Methodology

How the report is built

The product's one claim: every number traces to a public dataset, states its vintage, and carries its uncertainty. This page is the whole method — the sources, the rules, and the places where we decided a number was too weak to print without a caveat.

The rings

Every figure is computed inside three rings drawn around your address at 2, 3 and 5 miles, with the 3-mile ring as the headline. Those radii were chosen by measurement, not convention. Each candidate radius had to pass three tests:

1. A usable margin of error. A 1-mile ring runs ±9–11% on population in suburbs and up to ±24% in rural areas — wide enough that our own report would flag it as unreliable. At 2 miles the margin is about ±6%; at 5 miles, about ±3%.
2. Each ring must add information. In suburban markets each step outward adds 50–60% more population, so the three rings describe genuinely different markets rather than three views of the same one.
3. The numbers must be sayable. Dentists and brokers talk in whole miles.

In rural markets no small radius passes test 1 — at one validation site, 2 miles held 2,860 residents at ±18.6%, a figure too vague to act on. When the 5-mile ring holds fewer than 25,000 residents, the report widens to 5, 10 and 15 miles and says so on the cover. Some competing reports vary their radii too, but silently, which makes two of their reports quietly incomparable. We publish the rule, the threshold, and which set was used.

Demographics

Population, income, education, age, and housing come from the Census Bureau's American Community Survey, 2020–2024 5-year release on 2020 block-group geometry, the newest available. Block groups rarely align with a circle, so each one's contribution is apportioned by the share of its area inside the ring, computed in a site-centered equidistant projection. Margins of error are propagated through every derived rate — when the report says ±4%, that is the ACS uncertainty carried honestly through the arithmetic, not a decoration.

One honesty note we'd rather state than have you discover: a 5-year estimate is an average over its window, so its effective midpoint is about 2022. Commercial vendors advertise "current-year" and forward-projected figures; those are modeled from the same ACS, trading a citable number with a published margin of error for a more recent nominal year. We use the measured release and state its age — which is also why the report carries population growth separately rather than presenting one snapshot.

Growth, measured twice

The obvious approach — comparing two overlapping ACS releases — is invalid twice over: the 2015–2019 and 2020–2024 releases sit on different block-group geometry, and the Census Bureau warns against comparing overlapping 5-year windows because four of the five years are the same sample. So the report measures growth two ways, each labeled with what it can and cannot say:

• Ring level: the 2020 Decennial Census against ACS 2020–2024 — identical geography, flowing through the same apportionment as everything else. Because the ACS midpoint is about 2022, this understates a fast-growing market, and the report says so.
• Recency: the Population Estimates Program, vintage 2025, at county level — the most current measured population the government publishes. Coarser than a ring, so it is reported as the direction the surrounding county is moving, never silently applied to the ring.

Demand: patients, not headcount

Most site reports divide residents by offices and compare against a rule of thumb. That assumes every resident is a dental patient, and they are not: across Ohio census tracts, the measured dental-visit rate spans 19.5% to 82.9%. The report instead applies CDC PLACES tract-level dental-visit rates to the ring population, producing active patients per general-practice office — the pool of people who actually saw a dentist in the past year, divided among the offices already there.

That figure is read against what one office can absorb: a solo general dentist with one hygienist tops out around 1,300–1,500 active patients (industry practice-management figures from AFTCO and others — the ADA deliberately publishes no recommended ratio, on the grounds that one number cannot fit every market, and the report says that too). More patients per office than an office can serve means unmet demand; fewer means the incumbents are competing for a pool too small to fill their chairs.

Health tracts load for every state the rings touch, not just the one the address sits in, and any share of a ring with no published dental-visit rate is reported as a caveat rather than absorbed into the count. The old residents-per-office figure still appears in the report — every competing report leads with it, so you arrive anchored to it — but it is labeled as the weaker metric it is.

The competitor census

The count of competing offices is enumerated, not sampled. Discovery runs an adaptive map-tile search: Google Places returns at most 60 results per query, so any tile at that cap is hiding businesses and gets subdivided, down to tiles roughly 0.4 miles across — finer than any real dental density in the US. A tile still saturated at maximum depth is reported as a coverage gap instead of being silently undercounted.

Each discovered listing then has to survive three passes:

1. Dedupe. Listings are merged by street address, with directionals preserved — N Main and Main St S are different streets — so one practice with four provider records counts once.
2. Exclusion. Labs, suppliers, and non-practice businesses are removed and shown in an expandable excluded list, so you can audit what we didn't count.
3. Classification. Practices are classified general vs. specialist from their own websites, because a specialist in the count inflates your competition. Ambiguous cases get checked; a practice whose name already states its specialty does not, because confirming it changes no decision.

Each surviving office carries a confidence label — verified active, unconfirmed, or likely closed — built from corroborating signals: operating status, reviews, a live website, and the federal NPI provider registry. Only verified-active offices count toward the grade. The registry is enrichment, never the count: raw NPI records overstate real offices by a measured 2.49× at five miles. As an outside check, our office count was validated against Census payroll records — 95 vs. 95 across four fully-swept ZIP codes.

The grade

The letter grade blends the three rings' active-patients-per-office at weights of 25/50/25 — the 3-mile ring carries the most — then applies small, published demand modifiers (high income, high education, rapid growth, low owner-occupancy, high uninsured share) before reading the cutoffs:

Every input to the grade — each ring's count, the blend, each modifier that fired — is printed in the report. There is no adjustment that doesn't appear on the page.

The written reading

The short narrative at the top of the report is generated by a language model that is given only the report's published facts. Then we don't trust it: the text is read back and every figure it quotes is checked against those facts. If any number fails the check, the narrative is dropped entirely and the report falls back to fixed grade copy — a wrong sentence is worse than no sentence.

What the report does not claim

• No projections. Growth figures are measured history with stated vintages.
• No site selection. It screens a market; it doesn't pick your suite or negotiate your lease.
• No production modeling. What a practice would collect here depends on you.
• No false completeness. Censored income tracts, unverifiable listings, coverage gaps, and unpublished health rates are all reported as caveats, not smoothed over.

Sources and vintages

The same appendix, with per-figure citations, ships inside every report — including the free Prosper sample.