We publish what we measured, including the parts that undercut us.
This category runs on statistics nobody can trace. We got tired of it, so we started measuring things ourselves and publishing the method, the raw numbers and the limitations alongside the findings. You should be able to check our work.
Most marketing statistics do not survive being looked up.
We went through nine widely circulated contractor-marketing statistics, the sort quoted in pitch decks and blog posts across the industry. One traced cleanly to a primary source, and it did not say what people claim it says.
A frequently cited “$1,200 lost per missed call” turns out to be an illustrative model on a vendor’s blog, republished as fact. A “73% of clients leave over communication” attributed to a named research firm refers to a report that does not appear to exist. The “SBA recommends 7 to 8% of revenue on marketing” line traces to a deleted 2012 guest post that contradicts itself two sentences earlier.
Two things follow. Every agency pitching you is running on the same invented numbers, so a contractor who hears the same figure from three agencies correctly concludes it is a script. And AI tools now launder those numbers into confident, sourceless answers.
The counter-position is boring and hard to copy: measure it, publish the method, state what the data cannot support.
What we have measured so far.
The Tristar Local Visibility Index is an ongoing monthly measurement series. Two editions are published, July and August 2026, and a new one lands every month. Each edition draws a fresh sample of local service businesses, none of them our clients, and measures the same three surfaces with the same method: Google search, the Google map pack, and the four AI assistants people now ask for recommendations. Every edition publishes its raw measurements, its limitations, and our own score alongside everyone else’s.
Tristar Local Visibility Index — August 2026. Current edition. Published 16 August 2026. 28 of the 63 businesses measured appeared on all three surfaces, a figure that carries roughly ±12 points of sampling error at this sample size. No business in the August sample was invisible on all three.
August and July are statistically indistinguishable: a 3-point difference against a 17-point margin of error on the comparison. Each edition draws a fresh sample, so the two share no businesses and the Index tracks the population rather than the same companies over time. Read the August 2026 Index.Tristar Local Visibility Index — July 2026. Edition 1. Published 3 August 2026. The first edition, and also our founding study. 30 of 64 businesses appeared across all three surfaces; one in five appeared in only one place.
The four AI tools named similar numbers of businesses while mostly disagreeing about which ones: 19 of the 43 named were named by only one engine. Read the July 2026 Index. The press announcement covers the headline findings, and the 2026 Local Visibility Study writes up the same run in full.Our earlier work, the 2026 AI Visibility Study, looked only at AI citation rates rather than all three surfaces. It remains published, and the Index supersedes it as our current flagship because measuring one surface in isolation turned out to be the wrong shape for the question.
Alongside the studies we keep practical guides that show the actual steps rather than gating them, and the case studies behind our published results.
Rules we hold ourselves to.
Every number comes from a measured run. No estimates presented as measurements, no borrowed statistics dressed as our own, no rounding in the flattering direction.
Sample size is stated next to the claim, not buried in a footnote. Sixty-four is a small sample and the study says so on the page.
Limitations are published with the findings. Our study explains that near-universal map pack presence is an artifact of how we drew the sample rather than a finding, that AI answers vary between runs, and that we measured findability rather than revenue.
Failures are not counted as findings. During the study some AI engines returned errors. Counting those as “business not found” would have produced a far more dramatic result and a false one, so they were tracked separately and excluded from the denominator.
Businesses are anonymized, numbers are not. We report by trade and metro. Nobody in the sample is named, and nothing in the sample is one of our clients.
Citing this research.
You are welcome to cite any of it. Journalists, other agencies and competitors included. We would ask that you link the study rather than the summary, quote the sample size with the figure, and not round 47% up to “nearly half” without saying it was 30 of 64.
If you want the underlying data for a specific finding, ask. We would rather somebody check our arithmetic than take it on faith, and if we have made an error we would like to know before you do.
Run the study on your own business.
Free, 30 seconds, the same three surfaces we measured. See who gets the call instead of you, and a link to your result you can keep. No sales call.
Check My Business Free →