How the score is built.
Every rank is a weighted percentile across five forces, computed from every public review we can read. Here are the weights, the inputs, and the arithmetic — because a ranking you can't take apart is just an opinion with a number attached.
Most local rankings reduce a practice to two integers: a star average and a review count. Buyers do not decide that way, and operators cannot act on it. A 4.8 tells you nothing about whether your momentum is accelerating or stalling, whether the practice three blocks away just passed you, or which of the five things that decide a booking you are quietly losing.
Local Standings scores each of those five things separately, weights them, and ranks every practice in a market against the others. The output is a single composite percentile — but the point of this page is that the number is never the product. The reasoning behind it is. What follows is the entire machine: the forces, the weights, the math, a worked example you can check by hand, and a frank accounting of what the score does not yet measure.
Five forces, weighted into one number.
Each practice is scored on five forces. Practice, Pulse, Place, and Pricing are expressed as percentiles — a 73 means the practice ranks ahead of 73% of the eligible field in its own market. Presence is different: it is raw AI-search coverage (appears_in_n / 12 × 100, 0–100 linear) fed directly into the composite — not percentiled. The composite is the weighted sum of those inputs. The weights are fixed, published, and the same for every practice in the cell.
| Force | Weight | What it measures |
|---|---|---|
| Practice | 25% | The work, and the people doing it. Service quality, clinical breadth, retention, and the named providers — the injector or RN who shows up by first name across the corpus. The closest thing to a quality score, read from what customers actually said happened. |
| Presence | 20% | AI-search visibility. Presence = share of the 12 tracked AI-search queries where this practice appears (0–100 coverage, not percentiled). Invisible practices score 0 and differentiate on the other four forces. |
| Pulse | 20% | Reputation velocity. The rate, recency, rating, and substance of new reviews over a rolling ninety days. The leading indicator — the force that moves first when a practice is rising or cooling. |
| Place | 15% | The physical experience, as customers describe it. Atmosphere, the chair itself, scheduling friction, wait times, parking, hours. |
| Pricing | 20% | Posture and transparency. Whether the buyer can see what they'll pay before walking in, value perception in the reviews, and structural signals — membership, packages, financing. |
Composite = 0.25 × Practice + 0.20 × Presence + 0.20 × Pulse + 0.15 × Place + 0.20 × Pricing. The weights reflect what decides a booking in this vertical; they are published per market and they do not change between practices. There is no editorial thumb on the scale, and there is no pay-to-play — a practice cannot buy a rank, a callout, or a removal.
The one force that moves first.
Eight hundred reviews from 2019 lose to eighty from this quarter. Buyers read recent, and recent is where competitive position actually shifts — so Pulse is the force we watch hardest. It is itself a weighted composite of four components, each measured over the trailing ninety days and benchmarked against the practice's own trailing-twelve-month baseline.
| Pulse component | Weight | What it measures |
|---|---|---|
| Recent volume | 40% | Count of reviews in the last 90 days. Raw demand signal — are customers showing up and talking. |
| Recent rating | 30% | Mean star rating of those last-90-day reviews. Recency-weighted quality, not the all-time average. |
| Velocity | 20% | Recent monthly pace ÷ the trailing-twelve-month pace. A 2.0× means the practice is reviewing twice as fast as its own yearly norm — acceleration, not just volume. |
| Substance | 10% | Median character length of recent reviews. Long, specific reviews carry more buyer-decision weight than a bare five stars. |
Each component is percentiled within the market, then combined at the weights above to produce the Pulse percentile, which then enters the composite at 20%. Velocity is the editorially distinct piece: it is what separates a large, static practice coasting on history from a smaller one building a book of business right now.
The entire computation behind rank 1.
Here is the actual arithmetic for the top of the June 2026 Phoenix board. Nothing is rounded for effect; these are the numbers in the dataset.
GIRLTOX Aesthetics
And the Pulse 76.61 itself unpacks the same way: recent volume 85.25 × 0.40, recent rating 63.62 × 0.30, velocity 89.88 × 0.20, substance 54.50 × 0.10 → 76.61. Thirty-one reviews in ninety days at 5.0 stars, reviewing 1.97× its own yearly pace. Every layer decomposes; nothing is asserted.
On the leaderboard and every scorecard, the same drill-down is one hover away: each bar shows the percentile, its weight, and the points it contributes. This is a ranking you are meant to argue with — which is the entire point of publishing the formula instead of a black box.
Who gets ranked, and against whom.
Percentiles are computed within the market cell — Phoenix metro med spas are ranked against each other, never against a national set. A practice needs a minimum of fifteen substantive public reviews to be ranked; below that, the signal is too thin to score fairly, and the practice is tracked but left off the board rather than ranked on noise. For the June 2026 Phoenix issue that leaves 406 ranked practices out of the full set the pipeline tracks.
New practices are handled separately. A practice with under four months of review history has no stable Pulse — with every review inside the recent window, velocity pegs artificially high. Those practices are scored on their other four forces on the main board and surfaced in a dedicated New & Rising section ranked on raw momentum, so a three-week-old practice never vaults the standings on a number that isn't real yet.
What "since May" actually means.
Each issue ships monthly, and the dataset shows velocity: a practice that ranked eighth in May and fourth in June moved for a reason the data can name. Movement is displayed only when it clears a significance bar — a meaningful change in rank backed by a real change in the underlying signal, not the one- or two-review jitter that reshuffles a tightly packed mid-pack. Everything else reads Holding, because a ranking that claims precision it doesn't have is worse than one that admits a tie.
A note specific to this first issue. June 2026 is issue one, so there is no prior monthly snapshot to diff against directly. The May baseline is reconstructed from the timestamps on the reviews themselves — every review carries a date, so we can recompute each practice's standing as of May 1 and compare. That reconstruction is exact for the review-driven forces and is why this issue's movement is anchored on Pulse. From the July issue forward, every force carries true month-over-month movement from two real collected snapshots.
What the score does not yet claim.
The fastest way to lose an operator's trust is to overstate the instrument. So, plainly:
Presence is AI-search coverage, not a reviews-side read.
Presence = share of the 12 tracked AI-search queries where this practice appears (appears_in_n / 12 × 100, 0–100 linear coverage). It is fed directly into the composite — not percentiled. A practice invisible in AI search scores 0 on Presence and differentiates on the other four forces. It does not measure Google Maps rank or organic position for procedure-specific queries beyond those twelve grounded AI-search probes.
Cap-limited corpora are flagged, not hidden.
A handful of very high-volume practices return a truncated slice of their lifetime reviews, which can distort the velocity read. Where that happens the baseline is corrected, and any practice whose history is too shallow to score Pulse reliably is moved to New & Rising rather than ranked on a suspect number.
No pay-to-play, ever.
A practice cannot buy a rank, a callout, a better score, or a quiet removal. The publication is subscriber-funded for exactly this reason — the only customer is the reader, and the only input is public signal.
Corrections are public.
If a number is wrong — a misattributed review, a practice scored in the wrong category, a provider name we got wrong — write to geoff@localstandings.com. We verify, fix it in the next refresh, and note the correction. The methodology is only as good as its willingness to be challenged.
Every public review, every month.
The pipeline reads every public review for every practice in a market — not a summary, the full text — and extracts signal across the five forces, the same way a buyer reads the words before booking. It reads each practice's website for service mix and pricing signals, timestamps the result, and stores a reasoning trace for every score so the number can always be traced back to the sentence that produced it. The corpus is refreshed every thirty days; the time series is the asset.
That's the whole machine. Now go argue with it.