ProFix Directory

How we score reviews

Plain-English explanation of the math behind every ranked list on ProFix Directory.

Where the data comes from

We pull each pro's average rating + review count from the public Google Places aggregate. We don't write reviews ourselves. We don't accept reviews submitted directly to ProFix. The number you see is the same number any homeowner sees when they look up the business on Google Maps.

Why a 5.0 with 3 reviews doesn't win

A new business with 3 perfect reviews looks "5 stars" — but the sample is too small to mean much. A 4.8 across 200 reviews is dramatically more reliable signal. To prevent new (or fake) shops from gaming the rankings, we use a quality score that weights review volume:

score = rating × log10(reviewCount + 10)

This is a "Bayesian shrinkage" approach — pull thin-evidence pros toward the mean until they earn enough reviews to climb. Featured pros (paid placement) are pinned top-of-list only if their score would have ranked them top-10 anyway.

What we screen for

  • Sudden review bursts. 30 new 5-star reviews in 90 days = bought reviews. We flag and downweight these patterns.
  • License lapses. Every plumber/HVAC/electrician/gas pro is re-verified at the Ohio eLicense Center every 90 days. A lapse drops them off the list entirely.
  • Categorical mismatch. If a business shows up under "plumber" but its Google place type is "home_goods_store" (it's a parts retailer, not a service pro), we filter it out.
  • Geographic mismatch. Cincinnati or Cleveland businesses that show up in Toledo searches don't get listed under Toledo.

Star distribution histograms

On any pro page with 50+ reviews, we show an estimated 5/4/3/2/1-star breakdown. The estimate is derived from the public average + count using a Beta-distribution shape calibrated against home-services Google Places data. It's an estimate — the individual review breakdown lives on Google Maps. We're transparent about that on every histogram.

Where we'd improve

Our biggest gap right now: we can't verify which reviews are paid vs organic without machine-reading review-by-review. Future iteration: text-pattern detection on review content. For now, a sudden burst pattern is our only fake-review heuristic.

Read our pricing policy →
Emergency