How to read our commerce reviews assessment method

Agreements stop at written terms

The contract row records notice periods and data portability, weighted like each other area. It does not assess conduct during disputes because the available material cannot establish how a supplier behaves then. Retailers can consult contract clauses for separate checks; those checks do not alter this area's score.

Connections remain a separate category

Connections to ecommerce platforms and helpdesk tools receive a distinct area score, even when a retailer would rather rank them alongside analytics or another feature. Equal weighting describes this assessment only; it does not prescribe shortlist priorities. The comparison framework can help readers interpret separate results without merging unlike criteria.

Collection measures the process

Review collection is checked through request timing, channel choice and suppression rules. These are process details, not outcome measures; supplier descriptions do not independently verify accuracy or prove results. Teams may use a retail quality checklist for workflow checks, but it does not add evidence to the supplier's collection score.

Analytics depends on the record

Analytics is scored through reporting depth and export, using only the material available to support those checks. A feature description alone is not evidence of results in a retailer's operation. Retailers may set different measurement priorities, but those preferences do not combine this area's score with the other categories.

Display features do not establish results

The display category records material on star widgets, structured data markup and syndication to retail listings. It does not treat promotional wording as proof of an outcome or establish that a display feature produced a specified result. Separate advertising guidance may inform a retailer's own review of promotional claims.

Moderation evidence has a narrow reach

Moderation scoring uses evidence about how quickly fake or abusive reviews are caught. The score is neither a verified fraud rate nor the result of examining an internal moderation queue.

It describes the available material about the stated process, not fraud across a supplier's customer base or access to internal systems.

Labels distinguish claims from support

The method labels supplier claims, evidence and editorial judgement separately. A provider's statement is a claim; evidence can include a published rule, filing, dataset or official note. Editorial judgement is reasoning drawn from sources, not independent proof, so a claim on its own cannot raise an area score.

The source and date remain visible, and judgement is identified as reasoning. A supplier's own statements and its source records do not convert a claim into independent confirmation. Advertising claims and possible future rulings are not predicted by scoring; relevant evidence remains separate from analysis.

Published research about customer engagement and user-generated content is secondary evidence, not proof that a product performs well. Promotional wording may be recorded as a supplier claim, but it cannot establish performance by itself. Keeping those labels distinct prevents a claim from serving as confirmation of its own accuracy.

A number needs a trail

Every figure is dated and paired with a named source. If the source cannot be located, the number is removed or labelled illustrative. This source-and-date rule keeps the record traceable and helps readers distinguish a published rule or supporting material from editorial analysis.

For example, a monthly software charge of £400 over twelve months totals £4,800 before implementation; it is illustrative, not a market quote. Apply the same source-and-date discipline to cost figures, giving both source and date or clearly marking an amount as illustrative.

Commercial unknowns remain outside scoring

Price negotiations, sales demonstration quality and implementation speed remain outside the score because available access cannot verify them. Awards, logos and directory placements are excluded too: badges are marketing assets, not performance evidence. Supplier pages can log claims, but they do not independently confirm a provider's performance.

For an England-based homeware retailer with 80 staff, example monthly quotes of £300, £400 and £550 give annual software costs of £3,600, £4,800 and £6,600. An example £2,000 integration and £150 monthly moderation retainer stay separate cost lines.

The shortlist follows area results rather than adding them together. If the cheapest option scores poorly for fraud controls and syndication, the assessment records a risk instead of calling its lower price a saving. The provider examples apply the method to named providers; this article names none.

How to read an area score and the total

Each of the six areas receives a score from 0 to 2. A 0 means no usable evidence was found. A 1 means a supplier claim exists with some supporting source, but gaps remain. A 2 means specific, dated evidence from a source covers what the area examines. A score of 2 describes the available material, not the supplier's performance.

When sources conflict, the area takes the lower score and the reason is recorded. Equal weights mean the six scores add to a total out of 12, but the total is a starting point only. A total of 8 built from 1s and 2s across all six areas differs from a total of 8 made of four 2s and two 0s, such as 0s on collection and moderation. Read the breakdown first, then the total.

Updates respond to source changes

Each scored area is checked against its current source at least annually, with an earlier review when a relevant rule or platform note changes. Dates stay attached to figures, and changes to the assessment follow a relevant source update rather than an unsupported assumption.

A score more than twelve months old is marked stale, not silently refreshed. Published platform release notes take precedence over blog commentary as the record of platform changes. This leaves the last source check visible without implying that an unverified change has occurred.

No trial accounts, sandbox assessments or interviews with named customers are used. Internal moderation queues, actual fraud rates and contract behaviour during disputes remain unassessed. Treat a total as a starting point for supplier due diligence, not as a replacement for checks your own team carries out.

A supplier claim is logged, dated and cross-checked before it informs judgement; a claim alone receives no score. If a reader thinks relevant material is missing, they can point us to the source. Evidence that changes an area result leads to a correction in the body text, while a relevant source change brings a new update date.