers (steel, electronics, transport)VolumeScale purchasingManyNo single-source dependence is disclosed; Harvia says it can shift production between factories and change suppliers for many key materials (harvia-q3-2026 p.11). Unlike TerraVest's boiler business, no supplier controls a key component
Regulators (US safety certification, CPSC, EU rules)Safe productsCertified products—A heater or room safety recall at Harvia would damage the dealer channel the advantage rests on
Shareholders———No controlling holder: the largest, Onvest Oy, held 4.43% at end-2025; ten largest registered holders 27.4% in September 2026 (harvia-q3-2026 p.8). No one can force a sale or block a strategy

Read in both directions, the main dependence is on dealers and the largest retailer, and it is mutual: they need the category leader's range and US supply, and Harvia needs their shelves. Nothing found suggests a counterparty is positioned to take Harvia's share of value. Sources: as cited; channel structure from harvia-annual-report-2025 pp.8 and 141.

Management & incentives

Capital allocation is decided by a seven-member board with no controlling shareholder and a CEO hired from outside in 2023. Several long-serving managers from the pre-IPO team are still in the management team and hold most of the insider stake. Judge capital allocation by the cash returns of completed investments and by what the next factory or acquisition costs; a high group ROCE doesn't establish either. Sources: harvia-annual-report-2025 (governance statement and remuneration report, pp.29–52 and 130), harvia-q3-2026 pp.9–10, harvia-manager-lengauer-2026-10-30. Values at EUR39.70 (3 November 2026).

Ownership and incentives

Holdings at 31 December 2025, including controlled companies (harvia-annual-report-2025 pp.32–39, 130):

PersonRoleSinceHarvia sharesValue at EUR39.70Notes
Matias JärnefeltCEO202300Earned 8,032 gross LTI shares for 2023–2025, delivered net of tax in March 2026; no open-market purchase disclosed
Ari VesterinenCFO2014151,6506.0mPre-IPO management
Markus WörmansederHead of Products and Solutions2017106,9604.2m
Timo HarviaHead of Innovation and Technology2014103,3164.1mFounding family; also represents Tiipeti Oy (342,790 shares, 1.83%) on the Nomination Board, not disclosed as a controlled company
David AhonenHead of APAC & MEA201482,9833.3m
Anssi PelkonenHead of Northern Europe201463,0522.5m
Mika SuojaHead of Operations201635,0001.4m
Päivi JuolahtiHead of Marketing and Brands20203,3710.1m
Nathan Hagemeier, Ivan Sabato, Philipp KrauthHeads of North America, Continental Europe; EOS2024–2500New regional leaders hold nothing
Olli LiitolaDirector; Chair 2014–2024201442,1741.7m
Hille KorhonenDirector20215,4070.2m
Heiner OlbrichDirector; Chair until Nov 202520223,0000.1mLeft the board at the 2026 AGM
Catharina Stackelberg-HammarénChair from Nov 202520231,8360.1m
Anders HolménDirector20211,9860.1mManaging Director, Fyrklöver-Invest
Markus LengauerDeputy Chair from April 20262024880 + 660 bought 30 Oct 2026 at EUR38.100.1mThe only open-market purchase found in 2026
Petri CastrénDirector, audit chair20253010.0mCFO of Kemira
Board, CEO and management team total604,076 (3.2%)24.0m507,816 (2.7%) a year earlier

Pay. CEO total remuneration paid in 2025 was EUR839,153: fixed salary 566,545 (up 11%, partly because a Finland-wide personnel bonus was converted into base salary), 2024 short-term incentive 243,556 and a 29,052 performance bonus. The 2025 STI earned 213,717 (54% of maximum). STI: up to 70% of fixed salary, 70% on consolidated adjusted operating profit and 30% on personal targets (2025: steam business, Continental European D2C, infrared strategy). LTI: target 70% and maximum 140% of fixed salary in shares; criteria for 2024–2026 and 2025–2027 are absolute TSR 50%, revenue growth 40% (payable only above an undisclosed EBIT-margin threshold) and CO2 reduction 10%. The plan has no holding period or shareholding requirement (harvia-annual-report-2025 pp.46–52). Board: chair EUR63k, others EUR49k on average in 2025 (up 30%), about 40% of monthly fees paid in shares.

Whose interests dominate:

  • CEO: stake EUR0, annual pay EUR0.84m. If the stock compounds 10% a year, his maximum LTI shares per plan (11,894) are worth about EUR0.5m at today's price, roughly 60% of one year's cash pay. Pay, which grows with revenue growth and adjusted operating profit, dominates his economics. The TSR leg is the only per-share measure.
  • Long-serving managers (CFO, Wörmanseder, Timo Harvia, Ahonen, Pelkonen): EUR2.5–6.0m each. At 10% compounding the CFO's stake gains about EUR0.6m a year, probably more than his salary (not disclosed individually). Their interests are closer to owners'.
  • Board: small holdings; the chair holds about EUR0.1m. Shareholders discipline the board through the Nomination Board of the four largest holders (2026: representatives including Nordea, Evli and Onvest), not through director ownership.

No compensation metric measures return on capital, cash conversion or per-share value other than TSR. Revenue growth and adjusted operating profit can both be raised by spending: overhead to grow sales, or acquisitions bought with debt.

Track record and references

Two eras.

Pajuharju era (CEO to May 2023; IPO 2018). Revenue 74.1 (2019) to 179.1 (2021) through Almost Heaven, EOS and Kirami and the pandemic surge. Measured outcomes: Almost Heaven, taken over in December 2018 without cash consideration, became the core of a North American region with 75.8 of 2025 revenue. Kirami cost 9.5 in total for 22.1 of revenue and 3.3 of adjusted EBITDA (about 2.9×). EOS cost 19.7 for 78.6% in 2020 on a 25.5 debt-free value (17.3 of 2019 revenue), but the 21.4% minority was then bought in July 2022 for 19.0 at the same EBITDA multiple. That implies a 100% value about 3.5× the 2020 valuation, set on boom-period EBITDA just before the downturn (harvia-fsb-2020, harvia-fsb-2022). Inventory was built to 240 days in 2021 and then released through 2023. Russia (EOS Russia and Harvia operations) was exited in 2022 with the brand rights retained. Adjusted EBIT margin peaked at 26.4% in 2021. Pajuharju still held 139,000 shares at end-2025 (harvia-annual-report-2025 p.130); his sales history wasn't examined.

Järnefelt era (from 2023). Revenue 150.5 (2023) to 218.9 trailing September 2026 (+45%, of which ThermaSol about 13). Adjusted EBIT margin 22.4% (2023) → 19.6% (2025) → 18.7% (9M 2026). Adjusted ROCE excluding goodwill 54.5% (2023) → 41.3% (2025) → 39.9% (September 2026). Headcount 605 → 816. Decisions:

  • Worked: North American growth (+62.7% in Q4 2024, +22.1% in 2025, +21.9% in 9M 2026); APAC (China, Japan, Australia); Northern Europe returned to growth in 2025; gross margin rose to 64%.
  • Failed or not yet proven: ThermaSol (about 11–13× EBITDA in 2024, steam −8.2% in 9M 2026); the Muurame IT/ERP cut-over (about EUR4m of Q2 2026 deliveries postponed); a North American revenue-recognition correction in Q3 2026 (−0.7 revenue, −0.8 EBIT); North America head Jennifer Thayer resigned in May 2025, interim cover until Nathan Hagemeier joined in November 2025; opex outgrew revenue in 2024, 2025 and 9M 2026.
  • Shareholder return: from EUR17.68 at end-2022 to EUR39.70 plus about EUR2.84 of dividends paid 2023–2026, about 26% a year. Most of it is re-rating from a trough (EV about 12× cash NOPAT at end-2022 against 22× today), not per-share earnings growth: EPS rose from EUR1.25 (2023) to EUR1.41 (2025).

Since end-2019 (EUR10.45): price ×3.8 plus about EUR4.3 of dividends, about 23% a year over 6.8 years.

References. None collected. No former employee, dealer, competitor or customer has been interviewed; the issuer's annual customer and employee surveys (cited in harvia-fsb-2025) are self-reported. Bench depth: the CFO and four other managers have 9–12 years in the management team; the North American and Continental European leadership changed in 2025. This is a gap in the research.

Insider trades. The only 2026 open-market trade found is Lengauer's purchase of 660 shares at EUR38.10 on 30 October 2026 (harvia-manager-lengauer-2026-10-30), about EUR25,000, the day after the Q3 fall. That is too small to signal much. Management sales over the full period weren't reviewed (open check).

Capital allocation principles

Governing objective as stated: average annual revenue growth of 10%, adjusted operating margin above 20%, net debt/adjusted EBITDA below 2.5×, and a regularly increasing dividend paid twice a year (harvia-q3-2026 p.2). These are growth and margin targets, not per-share value or return-on-capital targets.

