Venture assessment · August 2026
The opening thesis was that radon detection is under-penetrated relative to risk in Toronto and the GTA, and that the opportunity was selling detection devices behind an awareness campaign. The under-penetration was real. Almost nothing else about the thesis held. This is the reasoning path, including what we got wrong along the way.
How to read this
They are numbered because they are genuinely sequential - every choice only makes sense given the one above it. Where a finding later proved wrong, it is marked as withdrawn rather than quietly removed.
Geography rejected - Toronto and the GTA
The thesis assumed GTA homeowners were under-protected relative to their radon risk. Testing penetration is genuinely low - Statistics Canada’s 2023 survey puts Ontario at 10% of non-apartment households ever tested, against 19% in New Brunswick. But low testing in Toronto turns out to be partly rational.
| Metro | Mean Bq/m³ | Above guideline | |
|---|---|---|---|
| Edmonton | 106.4 | 1 in 6 | |
| Calgary | 102.5 | 1 in 6 | |
| Ottawa–Gatineau | 85.9 | 1 in 6 | |
| Montréal | 82.4 | 1 in 6 | |
| Toronto | 43.0 | 1 in 22 | |
| Vancouver | 17.1 | 1 in 113 |
Toronto has the second-lowest exceedance rate of Canada's major metros. The consequence is economic, not epidemiological: roughly 1 in 22 tests produces a billable finding in the GTA versus 1 in 6 in Calgary or Ottawa. A test-to-mitigation model that works in Alberta is four to five times weaker in Toronto on hit rate alone.
The most decisive evidence came from the region itself. York Region Public Health ran a purpose-built study - 91-day protocol, 88% response rate, 474 valid tests:
Out of 474 tested. Average across the region was 47.4 Bq/m³, with no municipality averaging above 75.
“Based on the findings, a follow up study was not recommended.”
The self-selected survey put York at 4.0%. The region’s own study found 0.2% - the direction sampling bias predicts.
Unresolved
The York figures cannot currently be reconciled with the Toronto metro figure: no lognormal distribution with a mean of 43 Bq/m³ produces both 0.2% above 200 and roughly 6% above 100. The two are measuring different things, or one is wrong. Both are reported here rather than the convenient one being chosen, and this is flagged for resolution before any market sizing depends on it.
Scope clarification, public health versus unit economics
Read carelessly, Decision 01 says GTA radon is not worth worrying about. It does not, and the epidemiology says the opposite. The exceedance rate is close to the wrong metric for harm.
Lung cancer risk rises linearly with radon concentration with no threshold, an effect that remains statistically significant when the analysis is restricted to concentrations below 200 Bq/m³. The 200 Bq/m³ guideline is an administrative action level, not a biological cliff. It is also twice the World Health Organization's reference level of 100.
Because risk is linear and far more people live in the middle of the distribution than in its tail, most radon-caused cancers occur in homes that would pass a test. The WHO states it directly: the majority of radon-induced lung cancers are caused by low and moderate concentrations, not high ones. Quantified for Ontario:
| If every home above this level were fixed | Deaths prevented | Share of burden |
|---|---|---|
| 200 Bq/m³ (Canadian guideline) | 91 | 10.7% |
| 100 Bq/m³ (WHO reference) | 175 | 20.7% |
| 50 Bq/m³ | 389 | 46% |
Radon-attributable lung cancer deaths per year, range 230–380. Comparable to the region's annual road fatalities. Derived, not published; see caveat below.
Toronto's attributable fraction as a share of Ontario's. A 39% lower exceedance rate yields only a 20% lower burden, because burden tracks the mean, not the tail.
GTA homes above 100 Bq/m³, versus 1 in 22 above the Canadian guideline. Roughly 422,000 households.
Two honest consequences follow, and they cut in opposite directions.
For an individual, the absolute numbers are moderate and dominated by smoking. Lifetime residence at 200 Bq/m³ raises lung cancer death risk by age 75 from 0.41% to about 0.54% for a never-smoker, and from 10.1% to about 13.3% for a smoker. Roughly 86% of radon-attributable deaths occur in smokers and ex-smokers. Radon is, in absolute terms, largely a multiplier on smoking risk.
Caveat on the GTA estimate: it is calculated from Ontario's published figure scaled by population and exposure, not taken from a published source. Its largest uncertainty is Toronto-specific lung cancer mortality, which could not be sourced. The GTA's high share of apartment and condominium units, which sit near background, probably makes it an overestimate.
Geography rejected - Ottawa
Ottawa was the intuitive pivot: highest-radon major metro in Ontario, and the densest concentration of federally regulated employers in Canada facing a new exposure limit. Both halves fell apart under scrutiny.
