Your contractor has 4.5 stars. Three times out of four, that rating is lying to you.

The criticism, fairly stated: The report leans on a single 0–100 "trust score," but never publishes the formula, the input weights, or why the "Trusted" line sits at 65 rather than 60 or 70. Every headline — "18% are Trusted," "0% of bathroom remodelers clear the bar" — is a direct function of choices readers can't see. An unexplained composite that produces dramatic numbers deserves skepticism until the machinery is open.

Our answer: This is a legitimate gap and we are closing it. For this edition the score combines three families of signal — (1) review volume and consistency, (2) business verification and active operating status, and (3) available business-history signals — with review volume and consistency deliberately weighted most heavily. We have not yet published the exact coefficients or the 65 threshold's derivation. That is a shortcoming of a v1 benchmark, not a defensible choice. Disclosure we are adding: a public methodology appendix with the component weights, the score distribution, and a sensitivity table showing how the "% Trusted" figure moves as the threshold slides from 55 to 75. Until that appendix is out, treat the 65 cutoff as one reasonable line on a continuum, not a bright natural boundary — and read the distribution (most contractors cluster in a wide middle) as the durable finding, rather than any single side of the cutoff.

The criticism, fairly stated: Review volume is the heaviest input. That makes several headline findings partly definitional. "36% are effectively invisible" and "the review desert" aren't discoveries about trustworthiness — they're restatements of the input. And "stars lie" compares a Google rating against a score that itself ingests Google reviews, so the two are correlated by construction.

Our answer: This is the sharpest fair hit in the report, and we want to be precise about where it lands and where it doesn't.

Where it lands: the "invisible contractor" finding is, by design, a coverage statement. A contractor with two reviews is not necessarily bad — they are unmeasurable, and our score reflects low confidence, not proven low quality. We should not, and going forward will not, let the word "trust" blur that. Disclosure we are adding: we will relabel the low-volume cohort as "low-transparency / low-evidence" rather than implying poor performance, and state plainly that absence of a track record is a reason to verify independently, not proof of a bad contractor — the same standard we hold Google's stars to.

Where it doesn't land: the score is not only volume. Two contractors with identical review counts can score differently on consistency, verification, and operating history. And the star-vs-score comparison isn't pure circularity: the score down-weights a high average that rests on a thin or erratic sample, which is exactly the information a raw star average discards.

The honest one-line version: our score measures how much verifiable evidence of accountability exists, weighted toward review track record. It is a strong proxy for trust, not a measurement of workmanship, and it is partly correlated with the very stars it critiques.

The criticism, fairly stated: The dataset is the population of Google-listable businesses, not the population of contractors. Excellent operators who run on word-of-mouth, referrals, or trade-association channels and never cultivate a Google profile are either absent or land in the "invisible/low-trust" bucket. So the report may be measuring digital-marketing behavior as much as trustworthiness.

Our answer: Correct, and we should not have used "every active contractor in Ontario" without this qualifier. Our frame is every active, publicly-listable Ontario home-service business we could identify from public sources, which skews toward businesses that maintain an online presence. A genuinely great referral-only contractor can score low here purely for lack of a public footprint. Disclosure we are adding: an explicit coverage statement in the methodology, and a reframe of low-evidence scores as "we cannot verify," not "we verified them as untrustworthy." This also sharpens the homeowner takeaway: if you can't find a public record for someone, that's a prompt to ask for references, licensing, and proof of insurance directly.

The criticism, fairly stated: Trust in the trades has many public signals — HomeStars and TrustedPros reviews, BBB accreditation and complaint history, ECRA/ESA and provincial licensing, WSIB coverage, small-claims and Consumer Protection Ontario actions. Using Google alone imports Google's specific weaknesses (fake-review susceptibility, no licensing awareness) and omits the records that most directly indicate whether a contractor has actually harmed customers.

Our answer: True today. This edition's review signal is Google-based, and licensing/insurance fields were too sparse for us to publish (see Q9). We think Google is a reasonable single anchor for a v1 — it's the most complete and most consulted public source — but a single source is a real limitation, not a strength. Disclosure and roadmap we are adding: we name Google as the sole review source in the score for 2026, and commit to integrating licensing-registry and complaint-record signals in the next edition, which is precisely the layer that would move the score from "evidence of reviews" toward "evidence of accountability." Until then, read this as a review-and-verification index, not a compliance-and-litigation index.

The criticism, fairly stated: The publisher scores the exact market it monetizes. A report whose central message is "the free signal everyone uses is broken, and the thing we sell is the fix" has an obvious incentive to make the market look as untrustworthy as possible. Lowering the "Trusted" threshold, weighting toward a metric that produces alarming scarcity, and headlining "checking reviews won't protect you" all point the same commercially convenient direction.