  1. Zero-based allocation. Mixed. Capital has moved toward North America (Lewisburg expansion, ThermaSol, D2C, IT), and Harvia exited Russia (2022) and sold its Estonian operations (2025). But the 2023–2026 overhead build was allocated across all regions at once, and no business has been shrunk to fund another.
  2. Funding strategies, not projects. Partly. The four strategic priorities (full sauna experience, strategic markets, key channels, operations) are funded as a program; no project-level returns are reported. The cost of the program shows up as opex, so it bypasses capital-budgeting discipline.
  3. Capital has a cost. Share compensation is small: about 79,000 maximum LTI shares per plan (0.4% of shares), settled from treasury shares bought in the market (17,000 at EUR39.01 in February 2026), so dilution is offset with cash, a financing cost the cash flow statement records as a buyback. Debt is used for acquisitions (EOS 2020, EOS minority 2022, ThermaSol 2024: each year net borrowing was about 20). Leverage 1.2× against a 2.5× ceiling.
  4. Zero tolerance for bad growth. Weak evidence. Opex was cut in the 2022–23 downturn (60.9 → 56.0), so management can retrench. Since then three years of opex growing faster than revenue haven't been reversed, and ThermaSol hasn't been reported on separately.
  5. Price versus value. Passive. No buyback at the 2022 trough (EUR17.68, about 12× cash NOPAT), when the stock was cheap against the business; buybacks are only for incentive delivery. No equity was issued at the 2021 peak (EUR58.70, 29×). The 2022 EOS minority purchase followed a pre-agreed option formula at a boom multiple, so price discipline was given up in advance.

Does management behave as if long-term value per share is the goal? The dividend has risen every year since the IPO and acquisitions have mostly been small and cheap (EOS majority, Kirami). But incentives and stated targets point at growth and margin, and the current program is buying growth with overhead at a lower incremental return than the business historically earned (01-business.md, unit economics). The record so far is "good business, competent operators, unproven allocators of the new overhead".

Integrity and governance

Evidence:

  • Board independence: all seven directors elected in 2025 were assessed independent of the company and of its major shareholders (harvia-annual-report-2025 pp.32–33). One share class; no dual-class or super-voting control.
  • Ownership is dispersed: the ten largest registered holders owned 27.4% in September 2026; nominee and foreign holders 38.4%; households about 19% at end-2025. A Shareholders' Nomination Board of the four largest holders proposes directors.
  • Adjustments: items affecting comparability were small (0.8 in 2025, 1.6 in 2024; 0.3 in 9M 2026), mostly transaction and restructuring costs. No pattern of adding back recurring costs.
  • Related party: the 2022 EOS minority was bought from EOS's managing director, a management-team member, at a pre-agreed formula; disclosed.
  • Pay: CEO fixed salary rose 11% in 2025 and the board's average fee 30%; the 2025 STI paid 54% of maximum with operating profit below target. The remuneration report was approved in an advisory vote at the 2025 and 2026 AGMs (vote counts not reviewed).
  • Controls: the Q3 2026 North American revenue-recognition correction (EUR0.7m revenue) is small but is a control finding in the fastest-growing region, one quarter after the Muurame IT cut-over. Deloitte is the auditor.

Agency costs (Mauboussin and Callahan pp.47–48):

  • Control of resources: present. Revenue growth is 40% of the LTI and operating profit 70% of the STI, with no return-on-capital measure, while the CEO owns no shares. This rewards growth spending and acquisitions whether or not they earn their cost. It is the agency cost that matters here.
  • Risk tolerance: low concern. Leverage is modest and the CEO's undiversified exposure is pay, not stock.
  • Time horizon: LTI periods are three years with no holding requirement, shorter than a five- to ten-year investment horizon.

Valuation treatment: the model charges all opex as reported (no add-back of growth spending) and gives no value to future acquisitions. The agency cost appears as the risk that the 29.2% margin recovery never comes because spending keeps rising with revenue. Since 3 November 2026 the base and downside hold the margin at the 25.7% cycle average, which matches management's stated reinvestment policy; only the upside assumes the 29.2% normalization (05-thesis.md, 06-risks.md).

Capital allocation

Where Harvia's cash went over FY2021–FY2025, what each use earned, and how the model funds growth. Money is EUR millions. Facts are in financials/facts.json (cash-flow statements of harvia-fsb-2021 … harvia-fsb-2025); arithmetic is in valuation/calculations.json.

Sources and uses of cash

FY2021FY2022FY2023FY2024FY2025Five years
Cash flow before working capital (pre-tax)51.242.439.642.646.4222.3
Working capital (− = investment)−20.9−5.4+7.8−2.9−5.2−26.6
Income taxes and operating interest−8.5−12.7−8.3−8.0−6.2−43.7
Net cash from operating activities21.824.339.131.735.0151.9
Capex net of asset sales−11.7−3.5−3.0−6.1−14.3−38.7
memo: D&A5.86.56.37.07.433.0
Lease repayments−0.4−0.5−0.8−0.9−1.0−3.6
Interest paid (net of interest received in investing)−1.9−1.0−2.9−1.8−4.2−11.8
Free cash flow to equity before acquisitions7.919.332.422.815.497.8
Acquisitions net of cash (and divestments)−7.6+0.1−2.8−24.8−1.2−36.3
EOS minority purchase—−19.0———−19.0
Dividends−9.5−11.3−12.0−12.7−14.0−59.5
Treasury share purchases (for incentive delivery)−2.5−0.3———−2.8
Net borrowing (excluding leases)−0.4+20.9−1.9+20.1−0.3+38.4
Net change in cash−12.1+9.6+15.7+5.3−0.0+18.5
FX on cash+0.3+0.2−0.4+0.5−1.2−0.6
Cash at year end15.525.340.646.445.227.3 → 45.2

Each column reconciles to the reported net change in cash within rounding. FCF to equity is the same definition as the ownerCashPerShare history series.

What the table shows about priorities:

  • The dividend comes first and rises every year: 59.5 over five years, 61% of FCF to equity, paid through the 2022–23 downturn.
  • Acquisitions (55.3 including the EOS minority) were funded with debt: net borrowing of 38.4 came in 2022 (EOS minority) and 2024 (ThermaSol).
  • Organic growth capital appears inside operating cash: net capex was 38.7 against 33.0 of D&A, and working capital absorbed 26.6, of which 20.9 was the 2021 inventory build.
  • There have been no buybacks for value. Treasury purchases only deliver incentive shares.
  • Cash built from 27.3 (end-2020) to 45.2 while gross bank debt rose to 95.4. Net debt was 57.7 at end-2025 (1.2× adjusted EBITDA).

Investment rates per sales change

Window FY2019→FY2025: reported sales +124.8, of which organic +82.6 (issuer organic growth rates; acquired businesses count as organic after twelve months). Rates are computed in valuation/calculations.json.

RateDefinitionFY2019→FY2025Step
Incremental M&A rateAcquisition cash ÷ Δsales43.7% (58.9% incl. the EOS minority)acquisitionRate2019to2025
Expensed growth (intangible) rate, upper boundΔ(personnel + other operating cost) ÷ Δsales43.9%expensedCostRate2019to2025
— of which FY2023→FY2025same56.7%expensedCostRate2023to2025
Fixed capital rate(Capex − D&A) ÷ Δsales3.1%fixedCapitalRate2019to2025
Working capital rateCash-flow working-capital investment ÷ Δsales19.4% (29.2% of organic Δsales)workingCapitalRate2019to2025

Where Harvia really invests:

  • Not in fixed assets. Capex barely exceeds D&A over the window, though D&A includes acquired-intangible amortization and right-of-use depreciation. Excluding those, organic fixed investment is about 0.20 per organic sales euro (0.49 total growth capital less 0.29 of working capital).
  • In working capital: 0.29 per organic sales euro, because inventory runs at 230–290 days of materials.
  • In acquisitions: 44 cents per sales euro.
  • Increasingly in overhead. Over FY2019→FY2025, personnel and other operating costs rose 54.8, 44% of the sales increase. Under the current program (FY2023→FY2025) they absorbed 57% of each incremental sales euro, against 27% over FY2019→FY2021. R&D expensed is only 1.6–2.1 a year (harvia-annual-report-2025), so the growth investment is mainly selling, marketing, IT and regional organizations, expensed as incurred. This rate includes inflation and acquired overhead (ThermaSol), so it is an upper bound on investment. It is still the main inflection point in the record.

Who was in charge when the pattern changed. The working-capital surge (2021) and the debt-financed EOS minority (2022) were under Tapio Pajuharju. The overhead build (2023–2026), ThermaSol (2024), Lewisburg and the IT platform (2025–2026) are under Matias Järnefelt.

Reinvestment and returns

Owner earning power on the model's cash definition ((adjusted EBITDA − 1.26% × revenue) × 0.77): 12.4 (FY2019) → 33.9 (FY2025) → 36.1 (trailing September 2026). The 2.7× rise came from acquisitions, organic growth and FY2019's lower margin. Start and end NOPAT margins are almost equal (16.7% vs 17.0%).

