Why commerce reviews need a documented method before 2027

What to take away

What a review method must record

A star average is an output. The method behind it is the product. Two retailers can show 4.3 stars and mean different things. One holds verified purchase data, a published moderation policy and a collection window. The other holds a widget and a hope.

A method states the sample. A 4.3 average from 1,200 verified reviews over 12 months is not the same as a 4.3 average from 18 reviews collected in one week. Record the verified purchase share, the collection window, the moderation policy version and any incentive disclosure.

Buyers behave like auditors. They read two-star reviews first, check whether the reviewer is verified, and look for a pattern in the dates. Reviews that cluster in a fortnight and then stop send their own message.

For teams choosing software, the audit trail is the differentiator. A platform that timestamps collection, stores consent and exports a moderation log is worth more than one with a prettier carousel. Our commerce reviews selection checklist in England sets out the questions to ask vendors before a contract is signed.

Name one owner. That person holds the method, the version history and the vendor contracts. Shared ownership is how methods drift, and drift makes an audit expensive.

The legal frame: fake reviews and consumer protection

The Digital Markets, Competition and Consumers Act 2024 received Royal Assent on 24 May 2024. The commencement regulations made under the Act set 6 April 2025 for the consumer protection provisions.

From that date the CMA can act directly on misleading commercial practices, including fake and incentivised reviews, rather than going to court for every breach. The live provisions include the unfair commercial practices regime, the banned practices list and the CMA's power to impose fines of up to 10 per cent of global turnover.

Three behaviours carry the most risk: buying reviews, suppressing negative reviews selectively, and presenting incentivised reviews without disclosure. Each is a documentation problem before it is a legal problem. The evidence needed to defend a practice is the same evidence needed to run it properly.

Sector regulators add another layer. Advertising claims that lean on review evidence can be challenged, and the ASA's independent reviews of its own decisions show how rulings are revisited on appeal. That matters for due diligence: a competitor's published claim may have been tested, and the reasoning is public.

Build a simple evidence file. For each review campaign, keep the brief, the supplier contract, the disclosure wording and the date it went live. If a complaint arrives eighteen months later, that file is the defence.

Incentives need a written method, not a policy sentence. Record the incentive type, value, campaign and date. Store the exact disclosure wording shown to the reviewer. Flag every incentivised review in the database so reports can include or exclude them.

Data protection in review collection

Reviews are user-generated content with personal data attached. A display name, an order reference, a photo or a location hint can identify someone when combined. Collecting any of them means handling them under UK data protection law.

The Data (Use and Access) Act 2025 amends parts of that framework. The ICO's Data (Use and Access) Act 2025 page explains the scope and the regulator's role. The practical points are unchanged in substance: have a lawful basis, set a defined retention period, honour erasure requests, and publish no more personal detail than the review requires.

The ICO does not set one universal retention period for reviews. Its storage limitation principle asks you to justify the period you choose.

HMRC expects records supporting tax returns to be kept for 6 years, so review data tied to a transaction may need that long.

Published review text can often stay while the product page is live, but reviewer email, order reference and IP address should be deleted once the dispute window closes.

Consent for marketing use is separate from consent to publish. A customer who agrees to a review appearing on a product page has not automatically agreed to appear in a paid social ad. Keep those permissions in separate fields, with dates.

Test the erasure route before you need it. Submit a request through your own process and time how long it takes to remove a review from the platform, the search index and any syndication partner. The answer belongs in the method document.

Agencies that collect or respond to reviews sit inside that processing chain. Our commerce reviews agency reviews in England guide covers the contract terms and data clauses to check before work starts.

Platform updates change what you can measure

Review features are not static. Google maintains release notes for the Places API. The Places API (New) release notes record changes to place data and review-related fields over time. One concrete example is the places.reviews field in Places API (New).

A field mask must request it, and the legacy Places API exposes reviews through a different object. Monitor the notes if your local listings feed into a review strategy. A field can be added, deprecated or re-scoped, and any dashboard built on it inherits that change.

The lesson generalises across vendors. Any review metric that arrives through a third-party API is a dependency. Record the API version, the field name and the date you captured the figure. When the number moves, you will want to know whether customer sentiment changed or the schema did.

Write deprecation cover into contracts. Ask the vendor what notice they give if a feed, field or integration is retired, and what they do with historic data if it is.

Change control is the practical answer. Keep a one-page log of every review data feed you depend on, who supplies it and when it was last checked. Review it quarterly.

Sizing the market with official data

Review volume should be read against demand, not in isolation. The ONS digital economy statistics cover internet industry activity and online retail. The ONS digital economy data is a defensible reference point when you need to argue that a review programme is under- or over-resourced.

The relevant ONS series are Digital Economy Statistics and the Retail Sales Index: Internet Sales.

A workable internal benchmark is reviews per hundred orders. Typical planning ranges are 1 to 3 reviews per hundred orders for general retail, 2 to 5 for fashion, and 2 to 4 for consumer electronics. For hospitality the range is 5 to 15, and for automotive it is 10 to 20.

Test those against your own order data. Those bands are not targets. They are starting points for a monthly ratio that you track by category.

Pair it with a quality measure. The share of reviews containing a photo, a product-use detail or a specific complaint tells you more than the average alone. Ratings give you the mood. Content gives you the cause.