Our answer: The conflict is real and we won't wave it away. We can't prove our own good faith by assertion, so instead we've tried to make the analysis falsifiable and reproducible by outsiders: fixed, stated denominators; a documented snapshot date; raw counts alongside every percentage; and (per Q1) a forthcoming weights-and-threshold appendix so a critic can recompute our headline under different assumptions. Two structural points partly blunt the incentive claim: we score every contractor whether or not they engage with us, and we don't sell placement or rank — so the report doesn't advantage any paying customer. But the deepest mitigation is the threshold-sensitivity disclosure: if our "only 18% are Trusted" figure only looks alarming because of where we drew the line, an open sensitivity table lets you catch us doing it. We would rather you check that than take our independence on trust.

The criticism, fairly stated: East York, North York, Scarborough, Etobicoke, and "York" are not suburbs of Toronto — they are Toronto, amalgamated into the single City of Toronto in 1998. The ranking treats them as separate cities (East York 3rd, "Toronto" 44th, North York 38th, Scarborough 27th, Etobicoke 25th) based on Google's neighborhood labels. So the flagship "big-city surprise" is substantially an artifact of splitting one city by postal-area labels: the "Toronto" row is really just the downtown/old-Toronto core, and one of its own districts (East York) ranks 3rd province-wide. The same problem applies to Ottawa, whose Nepean, Kanata, Orléans, and Gloucester districts are listed separately.

Our answer: This is the most important factual correction and we accept it. The city field reflects the locality label on each business's public listing, not municipal boundaries, so amalgamated Toronto and Ottawa are each spread across several rows. The "Toronto 44th" figure is the old-Toronto core, not the City of Toronto as a whole; aggregating all Toronto districts would produce a different, higher composite and would undercut the "below its suburbs" framing. Corrections we are making: (1) a clear note that "Toronto," "North York," "Scarborough," "Etobicoke," "East York," and "York" are districts of one municipality, likewise the Ottawa districts; (2) a supplementary ranking computed on true amalgamated-municipality boundaries; and (3) softening the "big-city surprise" claim to what the data actually supports — variation within the GTA and between listing-labeled areas — rather than a clean city-beats-suburb story. The genuinely robust point survives: trust varies widely across Ontario geography and is not concentrated in the largest urban labels.

The criticism, fairly stated: The 50-contractor floor still lets tiny samples (Innisfil 55, Midland 51, Aurora 56, Simcoe 56) sit on a ranked ladder against markets ten to twenty times larger, reported to a tenth of a point. With samples that small the confidence interval on an average trust score easily spans several rank positions, so most of the ordering — especially "1st," "64th," and adjacent ties — is noise dressed as precision.

Our answer: Fair, and partly self-inflicted by presenting a strict 1–64 ordering. A 50–90 contractor sample supports a band ("upper third / middle / lower third"), not a precise rank, and certainly not decimal-point separation between neighbors. Disclosure we are adding: confidence intervals (or standard errors) on every city average, a caution that cities within roughly a few points of each other are statistically indistinguishable, and a shift in how we present the extremes — "among the lowest-scoring markets" rather than a definitive "least trustworthy city in Ontario." The superlatives on the smallest samples are the least defensible sentences in the report and we are dialing them back.

The criticism, fairly stated: Ranks 3 through 20 span roughly 57.0 down to 54.0 — three points across eighteen positions, often decimals apart. Treating a 0.1–0.5 point gap as a rank difference implies a precision the data doesn't have, and invites readers to over-read "my city beat yours."

Our answer: Agreed. The rank ladder is more precise-looking than the underlying data warrants. The honest reading is tiers, not ranks: there is a real, several-point spread between the top tier and the bottom tier, but adjacent positions inside a tier are effectively ties. Disclosure we are adding: we will group cities into a small number of statistically-separated tiers as the primary presentation and demote the exact 1–64 ordering to a reference appendix with the precision caveat attached. The defensible claim is the spread between top and bottom; the indefensible one is any specific adjacent-rank bragging right.

The criticism, fairly stated: For a trust index, the omission of the two most concrete trust signals — is this contractor licensed for the trade, and are they insured — is conspicuous. It leaves the score resting on softer proxies.

Our answer: We agree these are the strongest signals, and we deliberately excluded them because our current coverage of those fields is too sparse to support an honest percentage. Publishing a "licensing rate" off a fraction of records would manufacture a misleading denominator, which we refuse to do. This is the one place where the report is conservative rather than aggressive, and we'd rather take the criticism of omission than the criticism of fabrication. Closing this gap (via provincial registries) is the single biggest planned upgrade for the next edition — see Q4.

