How to score a commerce reviews agency before you sign

What to take away

Score five documentary areas before you signcollection, moderation, disclosure, reporting and contract terms. A 100-point sheet makes the final call defensible.
A desk review cannot test widget speed, consent rates or dashboard usability. Fund a paid pilot of six to eight weeks on one product category.
Treat uplift percentages as advertising claims. Ask for a dated study with a named baseline, or record the figure as unverified.
The Digital Markets, Competition and Consumers Act 2024 gives the CMA powers over fake reviews, with fines of up to 10% of global annual turnover.
Pricing sits outside a method review. Split software subscription, retainer and internal hours before comparing quotes.

The five areas a desk review scores

Assess collection first. Ask how the agency invites reviews by email, SMS, on-site prompt or printed receipt, and what it offers in return.

Incentives are the fault line. A written policy must say who may be rewarded, and whether a voucher can be tied to a star rating.

Moderation comes second. Filtering negative reviews without a published standard is the main risk. Request the log format and the appeal route in writing.

Disclosure comes third. Widgets that set cookies or read device storage need a lawful basis, as the ICO guidance on storage and access technologies sets out.

Reporting comes fourth. A monthly star average is not evidence. Ask for review-level exports and a written definition of a verified buyer that names its data source.

Contract terms close the assessment. Ownership, exit rights and data portability belong in writing; check the seven clauses to watch in commerce reviews commercial contracts if your template is not standard.

Five areas to assess

Collection mechanics and incentive policy
Moderation log format and appeal route
Disclosure and cookie consent basis
Reporting exports and verified buyer definition
Contract ownership, exit and portability

Example: a 100-point scoring sheet

Give two agencies the same questions and score the answers without knowing which brand wrote them.

100-point scoring sheet

Scoring area	Evidence to request	Points
Collection and incentives	Invite policy, reward rules	20
Moderation and appeals	Log format, escalation path	20
Disclosure and consent	Cookie wording, lawful basis	20
Reporting	Review-level export, verified buyer definition	20
Contract	Ownership, exit terms, portability	20

Within each area, a published policy earns full marks. A written answer that matches it earns 12. A verbal promise earns 5. Nothing earns 0.

What the desk review cannot tell you

A method review reads documents. It cannot measure how a widget renders on a mid-range Android phone or how it interacts with your checkout.

It cannot verify storage security or how quickly policy-breaching reviews come down.

Assessed versus not assessed

Assessed in desk review

Collection method: Published policy, incentive rules
Moderation: Log format, appeal route
Disclosure: Cookie and consent wording
Reporting: Export format, verified buyer definition
Contract: Ownership, exit, portability

Not assessed

Collection method: Live invite performance
Moderation: Reviewer behaviour at scale
Disclosure: Actual consent rates
Reporting: Dashboard usability
Contract: Negotiated commercial terms

That gap is why the pilot matters more than the pitch deck.

Where the budget actually goes

Rates vary with catalogue size, order volume, and whether the agency supplies its own software or manages a platform you already pay for.

Split every quote into four lines: platform subscription, managed-service retainer, customer incentives, and internal hours for briefing and moderation.

Platform choice shapes the total. Trustpilot, Reviews.io, Feefo, Judge.me and Yotpo usually appear on shortlists. Some publish a free entry tier; others quote on request.

Ask for unit costs and a renewal date in writing. A commerce reviews budget template shows how to lay out those lines before you compare two quotes.

The rules that moved: DMCC Act 2024

The CMA's 2021 consultation on reforming competition and consumer policy set the direction, but it is superseded.

The Digital Markets, Competition and Consumers Act 2024 adds fake and misleading reviews to the banned practices list. It also gives the CMA enforcement powers, with fines of up to 10% of global annual turnover.

Ask how the agency detects incentivised, bulk-submitted or gated reviews. Inviting only happy customers, or hiding unhappy ones, is the classic breach.

Cookie and storage consent for widgets sits with the ICO. Uplift claims sit with advertising rules, so read the ASA advertising codes before accepting a pre-cleared headline.

England, Scotland, Wales and Northern Ireland share these rules, so an England-only shop faces the same CMA tests.

Run the pilot in five steps

Pick one product category with enough order volume to generate reviews inside the window.
Record a baseline for review count, request volume and product-page conversion.
Run the pilot for six to eight weeks. A single month rarely covers reminder cycles or delivery variance.
Set a floor before you start. Aim for at least 100 completed reviews in the test category.
Score the result blind, then pick the attribution method to use when you read the numbers, whether that is last click or a holdout test.

Common questions

What does an agency review actually assess?

Published methods, moderation logs, disclosure wording, report formats and contract terms. Live widget performance is not covered.

Are agency conversion claims trustworthy?

Only when the agency publishes a dated study with a named baseline and a method you could repeat. Otherwise treat the figure as a claim.

Which regulators apply?

The CMA enforces fake and misleading review rules under the DMCC Act 2024. The ASA covers advertising claims and the ICO covers consent for review widgets.

How long should a pilot run?

Six to eight weeks on one product category, with a review-count floor agreed in advance. A single month is usually too short.