Reinvestment rate. Organic growth capital FY2019–FY2024 (one-year lag) was 40.5 against 165.6 of matching cash NOPAT (FY2019–FY2024): 24%. Acquisitions (73.6 of cash including the EOS minority) take total reinvestment to 69%. Dividends took 56% of FCF to equity over FY2019–FY2025 (73.6 of 132.0).

Incremental returns (3 November 2026 filing test; table in 07-assumption-register.md):

  • Full cycle FY2019→FY2025, lagged: EUR0.49 of growth capital per organic sales euro, which is 44% in engine units (21.5% steady cash margin) and 34% at actual margins (measuredRoiic2019to2025).
  • FY2020→FY2025: EUR0.63, 34% in engine units at the 29.2% margin (measuredRoiic2020to2025). The downside uses the same EUR0.63 at its 25.7% margin: 29.8% (downsideRoiicAtCycleMargin).
  • Three-year windows swing from −7% to 67% with utilization. They measure operating leverage, not durable capital efficiency, and aren't used.

Operating leverage versus capital efficiency. Incremental adjusted EBIT margins were 30–33% in 2020–2021 and 8–14% in 2024–2025 (01-business.md). Capital per sales euro hasn't worsened. What changed is the operating leverage, through overhead. The 45% base ROIIC is a statement about capital, and the measurements support it. The 29.2% margin is a statement about overhead, and the recent record doesn't (06-risks.md).

Identity check (growth ≈ reinvestment rate × return on new capital). 24% × 34–44% predicts 8–11% organic growth a year. The geometric mean of reported organic growth FY2020–FY2025 is 12.1%. The gap of 1–4 points is explained by acquired businesses' growth counting as organic after year one (EOS, Kirami and ThermaSol's later growth was bought with acquisition capital, not organic capital) and by 2020–2021 demand filling existing capacity. The identity roughly holds, and it says the measured organic ROIIC is, if anything, slightly flattered.

Mix and return by use:

  • Organic: 34–44% after tax at full cycle (above).
  • Acquisitions: see below. Kirami is the clear success on price (about 2.9× EBITDA); ThermaSol (about 11–13×) is the weakest so far.
  • Payouts: 56% of FCF to equity over FY2019–FY2025 (61% over FY2021–FY2025); no buybacks for value.
  • Capacity projects: the FY2021 build earned its cost through the 2022–23 bust (stage 3 cohort, 01-business.md). Lewisburg's cost isn't separately disclosed and its return is unmeasured.

Acquisition record

DealDateStakeCash priceTarget revenue at purchaseTarget EBITDA / multipleLater results
Sentiotec (Harvia Austria)Nov 2016100%n/a (pre-IPO)n/an/aControls and heaters; now part of Continental Europe
Almost Heaven Saunas (US rooms)Dec 2018business assetsNo cash consideration (harvia-fs-2018 p.28)2017 c.USD9mn/aBecame the base of a North American region with 75.8 of 2025 revenue (incl. ThermaSol)
EOS Group (Germany premium/pro heaters, controls, steam; Russia)Apr 202078.6% Germany, 80% Russia19.7 (cash flow 18.1 net)17.3 (2019)Debt-free value 25.5 for 100% = 1.5× sales; EBITDA not disclosedRussia sold Nov 2022. Revenue per issuer increased particularly in heaters and controls
EOS minorityJul 202221.4%19.0—Same EBITDA multiple as 2020 → implied 100% value about 88.8, about 3.5× the 2020 valuePrice fixed on boom EBITDA; debt-funded
Kirami (Finland hot tubs)May 2021100%7.0 + 2.5 earn-out (paid Dec 2023) = 9.522.1 (LTM Apr 2021)3.3 adj. EBITDA (15.0%) → 2.9×Hot tubs and saunas group +1.6% in 2025, +18.8% in 9M 2026 (not separable)
Sauna-Eurox (Finland stones)Aug 2021100%part of 7.6 of 2021 acquisition cashc.3.2n/aStones and accessories +8.3% in 2025
Phoenix El-Mec (Italy controls)2023100%c.0.3 (2023 acquisition cash 2.8 less the 2.5 Kirami earn-out)c.1.4n/aSmall
Not itemized2025—1.3——Estonia operations sold in 2025

Aggregate (FY2020–FY2025): 73.6 of cash (including the EOS minority) for about 57 of acquired revenue: EOS 17.3, Kirami 22.1, Sauna-Eurox 3.2, Phoenix 1.4, ThermaSol 13.3. That is about 1.3× sales. Known EBITDA at purchase: Kirami 3.3 and ThermaSol 2.3. EOS's isn't disclosed, so an aggregate EBITDA multiple and a matched multi-year deal yield can't be computed. The bounds that can be stated:

  • Price paid per sales euro (1.3×) is well below Harvia's own EV/sales today (about 3.7×).
  • The acquired units are mostly lower-margin categories (rooms, tubs, steam) than heaters; the largest by margin (EOS) is undisclosed.
  • Group adjusted EBITDA margin at the window's start (2019: 22.9%) and end (2025: 23.4%) is similar. Integration didn't lift the group margin, and it didn't dilute it either.

Deployment pace: about 12 a year on average (73.6 / 6), lumpy: three years with near-zero deals. The model assumes no future acquisitions and no acquisition benefit. Management says it is "prepared to act swiftly" on M&A (harvia-fsb-2025 p.3); debt headroom to the 2.5× ceiling is roughly 63 at trailing adjusted EBITDA (2.5 × 49.6 − 60.8).

How the model funds growth

The model estimates incremental capital as incremental revenue × normalized cash-earnings margin ÷ organic ROIIC. Working capital is funded first within that requirement; growth capex is the remaining positive amount, so working capital isn't charged twice. If working capital alone exceeds the requirement, it is still fully funded. Base and upside use 45%. The downside keeps the measured FY2020–FY2025 capital intensity (EUR0.63 per organic sales euro), which is 29.8% ROIIC at its 25.7% cycle margin (07-assumption-register.md).

Recurring capital (maintenance, right-of-use and capitalized development cash) uses the source's FY2026 EUR2.9m / EUR230m sales ratio (1.26%) as a company-specific scaling estimate, not a generic maintenance percentage. Reported depreciation is kept separately for the tax and FCF bridge.

Dividends start at the known EUR0.77 and grow 5% a year. No acquisition benefit and no buyback accretion are assumed. Cash left after investment, financing and dividends reduces net bank debt, and accumulated cash enters terminal equity once. Lease cash is already charged in recurring capital, so lease debt isn't deducted again. Measured diagnostics: 10-capital-efficiency.md.

Thesis and valuation

Assessment:5 November 2026. Replacement heaters and controls support recurring demand; North American sauna adoption adds growth. Distribution, availability and manufacturing capacity matter more than exclusive technology. Growth still needs working capital and productive capacity, while personnel and other overhead have outgrown sales. A good product franchise does not by itself establish the speed of adoption or margin recovery.

The current scenarios are a separately dated judgment, not a newly attributed author forecast.

Valuation summary

Quote EUR 40.65, market time 2026-11-04T16:29:38.000Z; valuation date 5 November 2026. Fixed original exits:26 October 2031/2036. Current reported anchors, normalized earning power and forward estimates remain separate.

HorizonBearBaseBullWeighted cash-flow IRR
5 years-6.47%14.01%22.79%12.45%
10 years0.93%13.28%18.37%12.53%

Base initial growth is 10%, fading to 9% over ten years, supporting EUR348.956m sales in year 5. FY2025 regional weights are explicitly allocated to current TTM sales: North America15% growth, Europe5%, APAC/MEA10% produces approximately EUR348.5m after five years. Steam overlaps those regions. Base EBITDA recovers to the 25.7% cycle average; upside alone retains 29.2%. The downside uses 0→5% growth and current 22.683% EBITDA, with measured capital per sales euro unchanged.

Pre-mortem

The main failure is mistaking a plausible mechanism for an evidenced quantity. Adoption could fade while permanent overhead remains; low maintenance estimates and incentive-driven expansion could overstate owner cash. A supplier/competitor price response can destroy the assumed margin and a bad acquisition can consume cash already needed for growth. The adverse scenario represents a specific failure path, not a maximum loss or a frequency estimate. The independent advocates’ strongest surviving attacks are recorded in 12-debate.md.

Exit multiple

The live base/upside use 20× cash NOPAT, after recurring economic capital and operating tax, with debt/claims deducted once. Installed-base replacement, distribution and high incremental productivity support a conditional premium. An illustrative finite calculation requires roughly 12 postexit years at 9% cash growth, 38.5% productivity and 9% required return; that duration is uncertain, not measured. Historical trading multiples are context rather than proof. The adverse 11.8947× is calculated from 15 finite postexit years at 3% nominal growth, 10% required return and 26.0803% productivity, then no-growth value. Its exact annual cash receipts and capital costs are in the model-change record.

