An opportunity is a review candidate

A recommended product without a confirmed match can suggest assortment potential, but it can also be a variant-mapping problem. An existing product with weak recommendation support can suggest content or positioning work. A recommendation summary should identify the actual issue rather than putting every unmatched entity into a procurement list.

Frequency evidence and review suitability

The legacy product assessment combines relevant recommendation rate, positive-answer volume, distinct questions, provider coverage, and top-three support. The reviewed weights are 30%, 25%, 20%, 15%, and 10% respectively, with bounded normalisation before combination. This frequency-evidence score is different from the raw percentage in the recommendation-frequency column. Review suitability then incorporates catalogue uncertainty, category information, and freshness. Neither is a revenue forecast.

Small samples are capped

A narrow sample can look impressive as a percentage. The reviewed product assessment marks sparse evidence when there are fewer than four recommendation answers, fewer than two known questions, or at most two answers from fewer than two providers. Sparse evidence is capped at 38 for its evidence score and 48 for review suitability. Other opportunity views expose independent evidence and readiness labels; do not interpret all score-looking numbers as one universal confidence measure.

Readiness needs more than repeated answers

Identity must be resolved where required, and catalogue-dependent decisions need a sufficiently current catalogue. The reviewed opportunity confidence layer treats a known catalogue age of fourteen days or less as fresh. An unknown sync date is not automatically fresh. Provider agreement across questions and repeat consistency are more informative than simply rerunning one easy question many times.

Useful opportunity types

A board can distinguish assortment potential, mapping work, strengthening an existing offer, content opportunities, competitive advantage, and observations that need more evidence. Keep the associated source links and sample sizes. A weak observation can be retained for monitoring without inflating it into a high-priority recommendation.

External prices and freshness

An observed shopping price or product-generation status is supporting context. Compare the same variant, currency, tax treatment, shipping destination, bundle, and observation time. 'Availability unclear' should remain unclear. No external price observation proves that the merchant can buy at that price, obtain supply, or sell profitably.

Merchant decisions have their own lifecycle

Review, bookmark, dismiss, or close a finding using the supported board controls. A dismissal reason explains the business decision and is not a negative model observation. New evidence can be flagged without erasing the prior decision. When transferring a finding to Verbesserungen, preserve the evidence and distinguish a proposed task from a completed merchant action.

An evidence-first review checklist

Before acting, confirm the exact entity and variant; inspect positive answer passages; read the independent question/provider/run base; check the catalogue state and timestamp; verify any external offer; and record the actual merchant decision. Revisit a compatible panel after the change. An unchanged store cannot support a claim that ShopFlare caused a measured improvement.

How Commerce Intelligence discovers products and retailersCatalogue matching, coverage buckets, and the CI ScoreFrom a finding to a task and a measured follow-up