The criticism, fairly stated: An entire trade returning zero contractors above the line is a classic signature of a threshold/weighting artifact rather than a real-world fact. Remodeling is project-based and episodic: customers remodel a kitchen once a decade, so even excellent firms accumulate reviews slowly compared to a plumber doing hundreds of small jobs a year. If the score is volume-weighted, it will structurally punish low-frequency trades regardless of quality — which means "0% of bathroom remodelers are trustworthy" may be measuring review cadence, not trustworthiness.

Our answer: This is a strong and largely correct critique, and it's the finding we'd most caution readers about. A literal 0.0% is exactly the kind of clean-zero result that should trigger suspicion of a mechanism, and part of the effect is almost certainly structural: remodeling generates fewer, lumpier reviews per firm than high-frequency service trades, so a volume-weighted score will disadvantage it even where quality is high. Disclosures and fixes we are adding: (1) a plain statement that the remodeling result is confounded with review frequency inherent to project-based work, and should not be read as "no trustworthy remodelers exist"; (2) trade-normalized scoring in the next edition, so contractors are compared within a trade's natural review cadence; and (3) removal of the absolute "0%" as a headline superlative in favor of "remodeling trades score lowest on our current volume-weighted model, in part because of how their work generates reviews." The residual, defensible point is narrower but still useful: on exactly the biggest-ticket jobs, a star average is backed by the thinnest public evidence — so independent verification matters most there. That's a statement about evidence density, which the data does support.

The criticism, fairly stated: The report says review volume is weighted heavily, yet drywall simultaneously has the highest "% Trusted" (51.1%) and one of the highest "% with <5 reviews" (50.2%), on a small sample (311). Those two facts sit awkwardly together and suggest either a small-sample quirk or that the weighting doesn't behave as described.

Our answer: A fair catch, and worth explaining rather than glossing. Two things are happening: drywall is one of our smaller trade samples, so its rates are noisier; and the score isn't volume alone — verification and consistency can lift a contractor with a short but clean, consistent record above the line even with modest review counts. Still, the juxtaposition shows our current model can send mixed signals, and it reinforces why the weights need to be public (Q1) so readers can see exactly how a thin-but-consistent record is treated. Disclosure we are adding: we'll flag drywall's smaller sample and use this specific pair of numbers as the worked example in the published-weights appendix.

The criticism, fairly stated: Everything reflects a single capture on/before August 5, 2026. Reviews lag real performance, businesses change hands, and a one-day snapshot can't distinguish a firm on the way up from one on the way down. Annual publication means figures are stale for much of the year.

Our answer: Correct on all counts, and inherent to any point-in-time index. We disclose the vintage prominently precisely so no one over-reads month-to-month precision into it. Two honest caveats we're stating more clearly: individual contractors can move between our bands as their public record updates, so the Index is a snapshot for research and comparison, not a live verdict on any single business; and review lag means the score trails real-world changes by however long it takes customers to post. The population-level findings (distribution shape, trade and geographic spread) are far more stable across a snapshot than any individual score, and those are what we intend the report to be cited for.

The criticism, fairly stated: A model is only a trust measure if a low score actually predicts bad outcomes. Without a validation study against ground truth — consumer-protection complaints, small-claims judgments, licensing revocations — the score is an internally-consistent index that has never been shown to track the thing it's named after.

Our answer: This is the most important thing we can't yet claim, and we won't pretend otherwise. For 2026 the score is a construct-based estimate built from signals that plausibly relate to accountability; it has not been back-tested against an external ground-truth outcome set. That is a genuine limitation of a first edition. What we're committing to: a validation study in a future edition that tests whether lower ProScore bands are associated with higher rates of documented complaints or adverse licensing actions, published whether or not the result flatters the model. Until that exists, the accurate framing — which we'll state in the methodology — is that ProScore measures the strength and consistency of a contractor's public accountability record, and is a tool for comparison, not a guarantee or a verdict on any individual business.

The criticism, fairly stated: The word "trust" and the alarmist framing imply a measurement of whether a contractor will do right by you. What's actually measured is the density and consistency of a public paper trail. The gap between the label and the substance is where a skeptic — or a competitor — will attack hardest.

Our answer: Fair, and it's the throughline behind several answers above. We're tightening the language so the claim matches the substance: ProScore measures verifiable evidence of accountability, weighted toward review track record and verification. It is a strong, comparative proxy for trust — not a character judgment, not a workmanship rating, and not a prediction about your specific job. The headline advice we stand behind without hedging is the modest one: a star rating is a starting point, not a decision; look past the average to the volume, recency, and independent verification behind it — most of all on the biggest jobs. That's a claim the data genuinely supports, and it's the one we want cited.