Bear case

North American adoption slows, dealers destock and overhead fails to absorb. Flat initial sales gradually recover toward 5% growth while EBITDA stays at the reported 22.68% margin. Capital per incremental sales euro remains EUR0.6325, so model productivity becomes 26.08%; reducing the margin does not silently make growth cheaper. There are no acquisitions. Product recalls, aggressive price competition or a debt-funded bad acquisition can produce losses outside this representative case.

Variant view

The opportunity-backed base no longer requires unidentified sales beyond the regional ledger. Product capital efficiency is supported, while the recovery of overhead and long-run adoption remain contested. At the saved price, a14.01% five-year base offers less margin of safety than the previous 17.81% estimate, and the revised adverse result is negative. There is no hard rerating catalyst or basis to assume the market simply misunderstands the business. Evidence of slower overhead growth and funded regional adoption is needed to strengthen the view.

Why the opportunity exists

The following liquidity/holder observations retain their original 3 November measurement window; they are context, not a new float or short-interest survey.

  • Size and liquidity: market capitalization about EUR742m. On Nasdaq Helsinki, 9M 2026 trading was EUR260.6m and 6.63m shares, about EUR1.4m or 35,000 shares a day (harvia-q3-2026 p.8). Shares also trade on Cboe, Aquis, the London Stock Exchange and ITG Posit, which isn't counted here.
  • Float turnover: excluding the two most recently flagged 5%+ holders (Alecta 5.08%, April 2025; Nordea Funds 5.02%, August 2025) and insiders (3.2%), about 16.2m shares float, about 460 days of Helsinki volume. Liquid enough for institutions; not neglected.
  • Holders: 33,504 registered shareholders; households about 19% (end-2025); nominee and foreign 38.4% (September 2026, down from 47.7% a year earlier). Domestic funds and pension insurers dominate the register.
  • Coverage: covered by Nordic sell-side and commissioned research (Inderes publicly); the full analyst list and consensus weren't retrieved. Short interest wasn't checked (FIN-FSA register).
  • Recent events that drove sellers: the 29 October 2026 Q3 report (adjusted EBIT margin 17.7% vs 19.1%; North American revenue-recognition correction; Q2 delivery catch-up) cut the price from EUR44.05 to EUR38.15 in a day. The year's range was EUR31.05–47.65.

Neglect or justified discount? Neither: a widely held, liquid Nordic small-cap at a multiple near its own median. The Q3 drop is a short-term seller event on a real issue (overhead), not evidence of mispricing by itself.

What would prove us wrong

ObservationRefutesWhere and when
Opex (personnel + other operating) grows faster than revenue in FY2026 and again in 2027Any margin recovery above the 25.7% cycle average (the upside path)FY2026 results 10 February 2027; Q1–Q3 2027 interim reports
Trailing adjusted EBITDA to September 2027 below 56 (downside band; base 60.4)Margin recovery and base growthQ3 2027 report, already a frozen prospective check
Gross margin (1 − materials/revenue) falls below 62%The "overhead, not product" diagnosis; points to price competitionEach annual income statement
North American organic growth below 10% for two consecutive quarters while US housing is flatThe health-adoption explanation of NA demandQuarterly regional revenue tables
Largest customer share falls sharply or a private-label sauna program appearsCounterparty incentives (stage 4)Annual report note 2.1; retailer announcements
Another acquisition above about 10× EBITDA, or one funded beyond 2× leverageCapital-allocation disciplineStock exchange releases
Working capital per organic sales euro rising above 0.4The measured capital intensity behind the 45% ROIICAnnual cash-flow statements
CEO buys shares in size or the LTI adds a return-on-capital metric(Favorable) the agency-cost concernManagers' transactions; 2027 remuneration report

Sources: as cited; harvia-fsb-2019 … harvia-fsb-2025, harvia-annual-report-2025, harvia-q3-2026, harvia-q3-2026-presentation, masco-10k-2023-sauna360, tulikivi-fy2025. Formulas are in valuation/calculations.json; inputs and their evidence in valuation/model.json.

The independent debate links each surviving attack to the relevant reporting event. Earlier frozen predictions remain unchanged; new debate-vintage forecasts below are stored separately rather than backfitted to realized outcomes. Same-price/date returns, sensitivities and claim judgments; annual model-change arithmetic.

Risks and disconfirming evidence

Assessment: 5 November 2026. The independent debate sizes the risks below against current scenarios. These are business and financing mechanisms, not forecasts of exact failures or a worst-case loss bound.

Surviving attacks

Rebase regional weights without adding steam twice. Sales opportunity rather than a desired IRR determines growth. The reference author forecast/calibration remains intact. Evidence: harvia-q3-2026 pp.6–8; research/08-underwriting-plan.md opportunity ledger.

Use0→5% growth and current 22.68% EBITDA. Derive26.08% model productivity at unchanged 0.6325 capital per organic sales euro; do not retain convenient 30% ROIIC. Finite15year/3%growth/10%requiredreturn tail gives 11.8947x. The earlier 12.3–12.35x was an illustrative diagnostic, not the exact recomputation. Evidence: harvia-q3-2026 pp.4,8,11–13; research/06-risks.md; named capitalPerOrganicSale2020to2025.

Historical share-price median is not the proof. Shorter duration remains an adverse path and prevents high conviction.

Evidence gaps and conviction

Moderate in installed-base/replacement economics; low in exact adoption speed, overhead absorption and postexit duration. The revised adverse path represents surviving attacks rather than merely labeling them.

Reassessed changed joint scenarios. Retain coarse existing horizon weights; there is no new empirical frequency evidence for a precise odds change. Slower central demand and no adverse margin recovery are expressed in operating paths, not by tuning probabilities to returns.

What would change the view

10February2027 FY2026 results: regional growth, gross margin/opex, Lewisburg cost/ramp and cash conversion; subsequently reported FY2027 regional sales and operating cash.

Full bull case, bear case and rebuttals; same-price/date financial effects. Existing frozen predictions are preserved, with the new debate vintage dated separately.

Thesis-to-model register

Current assessment:5 November 2026. The source calibration is preserved separately. Reported facts, normalization estimates and new forecast judgments keep their own evidence labels.

Starting financial bridge

Model inputValueClassificationEconomic definition and evidence
revenue218.878derived reportedFY2025 + 9M2026 − 9M2025 using detailed reported figures. The Q2 deliveries completed in Q3 are already included; no second backlog addition. This is a TTM proxy, not reported October revenue. Sources: harvia-q3-2026
ebitda49.649derived reportedMatching trailing adjusted EBITDA from the detailed thousands table. Includes the North America timing correction; no additional EUR0.8m addback or continuing growth-opex addback. Sources: harvia-q3-2026
shares18.830dilution proxySeptember issued shares less treasury shares plus Q3 share-plan dilution allowance (0.140m). Current proxy; not a full future grant-by-grant valuation. Sources: harvia-q3-2026
netDebt60.828pro forma estimateSeptember bank debt less September cash, plus the October dividend once. Economic lease cash is charged in recurring capital, so lease debt is excluded. Unreported post-September operating cash is not invented. Sources: harvia-q3-2026, harvia-dividend-calendar
depreciation8.223derived reportedFY2025 + 9M2026 − 9M2025 income-statement D&A. Used for tax/FCF, separately from economic sustaining capital. Sources: harvia-q3-2026
financingExpense3.410derived reported proxyTrailing finance costs less finance income, excluding fair-value movements. Hold constant as a refinancing assumption, with no additional credit for growing cash. Sources: harvia-q3-2026
workingCapitalRate0.259derived current ratioSeptember working capital / updated TTM sales. This seasonal high is a conservative funding proxy, not a recurring quarterly cash outflow or free opening cash release. Sources: harvia-q3-2026
dividendPerShare0.770reported policy anchor0.77 annual dividend paid in 2026. Future growth and payment dates are explicitly estimated. Sources: harvia-dividend-calendar
otherClaims1.635carried book-value estimateSeptember current and non-current employee-benefit obligations, carried once as a separate terminal claim. Minority earnings are already removed; no second NCI book-value deduction. Sources: harvia-q3-2026

Scenario assumptions

Base initial growth is 10%, fading to 9% over ten years, supporting EUR348.956m sales in year 5. FY2025 regional weights are explicitly allocated to current TTM sales: North America15% growth, Europe5%, APAC/MEA10% produces approximately EUR348.5m after five years. Steam overlaps those regions. Base EBITDA recovers to the 25.7% cycle average; upside alone retains 29.2%. The downside uses 0→5% growth and current 22.683% EBITDA, with measured capital per sales euro unchanged.