The evidence splits into two families that disagree systematically - randomised recruitment versus self-selected volunteers - and the direction of disagreement is predictable.
| Source | Recruitment | City of Ottawa | Kingston |
|---|---|---|---|
| Health Canada 2009–11 | Random digit dialling | 6.9% | 15.6% |
| Health Canada CMA 2012–13 | Random digit dialling | GM 67 | GM 112 |
| KFL&A study 2018–19 | Local volunteer | - | 21% |
| Evict Radon 2024 | Self-selected | 12.0% | 23.4% |
On randomised data the City of Ottawa sits at 6.9% - below the Ontario average of 7.4%. Its hotspot reputation rests on volunteer samples and on an Ottawa-Gatineau figure inflated by the Quebec side of the river. There is a physical reason: Ottawa sits on the Ottawa-Bonnechere graben, sedimentary rock under deep Champlain Sea clay, not exposed Shield.
Certified professionals for ~1.0M people, versus roughly 16 across the GTA’s 6.7M. Ottawa is about five times better served per capita than Toronto.
Ottawa-Gatineau testing penetration in 2018, above the Ontario average of 10% in 2023. A 500-kit municipal challenge ran in 2021.
The federal public sector’s share of projected national compliance cost - and ~56% of it was already tested.
Geography selected - Kingston and the Frontenac Arch
Everywhere else the two sampling frames diverge by a factor of two to three. In Kingston they converge. Its randomised geometric mean of 112 Bq/m³ is the highest of the three Ontario metros ever measured under random recruitment - 67% above Ottawa-Gatineau and 2.5× Toronto. Four datasets, four collection methods, one answer.
The geology explains it. Kingston sits at the southern tip of the Frontenac Axis, where Grenville Province granite and uranium-bearing pegmatite meet the limestone plain. That gives a targeting rule sharper than any municipal boundary.
| Census division | n | Volunteer | Randomised | Ratio |
|---|---|---|---|---|
| Lennox & Addington | 38 | 44.7% | - | - |
| Lanark | 114 | 42.1% | 13.7% | 3.1× |
| Leeds & Grenville | 220 | 31.4% | 13.7% | 2.3× |
| Renfrew | 192 | 29.2% | 9.5% | 3.1× |
| Frontenac (Kingston) | 214 | 23.4% | 15.6% | 1.5× |
| Stormont, Dundas & Glengarry | 28 | 3.6% | 2.3% | - |
Lanark and Stormont-Dundas-Glengarry are adjacent counties. One is 42.1%, the other 3.6%. The Shield contact, not the county line, is the variable - so the service territory follows the geology. A drive-time radius drawn from Kingston would spend half its marketing budget on limestone.
One alternative was tested hard, because it would avoid relocating: the western arc of Hamilton, Halton, Waterloo and Guelph, all within an hour of Toronto and far denser. On volunteer data it looks workable at 10–13%. Under random recruitment it collapses to Hamilton 4.4%, Waterloo 6.5%, Guelph 7.3%, Halton 0%, every one at or below the Ontario baseline of 8.2%, against Kingston’s 15.6%.
Product rejected - consumer detection hardware
The original concern - that detectors are commodities - was correct, and understated. Three price tiers are already crowded and fully distributed through Home Depot, Canadian Tire, Best Buy and Amazon. There is no shelf-space gap.
More decisively, the profit sits two steps downstream. A detector is a $60–250 retail transaction. A mitigation install is $2,800–3,800. The device is a customer-acquisition cost wearing the costume of a product line.
Free and subsidised programmes also anchor the measurement price near zero in exactly the target counties. Two of the three corridor health units supply kits at or below cost - one sells them for $27 and refers elevated results onward to certified professionals.
Catalyst accepted, with limits
A federal regulation registered in January 2026 lowers the workplace radon ceiling from 800 to 200 Bq/m³, in force 30 January 2027. The government’s own analysis projects 23,253 workplaces testing in year one. Verified verbatim against the Canada Gazette.
We tested whether this could anchor a recurring-revenue business. It cannot. No jurisdiction examined mandates periodic workplace retesting, and a single 91-day heating-season test is the accepted way to demonstrate an annual average in Canada. Volume is projected to fall to roughly 90 buildings a year afterwards.
Of ~18,000 federally regulated private employers, roughly this many are trucking firms - 85% with fewer than 20 employees. A small-business motion distributed across eastern Ontario, not a Bay Street pitch.
Federal buildings measured at 200–600 Bq/m³ and never mitigated. Compliant under the old ceiling; non-compliant on the deadline. Old test data creates the obligation rather than discharging it.