Set the benchmark before you need it. A ratio agreed in advance turns a quiet quarter into a conversation about demand rather than a debate about the data.

How to compare review providers

Start with the data model, not the demo. Ask where review content is stored, who owns the export, and what happens to it if you leave. A provider that cannot answer within a day is a provider with a migration risk.

Test the moderation workflow against a real edge case. A review naming an employee, a review containing a medical claim, a review from a customer who later cancelled. Run those through the vendor's process and see what the audit log records.

Agree the scoring criteria before the shortlist is drawn. Weighted criteria decided in advance stop the final choice being made by the best presenter on the day. Keep the completed scorecard, because it explains the decision months later.

Check the exit terms next. Ask for the data export format and the notice period. Find out what leaving costs. A platform holding years of review content should return it in a usable file.

Price the whole thing. Per-review fees, per-order fees, syndication charges and syndication removal charges are where budgets drift.

An illustrative sum: 2,000 reviews a year at a per-review rate of £1 is £2,000 before platform fees. A platform licence of £500 a month is £6,000 a year. Add a syndication removal fee and the total can move by thousands.

Our commerce reviews best providers in England comparison sets out the cost lines that are usually quoted separately.

Run the shortlist through the same questions twice, once before pricing is shared and once after. Vendors answer differently when they know the budget.

Public examples include Trustpilot, Google Business Profile, Feefo, Reviews.io and Yelp. Trustpilot publishes a TrustScore and review invitation tools. Google Business Profile shows reviews and owner replies.

Feefo collects verified buyer feedback. Reviews.io offers review collection and Google Seller Ratings integration. Yelp hosts business reviews and review response tools. Check each against the same evidence rules.

Scoring and comparison methods that hold up

A comparison method needs four properties. It must be reproducible, weighted, versioned and bounded.

Reproducible means another analyst with the same inputs reaches the same score. Criteria are weighted to reflect what buyers care about, not what is easy to measure. Versioned means the method has a number and a date. Say what the score does not cover, and you have bounded it.

A simple weighted model works for review software.

Collection and coverage might carry 30 per cent, moderation and audit trail 25 per cent, data protection and consent handling 20 per cent, integrations 15 per cent and price predictability 10 per cent.

Score each criterion out of five and multiply by the weight. The table below works the example for two vendors.

Vendor B wins on weighted total even though Vendor A scores higher on data protection. Change any weight and the result can change, which is why the weights belong in the dated method before the shortlist.

Those percentages are illustrative. The structure forces the weighting decision into the open before the shortlist is drawn. Our commerce reviews review methodology in England article applies the same model step by step.

Publish the table, the evidence source and the date so someone who was not in the room can check it.

The steps are simple. Set the question and scope first. Choose criteria and weights before the shortlist. Define a 0 to 5 scale with written anchors for each score. Score independently and record the evidence source and date. Calculate the weighted total, publish the working, then version the method and set a review date.

Planning for 2027 without a statutory deadline

Two shifts are visible already. Verification is becoming table stakes, with purchase-linked reviews treated as the default and unverified reviews discounted by readers. Review content is also being reused across surfaces, from product pages to paid media, which pushes consent design up the priority list.

Enforcement is getting cheaper for regulators and more expensive for firms. Direct CMA powers, sector rulings and public reasoning raise the cost of a weak review policy.

Budget for reuse rights as a line item. If review content will appear in paid media, the cost of securing and recording that permission belongs in the campaign plan, not in a legal afterthought.

For planning assumptions into 2027, our commerce reviews trends and outlook for England in 2027 piece sets out the scenarios worth modelling, including verification-led collection and consent-scoped reuse.

Treat 2027 as a planning horizon, not a statutory deadline. Contract renewals, platform migrations and data retention cycles take months. A method written now can be tested against at least one full reporting year before the next buying round.

2027 checklist for review operations

Use this before your next vendor renewal or campaign launch. Add a date to every line.

Common questions

How many reviews should a UK retailer expect per hundred orders?

There is no official target. The ONS digital economy data gives context on online retail activity, but review rates vary by category and delivery experience. Typical planning ranges are 1 to 3 per 100 orders for general retail, 2 to 5 for fashion, 5 to 15 for hospitality and 10 to 20 for automotive.

Track your own ratio monthly and treat sudden changes as a signal, not a score.

Are incentivised reviews allowed in the UK?

They can be, provided the incentive is disclosed and the review is not presented as independent when it is not. The Digital Markets, Competition and Consumers Act 2024 strengthened enforcement against misleading review practices, so disclosure should be explicit and recorded. The consumer protection provisions commenced on 6 April 2025.

Do we need consent to publish a customer's review?

You need a lawful basis for processing the personal data in the review, and publishing it is processing. Consent is one route, but it must be specific and separable from other permissions such as marketing use.

How often should a review method be reviewed?

At least annually, and whenever a platform, regulator or law changes. Version the method with a date so that historic scores can be explained against the rules in force at the time. The ICO storage limitation principle requires a defined retention period, so check that period at the same time.

What changes for commerce reviews before 2027?

The main change is enforcement and evidence. The CMA's consumer protection powers commenced on 6 April 2025, so they are already live. Platforms keep changing review fields. A method dated before 2027 gives you a baseline.