The live base/upside use 20× cash NOPAT, after recurring economic capital and operating tax, with debt/claims deducted once. Installed-base replacement, distribution and high incremental productivity support a conditional premium. An illustrative finite calculation requires roughly 12 postexit years at 9% cash growth, 38.5% productivity and 9% required return; that duration is uncertain, not measured. Historical trading multiples are context rather than proof. The adverse 11.8947× is calculated from 15 finite postexit years at 3% nominal growth, 10% required return and 26.0803% productivity, then no-growth value. Its exact annual cash receipts and capital costs are in the model-change record.

base

Regional demand bridge supports initial 10% growth fading to 9%, about EUR349m in year 5: current TTM sales allocated by FY2025 regional weights, North America15%, Europe5%, APAC/MEA10%. Steam is included in the regions. Retain the 25.7% cycle-margin recovery and measured capital intensity; growth above this needs additional identified opportunities.

DriverValue/pathReason and sources

downside

Adverse adoption/destocking with 0% initial growth recovering toward 5%, no recovery from current 22.68% EBITDA margin, and unchanged measured capital per sales euro. Terminal11.89x is a finite 15-year/3%nominal growth/10%required-return judgment, followed by no-growth value; no claim that this is a loss bound.

DriverValue/pathReason and sources
targetMargin0.227Adverse adoption/destocking with 0% initial growth recovering toward 5%, no recovery from current 22.68% EBITDA margin, and unchanged measured capital per sales euro. Terminal11.89x is a finite 15-year/3%nominal growth/10%required-return judgment, followed by no-growth value; no claim that this is a loss bound. Sources: harvia-fsb-2019, harvia-fsb-2020, harvia-fsb-2021, harvia-fsb-2022, harvia-fsb-2023, harvia-fsb-2024, harvia-fsb-2025, harvia-annual-report-2025, harvia-q3-2026

upside

Faster near-term adoption begins at 17% and fades to the same 9% long-run rate. Keep 29.2% normalized EBITDA, 45% ROIIC and 20x cash NOPAT. Favorable demand does not itself establish a higher terminal multiple or a permanently longer growth runway. No M&A uplift is assumed.

DriverValue/pathReason and sources

Evidence that would change the inputs

10 February 2027 FY2026 results: regional growth, gross margin/opex, Lewisburg cost/ramp and cash conversion; subsequently reported FY2027 regional sales and operating cash.

Reassessed changed joint scenarios. Retain coarse existing horizon weights; there is no new empirical frequency evidence for a precise odds change. Slower central demand and no adverse margin recovery are expressed in operating paths, not by tuning probabilities to returns.

Detailed claim-by-claim judgment. All shared-engine calculations retain stable fact IDs and audit links; original source arithmetic is not a current forecast.

Underwriting plan

This plan replaces 08-underwriting-method.md (driver map, 27 October) and 12-q3-2026-underwriting-plan.md (Q3 decision record, 29 October). The opportunity ledger and scenario review were written on 3 November 2026 after the stage 2–7 work, before returns were recalculated. Money is EUR millions.

Opportunity ledger

OpportunityStarting pointCustomer need and why Harvia winsReachable revenue in five yearsInvestment and cash costTiming and limitsEvidence against
North American adoption: rooms and heaters (volume, share)NA revenue 75.8 (2025), +21.9% in 9M 2026; Almost Heaven D2C, retail and dealersHome sauna for health; Harvia offers rooms plus heaters, US-made (about 70% of US end products), and the largest retailer relationship (about 11% of group sales)About 150–170 if NA grows 15% a year, half of the base case's incrementLewisburg expansion (in 2025–26 capex of 14.8 and 7.9; cost not itemized); working capital at about 0.29 per sales euroLewisburg completes Q4 2026. US consumer confidence low, housing weakSteam (housing-linked) is falling; room pricing is campaign-driven; new entrants (Tulikivi ETL heaters, imports); Sauna360/Masco
European replacement and renovation (replacement, share)Northern Europe 46.6 (+12.8% 9M 2026), Continental 55.6 (+5.1%)Replacement of heaters in a large installed base; Harvia's share of heaters and components is over 20%4–6% a year from replacement and modest share gains: about 125–135 in 2031Low: Muurame has spare capacity from 2021; working capitalFinnish construction weak; Northern Europe fell in 2023 and 2024Energy prices cut sauna use (2022); Q2 2026 IT disruption showed delivery risk
APAC & MEA (new markets)21.0 (2025), +9.6% in 9M 2026Growing awareness in China, Japan, Australia30–35 at 10% a yearSales organizations (Japan company set up 2023), Guangzhou factoryMiddle East (about 2% of group revenue) on holdSmall base; project-driven
Price and mix (price)Gross margin 64.2% (2025) vs about 60% in 2019–22Premium EOS, digital controls (MyHarvia sensor), heater/control mixSupports the margin; not a separate revenue lineProduct development (R&D expensed 2.1 a year)Campaigns and tariffs offset partRoom mix in NA dilutes; Black Friday discounting
Overhead absorption (margin)Opex 42% of sales (2025); adjusted EBITDA margin 22.7% trailingOperating leverage once the IT platform, regional teams and Lewisburg stop growing faster than sales29.2% EBITDA needs an opex ratio near 36%: opex growing 4–5 points a year slower than revenue for about four yearsNone in capital; it is a spending choiceThe program continued in Q3 2026 (headcount +12%)Four years of opex outgrowing revenue; no metric in pay rewards it
Steam (ThermaSol, EOS)17.2 (2025), −8.2% in 9M 2026Steam rooms and showers in US renovation and commercial projectsFlat to slightly upLittleTied to US housingAlready declining; possible goodwill pressure
Acquisitions73.6 deployed FY2020–FY2025, about 12 a year, lumpyFragmented sauna and spa market; Harvia paid about 1.3× salesNot modeledDebt headroom about 63 to the 2.5× ceilingManagement says it is ready to actThermaSol at about 11× EBITDA has underperformed; EOS minority bought at a boom multiple

Overlap. NA heater revenue sits in the first row, not in price and mix; European replacement excludes NA. The margin line is a cost line, not revenue. Acquisitions are excluded from all cases: cash that can't be placed reduces net debt, following the source and the company's dividend policy.

Management targets are a cross-check, not the demand forecast. The issuer targets 10% average annual revenue growth and adjusted EBIT above 20%. The current base now starts at 10% growth and fades to 9%, because the regional opportunity bridge supports that scale independently. The 25.7% cycle EBITDA margin remains partial overhead recovery; the 29.2% author normalization remains favorable only. Current returns and sensitivities are in the dated debate review rather than earlier-price examples.

Scenario mechanisms

Current5 November 2026 judgment: Base initial growth is 10%, fading to 9% over ten years, supporting EUR348.956m sales in year 5. FY2025 regional weights are explicitly allocated to current TTM sales: North America15% growth, Europe5%, APAC/MEA10% produces approximately EUR348.5m after five years. Steam overlaps those regions. Base EBITDA recovers to the 25.7% cycle average; upside alone retains 29.2%. The downside uses 0→5% growth and current 22.683% EBITDA, with measured capital per sales euro unchanged.

North American adoption slows, dealers destock and overhead fails to absorb. Flat initial sales gradually recover toward 5% growth while EBITDA stays at the reported 22.68% margin. Capital per incremental sales euro remains EUR0.6325, so model productivity becomes 26.08%; reducing the margin does not silently make growth cheaper. There are no acquisitions. Product recalls, aggressive price competition or a debt-funded bad acquisition can produce losses outside this representative case.

The forward opportunity ledger determines the operating path before return comparisons. It is not a projection of past CAGR or management targets. Capital, timing and ownership claims remain matched to the revenue/cash perimeter. Reassessed changed joint scenarios. Retain coarse existing horizon weights; there is no new empirical frequency evidence for a precise odds change. Slower central demand and no adverse margin recovery are expressed in operating paths, not by tuning probabilities to returns.

Current absolute scale and terminal opportunities

5-year base: revenue 348.957, owner cash 65.526, exit cash earnings 65.526 (EURm). 10-year base: revenue 542.406, owner cash 101.938, exit cash earnings 101.938 (EURm).

The live base/upside use 20× cash NOPAT, after recurring economic capital and operating tax, with debt/claims deducted once. Installed-base replacement, distribution and high incremental productivity support a conditional premium. An illustrative finite calculation requires roughly 12 postexit years at 9% cash growth, 38.5% productivity and 9% required return; that duration is uncertain, not measured. Historical trading multiples are context rather than proof. The adverse 11.8947× is calculated from 15 finite postexit years at 3% nominal growth, 10% required return and 26.0803% productivity, then no-growth value. Its exact annual cash receipts and capital costs are in the model-change record.

No return target determines this path. Source locators and disagreements are in 12-debate.md and 07-assumption-register.md.

Method self-evaluation - 27 October 2026

This review reread the investment-cycle and valuation arguments, inspected the original page 96 table, and checked the H1 2026 issuer figures. It is not an independent blind replication or a new complete company/industry review.

What failed

The prior prompt required source reproduction but did not require an explicit economic test distinct from that reproduction. The existing check rebuilt cash NOPAT and FCF from source operating inputs. That proved the earnings definitions and arithmetic, while still accepting future EBITDA and investment rows as inputs. It did not prove that the research process would choose a coherent forecast or downside.

