BuildMatchMethodology
How we calculate contractor ratings.
BuildMatch publishes one aggregate rating per contractor. It is drawn from multiple public review platforms and statistically adjusted so contractors with very few reviews don't unfairly outrank veterans. This page documents exactly how the number you see is computed.
3
Source platforms aggregated
m = 20
Bayesian shrinkage parameter
Hourly
Maximum manual refresh rate
Where the data comes from
Public review platforms — fetched directly.
Where the data comes from
Contractors link their own profiles on the platforms below. BuildMatch fetches the public data directly from each platform's API or business page — we don't scrape, and we don't accept self-reported ratings.
Yelp Fusion API
Aggregate rating, review count, and a sample of recent review excerpts (up to three) plus business metadata.
Google Business Profile
Link mode — Google's public data API is restricted, so we surface the linked profile and exclude it from the numerical aggregate unless it returns a rating.
Trustpilot
Aggregate rating and review count via the public Business Units API when available; otherwise link mode.
The formula
The math, in one line.
The Bayesian shrinkage formula
Bayesian shrinkage
adjusted = v · R + m · Cv + m
A contractor with three five-star reviews is treated as less certain than a contractor with three hundred four-star reviews. The formula pulls low-volume averages toward the platform-wide mean, so a tiny sample size doesn't produce a top spot on its own.
- R
- The contractor's average raw rating across linked sources.
- v
- The number of reviews they have.
- C
- The global average rating across all contractors — the "prior."
- m
- A confidence parameter that controls how much we shrink toward the prior. BuildMatch uses m = 20.
A worked example
Three reviews vs three hundred.
A worked example
Suppose the platform's global mean rating is C = 4.5. Compare two contractors:
01 · Newer contractor
Three five-star reviews.
- R
- = 5.0
- v
- = 3
- C
- = 4.5
- m
- = 20
(3 × 5.0) + (20 × 4.5) = 105
3 + 20 = 23
≈ 4.57
Their adjusted rating drops from 5.0 to about 4.57 — much closer to the prior of 4.5 because we don't have enough reviews yet to be confident.
02 · Established contractor
Three hundred reviews averaging 4.7 stars.
- R
- = 4.7
- v
- = 300
- C
- = 4.5
- m
- = 20
(300 × 4.7) + (20 × 4.5) = 1500
300 + 20 = 320
≈ 4.69
Their adjusted rating barely moves — from 4.70 to 4.69 — because 300 reviews is a strong enough signal that we don't shrink it much.
Without this adjustment, the newer contractor would beat the veteran on rating alone. With it, the veteran's much larger evidence base wins out — which matches how a careful homeowner would actually read the two profiles.
Multiple sources
How sources combine into one number.
How multiple sources combine
When a contractor links more than one platform, the published rating is a weighted average across sources, weighted by each source's review count. A source with 200 Yelp reviews has more influence on the published number than a source with 12 Trustpilot reviews — which mirrors how much evidence each platform actually carries.
Sources that have been disabled by an admin, or that returned no rating from their last sync, are excluded from the aggregate.
A rating is not a one-time snapshot. Whenever a sync runs — on initial setup, on a contractor's manual refresh (rate-limited to once per hour), or when an admin triggers a sync — the latest external counts and ratings flow through the same formulas.
Why this matters
Rating is one feature of nine.
Why this matters for matches
The Bayesian-adjusted rating is one of nine features in BuildMatch's contractor ranking algorithm. Geographic proximity, category fit, response time, availability, and reliability metrics (acceptance, completion, and dispute rates) all contribute alongside it.
In other words: rating is a key input, not the only one. Two contractors with similar adjusted ratings can still rank differently based on how close they are to the project, how fast they respond, and how reliably they complete jobs.
Frequently asked
Common questions.
Common questions
Why is the rating I see on BuildMatch slightly different from what I see on Yelp?
What if a contractor has zero linked reviews?
Can contractors edit or delete reviews?
Why use a Bayesian prior instead of just averaging?
How often do ratings update?
Is the published rating the only thing that drives matching?
Ready when you are
Match with contractors whose ratings are earned.
Tell us about your project. The contractor we introduce ranks against the same statistical adjustment described above — no inflated stars.
We will never share your information with third parties and will only contact you about your project.