Methodology and data: State of Med Spa Google Presence 2026

Full study (version 1.0)

• Profile figures and city figures use the full study: 555 US med spas across 20 metros, pulled with one consistent method (SearchApi) for every profile. At this size the national margin of error is roughly ±4%.
• City figures are published for the 19 metros with at least 10 med spas. Phoenix (9 med spas) is not reported on its own.
• City averages roll up to the national figures: the weighted average of the city rows matches every national figure.
• Review velocity and owner reply rate come from the review subsample (115 med spas) and are national only; the subsample is too thin to split by city.
• Review themes come from 2,990 reviews from the same subsample, classified as described below.
• Visibility score, secondary categories and attributes come from the 86 early-look records described below; they were not recomputed on the full run.
• The sections below document the original early-look run (115 med spas), which remains the basis for the subsample figures.

Prepared 2026-09-28. This documents exactly how the published early-look numbers were produced, for transparency and independent checking.

Sample

• Businesses analysed: 115 US med spas (an early-look cut; the full study targets 400 per specialty).
• Data collected: 2026-09-27 to 2026-09-27. Review dates observed in the corpus span 2019-09-28 to 2026-09-28.

Discovery

• Search terms: "med spa", "medical spa", "botox clinic", "aesthetic clinic".
• Metros searched (20): New York, Los Angeles, Chicago, Houston, Dallas, Phoenix, Philadelphia, San Diego, San Francisco, Austin, Miami, Atlanta, Boston, Seattle, Denver, Washington DC, Las Vegas, Nashville, Orlando, Scottsdale.
• For each metro a text search was biased to a 25 km circle around the metro centre, up to 2 pages (20 results each) per term.
• Inclusion: kept a result if its name matched a med-spa keyword (med spa, medical spa, medspa, aesthetic, aesthetics, botox, filler, laser, ...) or its Google category was one of spa, beauty_salon, wellness_center, skin_care_clinic, medical_clinic, health. Businesses marked not operational, or in clearly unrelated categories (hospital, pharmacy, drugstore, nail_salon, hair_salon, hair_care, supermarket, gym) without a matching name, were dropped.

Sampling

• From the discovered pool, a random sample was drawn per metro using a fixed seed (20260927) so the draw is repeatable, targeting an even spread across metros.
• This early-look reflects the 115 profiles that completed full detail collection before a daily Google Places quota limit paused the run. Metro spread: New York 20, Los Angeles 20, Chicago 20, Houston 20, Dallas 20, Phoenix 8, Scottsdale 7. Metros with fewer than 10 profiles are not reported individually.
• Exclusions: none of our own client profiles were in scope (the exclude lists were empty for this run); permanently-closed businesses were dropped at discovery.

Fields and sources

• From Google Places API (New) per profile: rating, review count, primary category, number of secondary categories, website present, phone present, opening hours present, days open per week, open on a weekend, attribute count (payment + accessibility options), editorial summary present.
• From SearchApi.io (Google reviews) per profile: up to 50 most-recent reviews with text, date and whether the owner replied.

Derived measures

• Review velocity (new reviews/month): count of pulled reviews dated within the last 6 months, divided by 6; the national figure is the median across profiles. (Because up to 50 reviews are pulled per profile, velocity is a floor for very high-volume profiles and is reported as a median to limit that effect.)
• Owner reply rate: share of a profile's pulled reviews that have an owner response; the national figure is the mean across profiles, plus the share of profiles replying to at least half of reviews.
• Review-count bands: profiles grouped as <50, 50-199, 200-499, 500+.

Visibility score

A 0-100 composite ("public Rank OS subset"). It combines four sub-scores from signals visible on a public profile, re-weighted to sum to 100:

• Relevance 30%: primary category (base 40 of 100) + a secondary category present (20) + editorial summary present (20) + attributes (>=5 -> 20, >=1 -> 10).
• Review health 34%: review volume (>=150 -> 40, >=50 -> 25, >=10 -> 12) + rating quality (4.3-4.9 -> 30, >=4.0 -> 20, >=3.5 -> 10) + velocity (>=4/mo -> 20, >=1 -> 10) + replies (>=50% -> 10) where a review sample exists.
• Freshness 22%: opening hours present (and a recent post where sampled).
• Answer visibility 14%: richness of common service keywords in the review text.
• Excluded: an entity-authority dimension used in the full internal score is left out here because it has no reliable public signal. Bands reported: 75-100, 60-74, 40-59, under 40.

Review theme classification

• Model: Google Gemini (gemini-2.5-flash), temperature 0.1, strict-JSON output.
• Each review was assigned exactly one theme from: results, staff/bedside manner, price/value, pain/comfort, wait time/scheduling, cleanliness, booking/communication, upselling, before/after outcome; plus a sentiment of positive, negative or neutral.
• Prompt (verbatim template): "Classify each numbered Med Spa Google review into ONE primary theme from this list: [themes]. Also give sentiment positive|negative|neutral. ... Return STRICT minified JSON: {items:[{i,theme,sentiment}]}". Reviews were sent in batches of 20.
• Reproducibility: language-model classification is not bit-for-bit repeatable, and per-review labels from the published run were not stored. The review corpus is preserved, so the classification can be re-run and human-spot-checked to validate the theme findings (notably booking & communication).

Known limitations

• Early-look sample size (115) gives roughly a +/-9% margin at the national level; metro-level figures are directional.
• Discovery is seeded on 20 large metros, so it under-represents small-town med spas.
• Review velocity is capped by the per-profile review pull limit (see above).
• Booking/reservation signals are not reliably exposed by the Google Places API and were not scored.
• Data-handling note: a later, larger refresh reused the same working table and overwrote the stored detail records for 29 of the 115 rows; completeness and visibility figures are therefore reported on the 86 untouched rows (n shown next to each figure).
• No business is named alongside any figure in published outputs; groups smaller than 10 are not published.