The current downside also permanently reduced normalized EBITDA from 29.2% to 25.5% without identifying separate evidence of impaired unit economics. The upside combined faster growth, a stronger terminal growth rate and 22x valuation. Both departed from the causal case more than their labels made apparent.

Revised test

valuation/method-check.json holds a separate, simplified ten-year economic reconstruction at the original EUR38 price and FY2025 opening balance sheet. The engine generates future revenue, margins, NOPAT, investment and cash. Numerical benchmarks are used only after forecasting. A regression test changes those benchmarks and confirms that the generated forecast is unchanged.

The reconstruction still accepts explicit author judgments: normalized margin, organic return on capital, tax, dividends and valuation. Those judgments are explained in the driver register; the test does not claim to derive their exact values uniquely from observations. Historical figures, source-derived recurring-capital and depreciation proxies, simplified growth fade and margin timing remain visible. The range is a sensitivity to a documented 15%-17% near-term growth cadence, not a probability interval or optimized fit.

The initial high-cadence reconstruction gave 17.34%, above 16.7%. Its ending revenue was EUR673.5m versus EUR637.2m in the report. That exposed an overly strong simplified growth path, not an excuse to alter taxes or the multiple. Keeping the source-supported cadence range gives the following comparison:

CheckReconstructed resultAuthor referenceInterpretation
Exact printed valuation bridge16.7303%16.7%Matches printed rounding.
Exact operating-input cash bridge16.7349%16.7%Matches printed rounding; annual maximum difference EUR0.139m.
Driver-generated base range16.37%-17.34%16.7%Economically consistent range; no claim of an exact independently rediscovered point.
Ending revenueEUR616.9m-EUR673.5mEUR637.2mOriginal outcome lies within the mechanically generated output range.
Ending cash NOPATEUR132.7m-EUR144.9mEUR137.3mSame.
Cumulative FCFEUR638.9m-EUR674.9mEUR649.9mSame; future FCF was generated, not inserted.
Ending net cashEUR391.0m-EUR427.0mEUR411.1mSame; distributions and retained cash reconcile.

Run npm run thesis:check -- harvia to reproduce the numerical consistency check. npm test also verifies that slower growth reduces required investment without silently damaging normalized margins, worse capital efficiency consumes more cash, and a changed expected return or cash target cannot feed back into forecast inputs.

Remaining limits

The exact growth and margin cadence is not determined by the prose alone. A range containing the source result is weaker evidence than independently producing the exact annual curve. This check supports causal consistency; the separate original table remains the precise reference. No tolerance was widened to relabel the economic reconstruction as an exact match.

The input range was assembled during an open-book review after the report was known.

The grower adapter does not explicitly model capacity cohorts or ramp delays. The report's project analysis informs its ROIIC judgment, but the saved check is not a new project-level replication. A later capacity-cohort model requires its own evidence and tests. No M&A uplift or scenario probabilities are inferred.

Current five-year returns use updated financials, a different starting price/date, a fresh forecast window and dated dividends. They cannot be made to equal the historical 16.7% merely by treating that figure as a target.

Capital productivity and reinvestment

Assessment:5 November 2026. Values below come directly from the same shared forecast used by the dashboard. Money is in EUR millions.

HorizonScenarioCapital-return definitionReturn/proxyPeriod-weighted reinvestmentGrowth capitalMatching earningsFunding residual
5 yearsdownsideOrganic ROIIC · model implied26.08%4.27%7.842183.8092.13e-14
5 yearsbaseOrganic ROIIC · model implied38.45%24.28%63.679262.2782.84e-14
5 yearsupsideOrganic ROIIC · model implied45.00%31.54%107.805341.8171.42e-14
10 yearsdownsideOrganic ROIIC · model implied26.08%9.76%38.559394.9713.55e-14
10 yearsbaseOrganic ROIIC · model implied38.45%22.84%158.382693.5785.68e-14
10 yearsupsideOrganic ROIIC · model implied45.00%24.78%249.6591007.6085.68e-14

Historical capital productivity is distinct from this table’s model-implied organic productivity. FY2019–2025 and FY2020–2025 lagged capital/sales measurements remain in calculations.json. Base capital intensity remains approximately 0.49 per incremental sales euro; adverse remains 0.6325. Lower adverse EBITDA changes productivity to 26.08%, not the required physical/working capital per sales euro. Total earning-power change can include margin recovery and is not a newly measured ROIIC.

Reinvestment equals total new growth investment divided by total matching earnings over the period, not the average of annual percentages. Cash generation less distributions/old claims plus opening cash drawn and net borrowing reconciles to new investment. No extra borrowing or invented working-capital release is introduced. All six funding ledgers reconcile within 0.000001 model millions.

Sources and definitions:04-capital-allocation.md, valuation/calculations.json and the shared ledger; docs/method/cash-flow-and-reinvestment.md. Dated review.

Scenario probabilities

Assessment:5 November 2026, after the independent debate. Low-confidence subjective joint-path weights; no claim of empirical calibration or source-author odds.

HorizonBearBaseBull
5 years25.00%55.00%20.00%
10 years30.00%45.00%25.00%

Reassessed changed joint scenarios. Retain coarse existing horizon weights; there is no new empirical frequency evidence for a precise odds change. Slower central demand and no adverse margin recovery are expressed in operating paths, not by tuning probabilities to returns.

Base initial growth is 10%, fading to 9% over ten years, supporting EUR348.956m sales in year 5. FY2025 regional weights are explicitly allocated to current TTM sales: North America15% growth, Europe5%, APAC/MEA10% produces approximately EUR348.5m after five years. Steam overlaps those regions. Base EBITDA recovers to the 25.7% cycle average; upside alone retains 29.2%. The downside uses 0→5% growth and current 22.683% EBITDA, with measured capital per sales euro unchanged. North American adoption slows, dealers destock and overhead fails to absorb. Flat initial sales gradually recover toward 5% growth while EBITDA stays at the reported 22.68% margin. Capital per incremental sales euro remains EUR0.6325, so model productivity becomes 26.08%; reducing the margin does not silently make growth cheaper. There are no acquisitions. Product recalls, aggressive price competition or a debt-funded bad acquisition can produce losses outside this representative case.

Moderate in installed-base/replacement economics; low in exact adoption speed, overhead absorption and postexit duration. The revised adverse path represents surviving attacks rather than merely labeling them. Review on 10 February 2027 FY2026 results: regional growth, gross margin/opex, Lewisburg cost/ramp and cash conversion; subsequently reported FY2027 regional sales and operating cash.

Weights sum to 100% independently for each horizon. These paths represent outcome regions and omit permanent-loss tails. Price alone does not change business-path odds; local scenario overrides suspend researched weights. Weighted IRR is calculated from probability-weighted dated shareholder receipts, not by averaging scenario IRRs. Evidence and rebuttals; method:docs/method/scenario-probabilities.md.

Bull–bear review — Harvia, 5 November 2026

Independent advocates prepared opening cases without seeing the other draft, then exchanged point-by-point cross-examinations. A separate judge who did not author the prior model assessed the competing claims. Source reconstructions retain their stated provenance; inaccessible private reports were not newly read. Quoted sensitivities test arguments and are not automatically adopted forecasts.

Fixed quote EUR 40.65, market time 2026-11-04T16:29:38.000Z; valuation date 2026-11-05. Original forecast exits retained.

HorizonBefore bear/base/bullAfter bear/base/bullWeighted before → after
5 years5.33% / 17.81% / 22.79%-6.47% / 14.01% / 22.79%16.34% → 12.45%
10 years8.24% / 15.71% / 18.37%0.93% / 13.28% / 18.37%14.73% → 12.53%

Bull case

As of 5 November 2026. No opposing advocate draft was read. Existing company research, fact/calculation ledgers and exact registered source locators were examined; original issuer sources additionally checked live: Lumine Q3, CSU Q2, Diploma Q3 transcript, NVIDIA CFO commentary, TerraVest Highland/KBK, WOSG FY26. Harvia Q3 PDF retrieval failed; its registered extraction remains the evidence limit. Private source interpretations below are the repository's documented reconstruction, not a new claim to have read an unavailable private original.

The strongest bear case is that growing North American adoption is being purchased with recurring overhead, with a dealer/consumer cycle and campaign pricing layered on top. The base's15% initial growth also exceeds the revenue independently reachable from its own opportunity ledger.

The existing ledger forecasts NA150–170, Europe125–135 and APAC30–35. These are regions: steam already sits inside them and cannot be added again. Allocate TTM218.878 by FY2025 regional weights 75.8/46.6/55.6/21.0 (explicit unchanged-mix proxy), then NA15%, combined Europe5%, APAC10% for five years: about 348.5m. The engine's10% initial growth fading 9% produces 348.96m versus 415.23m currently. I concede 15% base lacks~66m of named opportunities; a10% initial base is justified by the bottom-up bridge, not management authority.

