What the UK fake-review research can and cannot tell your reputation team

Reputation management benchmark research should establish whether an external number is comparable before anyone uses it as a target. An England business needs the original population, measurement period, unit and collection method. A familiar percentage from a UK report may still be unsuitable for a branch dashboard.

The evidence below concerns a government-commissioned study of fake product reviews. It provides a dated research reference and a worked comparability decision, rather than an invented national norm for ratings, reply speed or conversions.

What to take away

What the published research measures

The Department for Business and Trade published its fake-review research record on 25 April 2023. It reports that 11% to 15% of reviews in three product categories on the e-commerce platforms studied were likely fake. Those categories were consumer electronics, home and kitchen, and sports and outdoors.

The unit is a proportion of product reviews estimated to be fake. It is not a percentage of businesses, reviewers or customer complaints. The geography is UK e-commerce research, not an England-only branch sample. The publication date must remain attached to the number; a 2027 planning document does not make the observations contemporary.

The executive summary describes a predictive model applied to 2.1 million reviews across nine platforms. It separately describes an online shopping experiment involving 4,900 UK participants. These are different research components: a modelled prevalence estimate and an experiment about consumer choice. They should not become one general review-performance score.

Decide whether the figure belongs in your comparison

The original publication says its findings do not extend to services purchased online or the impact of inflated star ratings. For a service firm, the defensible decision is to retain the report as background and reject it as a direct performance benchmark. A numerical value can be well sourced and still fail the comparison.

Design a useful local reference series

Begin with comparable records from the business's own approved branch population. Define which platforms, dates and review types are included. Preserve branch-level results before constructing an aggregate so changes in branch mix remain visible.

If benchmarking response handling, use the same start event and time unit across branches. State whether the series includes only answered reviews and display unanswered cases alongside it. A supplier's advertised response time cannot substitute for a measured reference series with matching definitions.

For a peer study, request the selection method, number of participating businesses, sector mix, coverage dates and exclusions. Find out whether participation was voluntary and whether the publisher sells services to the businesses measured. Such questions are proposed due-diligence steps, not assertions about any named report.

Avoid labelling a branch above average until the calculation is reproducible. Averages of branch averages weight branches equally; pooling underlying records weights them by record volume. Choose the weighting for the decision and disclose it before inspecting the result.

Keep an evidence card with every benchmark

Store the original figure, page or table location, publisher, publication date, access date and population. Add the permitted use and the comparison you rejected. Retain ranges as ranges when the source does not support a single preferred estimate.

Before adding a benchmark to the next report, ask whether its unit can be reconciled with the dashboard measure without extra assumptions. If the answer is no, use an explanatory note and commission the missing research. Leaving a target unset is preferable to presenting a product-study estimate as measured performance for an unrelated English service.

Before you act

Common questions

What does the fake-review research actually measure?

It reports that 11% to 15% of reviews in consumer electronics, home and kitchen, and sports and outdoors on the platforms studied were likely fake. The unit is a proportion of product reviews, not businesses, reviewers or complaints, and the geography is UK e-commerce.

Why can the headline figure not be used for an England service business?

The research covers specified consumer product categories and a UK platform context, so no England estimate can be extracted. The original publication says its findings do not extend to services purchased online or the impact of inflated star ratings. Retain it as background only.

How should a reputation team build a defensible local benchmark?

Start with comparable records from your own approved branch population. Define platforms, dates and review types, preserve branch-level results before aggregating, and use the same start event and time unit. For peer studies, request selection method, participant numbers, sector mix, coverage dates and exclusions.