So the deadline is used as a wedge into commercial work and a reason to hold certification early - not as the business case.
Recurring model selected
Having ruled out repeat testing, we looked for where recurring obligations genuinely exist. Every one found attaches to mitigated buildings rather than to the general stock: the UK requires annual checks on installed systems because “those systems may also fail over time”; Ireland requires five-year re-verification; the US consensus standard governs operation, maintenance and monitoring of installed systems.
Core risk - unresolved
The single most important number in the research is not a prevalence figure. In Ottawa-Gatineau, 12% of households had tested but only 3% had mitigated. Among people who tested, found a problem, and knew it, the overwhelming majority did nothing.
No choice of geography solves this. Trebling prevalence does nothing to a conversion rate. Every incumbent leaks at the same two joints - the kit that is never returned, and the silence between “your reading is 340” and “the system is installed.”
What the evidence does support is specific and partly counter-intuitive:
Versus 21.9% for a control letter, in a randomised trial of 3,035 households. Reciprocity framing, frequency risk, a deadline, and 20% fewer words.
Adding a county radon map to that letter reduced uptake. A map lets a householder conclude their area looks fine - and nearly every radon site opens with one.
Health Canada’s national kit return rate, achieved with two structured phone calls and nothing more sophisticated.
These combine into an operational requirement rather than a marketing preference. Running measurement as sequenced county campaigns means positives surface geographically clustered 91 days later, so mitigation can be batched - four to six jobs in one town over two or three days. That is what makes rural work survivable when installs take a full crew-day and Perth is 75 minutes each way.
Research integrity
Every research report was independently audited by a separate validation pass instructed to fetch primary sources and attempt to refute. No fabricated sources or invented figures were found in any of them. Three claims did not survive, including two that had already informed decisions.
The claim that the category leader was “a public company fleeing consumer hardware for B2B.”
The annual report shows the opposite: it delisted in February 2026, and the segment it tried to divest was the business one. Its revenue decline was a deliberate inventory build-down, and its own filing states end-user sell-through kept growing. A genuine liquidity crisis at one firm had been read as evidence of category collapse.
A UK figure showing only ~6% of homes in targeted programmes were ever mitigated.
The cited government document contains no such data, and the underlying denominator was probably inverted - making the true figure closer to 0.4%. It had been described as the most important negative result in the research. It is now unresolved and unused.
A strategy pillar resting on four returned test kits.
An app-versus-brochure finding was reported as a 70% drop in kit returns. The underlying study is 2 of 29 against 2 of 9 - four kits across 38 participants. The direction survives; the strategic weight does not. The national 72.4% figure carries that argument instead.
The widely quoted “1 in 6 Ontario homes” figure is not an Ontario figure.
It is a Central Canada rollup combining Ontario with Quebec, which the source itself separates one paragraph later into Ontario 12.4% and Quebec 16.7%. The correct randomised Ontario baseline is 8.2%, from 3,891 randomly selected samples in a peer-reviewed study. Any figure in this category should be checked for which geography and which sampling frame produced it.
The finding that redirected the entire strategy.
The six-metro prevalence table was verified line by line against the primary 6.7MB survey PDF rather than its summary page - which turned out to contain an error the summary propagated. The regulation was verified word for word against the Canada Gazette, including its coming-into-force clause.
Where it stands
Not a device business. The verification steps below are deliberately ordered by how much they could still change the answer - each is cheap, and the first three are prerequisites to spending anything.
Aggregate the Health Canada open-data radon file to compute prevalence by metro directly.
Settles Kingston versus Ottawa definitively, and resolves a live discrepancy: randomised data puts Kingston’s mean 46% above the volunteer figure, the opposite of what sampling bias predicts. It decides whether Kingston is a strong market throughout or merely a base of operations.
Obtain five real price quotes from operators in the corridor.
No operator in eastern Ontario publishes prices. Five phone calls establish the actual ceiling; every figure currently in the model comes from aggregators.
Read the Canadian mitigation standard for existing buildings.
It governs the post-mitigation service that carries the recurring revenue, and it is the one document the research could not open.
Establish basement share across the corridor counties.
Shallow bedrock means crawl spaces and slab-on-grade are common, which raises ticket size but makes installs harder and riskier for a solo operator.
Begin certification, and start the mitigation mentorship clock.
Measurement certification takes two to six weeks. Mitigation requires a six-month mentorship - which is precisely why mitigation capacity in the region is scarce.
Run one behavioural direct-mail campaign in a single high-prevalence township.
Order rate, return rate, elevated rate and mitigation attach, measured once. That single test prices the entire funnel and interrogates the one risk that geography cannot fix.