How We Rank Agencies
Updated:
Two numbers, two purposes
Every agency profile shows two independent figures:
- a star rating (1.0–5.0), which summarises what reviewers said;
- a Trust Score (0–100), which summarises what we could verify about the agency itself.
They answer different questions. An agency can have delighted clients and an unverified licence, or a fully verified licence and mediocre reviews. We deliberately do not blend them into one number, because that would hide exactly the tension a consumer needs to see.
The star rating: Bayesian weighted average
We do not display the raw arithmetic mean of review scores. Raw means are unstable at low volumes: one agency with a single 5-star review would outrank another with 4.8 across two hundred reviews.
Instead we use a Bayesian weighted average with a prior mean of 4.0 and a prior weight of 8 reviews:
displayed rating = (8 × 4.0 + sum of review scores) / (8 + number of reviews)
In plain terms: every agency starts as if it had eight reviews at 4.0 (roughly the market's expected level), and each real review pulls the displayed rating toward the actual review average. The effect:
- An agency with one 5-star review displays about 4.11, not 5.0.
- An agency with twenty 5-star reviews displays about 4.71.
- An agency with two hundred reviews averaging 4.6 displays about 4.58.
As review volume grows, the prior's influence fades and the displayed rating converges on the true average. Removed reviews (see the review policy) are excluded from the calculation entirely.
The Trust Score: components and weights
The Trust Score is built from verifiable signals, each with a published weight:
| Component | Weight |
|---|---|
| Licence verification | 20 |
| Identity / business verification | 15 |
| Verified-review ratio | 15 |
| Claimed profile | 10 |
| Review volume | 10 |
| Review recency | 10 |
| Responsiveness to reviews | 10 |
| Profile completeness | 10 |
| Moderation penalty | up to −25 |
Notes on the components:
- Licence, identity, claim are earned through the documentary process in the verification policy. No document, no points.
- Verified-review ratio rewards agencies whose reviewers backed their reviews with evidence — a strong signal against fake-review inflation.
- Volume and recency reward a sustained, current review record over a handful of old reviews.
- Responsiveness measures whether the agency, via its claimed profile, actually answers reviews — including negative ones.
- Completeness rewards a fully filled, current profile.
- Moderation penalty subtracts up to 25 points for policy violations attributed to the agency: fake or incentivised reviews, forged verification documents, abusive flagging.
The score is capped at 0–100 and recalculated whenever an input changes.
How default rankings are ordered
Category and comparison listings are ordered by a combination of Trust Score and Bayesian rating, with verification-backed signals dominating. Filters and sort options let you reorder by rating, review count, or recency yourself.
What never affects rating or rank
- Money. Sponsored placements are always labelled "Sponsored" and never affect ratings, Trust Scores, or organic ranking positions. See the advertising policy.
- Editorial preference. Editors cannot adjust a score. Errors are fixed through the corrections process by correcting the underlying data, identically for every agency.
- Claiming a profile, beyond its fixed 10-point component and the abilities it unlocks. Responding to reviews earns responsiveness points whether responses are apologetic or combative — we measure engagement, not tone.
Changes to this methodology
If we change the prior, the weights, or the components, we will announce the change on this page with the date and recalculate all providers at once. We never apply methodology changes selectively.