HARV-B2 — Margin absorption is the strongest legitimate upside. Lewisburg capacity, IT implementation and regional teams can cease outgrowing sales; heater/control mix and higher gross margin can release operating leverage. Test29.2% EBITDA with ROIIC adjusted to preserve the base's measured growth-capital intensity; no free productivity uplift. The outside view is 25.7% through-cycle EBITDA; management's continuing growth-spend preference prevents making 29.2% central today.

HARV-B3 —20× has an economic duration cost. Heaters participate in rooms sold by others; replacement and distribution can outlast initial adoption.20× at 9% required return, 9% post-exit growth and 38.5% marginal return requires~12 additional high-return years. A15× test quantifies shortened runway. It is not a recommendation to retain 20× merely because historical multiples averaged more. Distribution concentration, competing heaters and health-fashion risk can truncate the duration. No acquisition upside is credited.

Isolated shared-engine effects

Saved quote 40.65; market timestamp 2026-11-04T16:29:38.000Z. Fixed valuation date 2026-11-05; model 09878f061ae1; exits remain the original model vintage. These are sensitivities, not joint forecasts or probability revisions. Baseline base IRR 17.81% / 15.71% (5/10 years).

ClaimExact input/path5-year base10-year baseChange, pp (5/10)
HARV-B1cases.base.inputs.growthRate19.33%16.67%+1.52 / +0.96
HARV-B2cases.base.inputs.targetMargin + roiic (constant capital intensity)21.19%17.38%+3.38 / +1.67
HARV-B3cases.base.inputs.exitMultiple11.65%13.03%-6.15 / -2.68
HARV-B4cases.base.inputs.growthRate14.01%13.28%-3.80 / -2.43

Bear case

Strongest opposing case

Heaters and controls participate in both new sauna adoption and later replacement; Harvia's distribution and US manufacturing can earn attractive returns without owning the whole room market. Gross margin is strong and the matched capital-intensity reconstruction through a boom and bust supports economically attractive organic growth.

Bear case

The present base assumes more demand than its own opportunity ledger explains. Starting from EUR218.9m trailing sales, identified North American adoption, European replacement/share gains, APAC and flat steam total onlyEUR320–360m in 2031. The15→9% base reaches roughlyEUR415m. Filling the gap needs faster conversion, share or pricing with named evidence; saying the latest organic growth was 16% is trend extrapolation, not a reachable-demand bridge. A10→9% sensitivity is consistent with the ledger's scale, chosen independently of the issuer's similar target.

The downside is internally inconsistent with the written bear. The modeled downside still grows 7% every year and lifts 22.7% EBITDA to 25.7%; it is slower compounding, not the described dealer destocking/adoption pause with persistent overhead. Yet Harvia itself had−7.7% and−9.4% organic years in 2022–2023. A meaningful adverse path can hold sales flat initially, retain current overhead and use a trough multiple without alleging permanent product obsolescence. [harvia-fsb-2022 p.2; harvia-fsb-2023 p.2; research/05-thesis.md Bear case; valuation/model.json downside.]

High gross margin does not automatically translate into owner-cash margin. Personnel and operating cost grew faster than sales for several years, despite a record gross margin. Continued investment is necessary to create the assumed adoption; simultaneously counting rapid growth and removing its sales/service cost overstates operating leverage. The base's25.7% is already a recovery. The upside 29.2% requires overhead absorption not yet observed. [harvia-fsb-2023 through harvia-fsb-2025 income statements; harvia-q3-2026 financials pp.15–23; harvia-annual-report-2025 remuneration pp.46–52.]

A20x exit exceeds the 11.1x no-growth value at 9% required return and requires 4–5% growth with attractive reinvestment after 2031. The current record does not establish how many additional years of North American adoption/share gain support that premium. The2022 trough 12.3x shows the price can discount a good product franchise severely when channel conditions reverse. Historical medians are observations, not a terminal entitlement. [research/05-thesis.md Exit multiple; harvia-fsb-2019…2025.]

Contested claims

  • HARVIA-B1 —Reduce base initial growth to 10%, fading 9%, to close the identified scale gap; if the bull supplies incremental reachable demand, reconsider.
  • HARVIA-B2 —Keep EBITDA at today's22.7%, rather than assuming recovery. This tests persistent investment cost; it is not a claim that gross product economics weakened.
  • HARVIA-B3 —Test15x on otherwise unchanged base: terminal assumptions matter much more than the notebook's stale 15.5% sensitivity suggests at the present model/quote.
  • HARVIA-B4 —Use the already-written bear as an explicit stress:0→5% growth, 22.7% margin, 30% ROIIC,12.3x. A negative return is not an excuse to discard it. It is not a probability-calibrated worst case.

The current issuer PDF was inaccessible through the web tool in this run; page-specific figures above come from the existing source-linked reconstruction. No new channel sample was performed.

Shared-engine effects

Fixed saved quote EUR40.65; market timestamp 2026-11-04T16:29:38.000Z; valuation date 5 November 2026; fixed original exit calendar. Base before: 17.81% / 15.71% (5/10years). These are one-change sensitivities to the base, except named bundles; existing scenario probabilities are not revised or claimed as validated.

ClaimChange5-year base IRR10-year base IRR5-year effect10-year effect
HARVIA-B1Base growth 10% fading 9%14.01%13.28%-3.80pp-2.43pp
HARVIA-B2EBITDA stays at current margin14.72%14.19%-3.09pp-1.52pp
HARVIA-B3Base exit 15x11.65%13.03%-6.15pp-2.68pp
HARVIA-B4Written bear:0→5% growth, 22.7% margin, 30% ROIIC,12.3x exit-5.87%1.25%-23.68pp-14.46pp

Rebuttals

Bull response

5 November 2026. Written after independently frozen openings were exchanged. Every numbered bear claim is addressed below; original bull openings and sensitivities remain unchanged.

  1. Bear10→9% growth: concede, with an independently rebuilt bridge. FY2025 regions sum 199.0m: NA75.8, Northern Europe46.6, Continental55.6, APAC/MEA21.0. Allocate current TTM218.878 at those regional weights (an explicit unchanged-mix estimate), then compound NA15%, combined Europe5%, APAC10% for five years:~348.5m. Steam is inside regions, never an extra 17m. Shared engine 10→9% gives 348.96m; present 15→9% gives 415.23m. No named~66m incremental opportunity bridges the excess. My17% is an upside sensitivity only; recalibrate its absolute market/share implications rather than treating a positive IRR as evidence.
  2. Bear flat 22.7% margin: concede as adverse; partly rebut as central. Continued growth overhead is observed; gross margin and completion of current capacity/IT support possible absorption (harvia-q3-2026, pp6–8; harvia-fsb-2025, p8; source reconstruction).25.7% cycle-average base is a reasoned partial recovery, not proof; 29.2% belongs in upside until costs demonstrably lag sales. Do not eliminate necessary growth opex while retaining all demand.
  3. Bear15× exit: concede duration sensitivity; no automatic haircut.20× is supported only if replacement, distribution and attractive incremental economics persist for roughly 12 post-exit high-return years in my stated diagnostic. No source empirically establishes that life.15× belongs to shorter/lower-return paths. A method correction should state the duration risk; replacing 20× merely because a lower multiple lowers returns is not analysis.
  4. Bear written adverse path versus modeled mild downside: concede substantive inconsistency.0→5% growth with current 22.7% EBITDA better matches adoption pause/dealer destocking than 7% perpetual growth plus margin recovery. Preserve the measured~0.63 capital per organic sales euro: ROIIC must fall with margin (~26.1%, not an independently chosen 30%).12.3× is an observed 2022 trough stress, not an economic entitlement. A finite diagnostic at 10% required return, 3% nominal post-exit growth,~26% incremental return and 15years gives~12.35×; that duration is a disclosed judgment, not evidence precise 12.3× is right. Reassess weights after changing scenario definitions. Conviction moderate in product economics, lower in the old forecast.

Bear response

5 November 2026, after independent openings were exchanged.

  1. Bull margin absorption: upside conceded, free capital productivity rejected. Finished capacity/IT and strong gross margin support possible operating leverage, but current growth overhead is recurring until shown otherwise.25.7% cycle-average partial recovery is defensible central judgment; 29.2% remains upside. Preserve measured growth capital/sales when margins move; this also corrects my opening's convenient 30% ROIIC in the genuine-bear stress.
  2. Genuine downside refinement. Current reported 22.6834% EBITDA versus 25.70545% old downside implies new ROIIC26.08033483% at unchanged 0.6324826082 capital per incremental sales euro:29.7594% ×(22.6834%−1.26087%)/(25.70545%−1.26087%). With0→5% growth, current margin, 12.3x exit and existing dividend path, shared engine gives−5.9204%/+1.1734% over 5/10years atEUR40.65,5 November. This is a judgment stress grounded in observed destocking and overhead, not a predicted worst case. Both advocates agree old 7% growth plus margin recovery failed the written adverse mechanism.
  3. Bull20x duration: risk conceded, automatic haircut rejected. Historical median is not an entitlement; the bull's12post-exit high-return years are an explicit forward judgment. My15x test also does not establish central fair value. For the adverse 12.3x, an independent finite diagnostic at 10% required return, 3% post-exit growth, 26.08% new-capital return and 15years produces about 12.35x, followed by no-growth value. A perpetual cross-check at 2.7% nominal replacement growth yields 12.28x. These chosen duration/growth judgments corroborate the historical 12.3x trough; neither is a reported forecast. Use the finite one consistently, rather than claiming the trough price proves intrinsic value (harvia-fsb-2022, source calibration; stage 9 terminal definition).

Reassess joint scenario probabilities after changing base/downside definitions. The original investor's calibration remains untouched; the material changes are this study's revised current demand and risk judgments. Gross-margin deterioration, unabsorbed overhead and lost dealers remain the most important refutations. Current issuer-PDF access failure limits this to the registered extraction and independent arithmetic, not new channel research.

Judgment

Existing estimates receive no presumption merely because they are already in the model. Positive company evidence, the relevant outside view and the capital/funding bridge determine each verdict; neither arbitrary conservatism nor an unsupported favorable sensitivity replaces it. The evidence and reconstruction below explain numerical changes or explicit no-change conclusions claim by claim.

Contested claimVerdictEvidence/locatorReason and model action
bear advocate: Current base exceeds independently identified regional demandadoptharvia-q3-2026 pp.6–8; research/08-underwriting-plan.md opportunity ledgerRebase regional weights without adding steam twice. Sales opportunity rather than a desired IRR determines growth. The reference author forecast/calibration remains intact.
bear advocate: The written genuine bear is absent from the actual scenarioadopt-modifiedharvia-q3-2026 pp.4,8,11–13; research/06-risks.md; named capitalPerOrganicSale2020to2025Use0→5% growth and current 22.68% EBITDA. Derive26.08% model productivity at unchanged 0.6325 capital per organic sales euro; do not retain convenient 30% ROIIC. Finite15year/3%growth/10%requiredreturn tail gives 11.8947x. The earlier 12.3–12.35x was an illustrative diagnostic, not the exact recomputation.

Reassessed changed joint scenarios. Retain coarse existing horizon weights; there is no new empirical frequency evidence for a precise odds change. Slower central demand and no adverse margin recovery are expressed in operating paths, not by tuning probabilities to returns.

Conviction

Moderate in installed-base/replacement economics; low in exact adoption speed, overhead absorption and postexit duration. The revised adverse path represents surviving attacks rather than merely labeling them. These are forward estimates with surviving attacks, not confirmed outcomes. Neither numerical calibration, a completed checklist nor passing software tests establishes investment skill. Reported facts stay separate from forward judgments, and adverse paths are not worst-case bounds.

What would change the view: 10 February 2027 FY2026 results: regional growth, gross margin/opex, Lewisburg cost/ramp and cash conversion; subsequently reported FY2027 regional sales and operating cash.

Exact inputs, isolated sensitivities, model hashes and same-price/date before/after returns.

Public source library

Primary filings, investor communications, and industry evidence.

primary

Harvia issuer release index ↗

Dynamic issuer release index; latest discovered 30 October manager transaction was read in full from issuer HTML and corroborating issuer wire.

Reviewed Nov 2, 2026
primary

Kesko September 2026 sales ↗

Building/home improvement versus technical trade; comparable sales

Published Oct 14, 2026 · Reviewed Oct 27, 2026
primary

Masco Q3 2026 Form 10-Q ↗

Note H; Item 2 pp.16–20: price/volume/mix, tariff benefits and segment discussion

Published Oct 28, 2026 · Reviewed Oct 28, 2026
primary

Harvia Q3 2026 interim report ↗

Full report pp.2–25; pp.6–8 demand/cash/investment, p.13 calendar, pp.15–23 financials and claims.

Published Oct 29, 2026 · Reviewed Oct 29, 2026
primary

Kesko Q3 2026 interim report ↗

Results and building/home-improvement versus technical-trade demand discussion.

Published Oct 29, 2026 · Reviewed Oct 29, 2026
primary

Pentair Q3 2026 results ↗

Full release: segment demand/channel inventory, income/cash statements and FCF reconciliation.

Published Oct 27, 2026 · Reviewed Oct 30, 2026
primary

Harvia financial statements bulletin 2019 ↗

PDF p.2 organic/comparable-currency growth; p.16 consolidated cash flow (working-capital lines, purchases/sales of tangible and intangible assets, acquisitions, lease repayments); p.21 key figures and APM reconciliation (adjusted EBITDA/EBIT, D&A, net working capital).

Reviewed Nov 3, 2026
primary

Harvia financial statements bulletin 2020 ↗

PDF p.2 organic/comparable-currency growth; p.17 consolidated cash flow (working-capital lines, purchases/sales of tangible and intangible assets, acquisitions, lease repayments); p.22 key figures and APM reconciliation (adjusted EBITDA/EBIT, D&A, net working capital). Also p.8 EOS acquisition (2019 revenue EUR17.3m) and p.17 FY2019 lease repayment comparative.

Reviewed Nov 3, 2026
primary

Harvia financial statements bulletin 2021 ↗

PDF p.2 organic/comparable-currency growth; p.18 consolidated cash flow (working-capital lines, purchases/sales of tangible and intangible assets, acquisitions, lease repayments); p.23 key figures and APM reconciliation (adjusted EBITDA/EBIT, D&A, net working capital).

Reviewed Nov 3, 2026
primary

Harvia financial statements bulletin 2022 ↗

PDF p.2 organic/comparable-currency growth; p.17 consolidated cash flow (working-capital lines, purchases/sales of tangible and intangible assets, acquisitions, lease repayments); p.21 key figures and APM reconciliation (adjusted EBITDA/EBIT, D&A, net working capital). Also p.8 EOS minority purchase EUR19.0m and p.17 transactions with non-controlling interests.

Reviewed Nov 3, 2026
primary

Harvia financial statements bulletin 2023 ↗

PDF p.2 organic/comparable-currency growth; p.17 consolidated cash flow (working-capital lines, purchases/sales of tangible and intangible assets, acquisitions, lease repayments); p.21 key figures and APM reconciliation (adjusted EBITDA/EBIT, D&A, net working capital). Also p.8 Phoenix El-Mec (2022 revenue c.EUR1.4m) and the advanced Kirami earn-out.

Reviewed Nov 3, 2026
primary

Harvia financial statements bulletin 2024 ↗

PDF p.2 organic/comparable-currency growth; p.17 consolidated cash flow (working-capital lines, purchases/sales of tangible and intangible assets, acquisitions, lease repayments); p.22 key figures and APM reconciliation (adjusted EBITDA/EBIT, D&A, net working capital). Also p.8 ThermaSol (2023 net sales USD14.4m, adjusted EBITDA margin 17.2%, price USD30.4m).

Reviewed Nov 3, 2026
primary

Harvia financial statements bulletin 2025 ↗

PDF p.2 organic/comparable-currency growth; p.18 consolidated cash flow (working-capital lines, purchases/sales of tangible and intangible assets, acquisitions, lease repayments); p.22 key figures and APM reconciliation (adjusted EBITDA/EBIT, D&A, net working capital). Also p.8 investment commentary (IT modernization, Muurame layout, Lewisburg expansion).

Reviewed Nov 3, 2026
primary

Harvia Annual Report 2025 (incl. Board report, IFRS statements, governance statement, remuneration report) ↗

p.29 Nomination Board holdings; pp.32-39 board and management holdings at 31 Dec 2025; pp.46-52 remuneration report (CEO pay EUR839,153; STI max 70%; LTI TSR/revenue growth/CO2; no shareholding requirement); p.55 ROCE and capital employed; pp.134-135 income statement and balance sheet; p.141 note 2.1 (largest customer c.11% of 2025 net sales; mostly annual contracts).

Reviewed Nov 3, 2026
primary competitor release

Tulikivi 2026 guidance reduction — original Finnish release ↗

Entire release read; sales EUR26.5–28.5m, comparable operating result EUR−1.0–0.0m, interrupted confidence recovery in Finland/Central Europe. Previously known context; no Harvia-specific order or margin measurement.

Published Oct 28, 2026 · Reviewed Nov 4, 2026
primary

Tulikivi Q3/9M2026 interim report ↗

PDF pp.2-10 operating, financing and product commentary; p.11 income statement visually checked. Mixed fireplace/sauna group, not Harvia regional orders.

Published Nov 6, 2026 · Reviewed Nov 6, 2026
primary

Harvia Fenix 2026 product support and remote-use licences ↗

Support page last updated12November, with6November OTA feature note. Control connectivity, remote-use licence and compatibility sections read; not a new financial release or a measured paid-user, attachment-rate, revenue or cash observation.

Published Nov 12, 2026 · Reviewed Nov 15, 2026