A star rating is the first thing most shoppers look at and the easiest thing to manufacture. But reviews leave a trail over time, and authentic reviews behave differently from purchased or incentivized ones. The rating isn't the signal — the shape of how it was built is.
Genuine reviews track sales. Bursts that don't track sales, reviews older than the listing, and merged histories are the fingerprints of intervention.
Patterns worth watching
- Bursts without sales — a sudden cluster of five-star reviews with no corresponding jump in sales rank suggests the reviews arrived by arrangement, not from buyers.
- Reviews older than the listing — a common tell of a "merged" listing, where an old product's review history is grafted onto a new one to inherit its rating.
- Stockout reviews — reviews arriving while a product was out of stock and couldn't be purchased.
- Rating jumps that don't match volume — the average moving faster than new reviews should allow.
- Review purges — when a marketplace removes reviews it judges inauthentic, the count drops. A falling review count is one of the few visible traces of that.
How LookCloser builds this over time
Most of these patterns are only visible historically — you need to compare the same product across dates. LookCloser records a dated observation of every product's rating, review count, and price on each check, building its own longitudinal history. A review count that falls between two observations, or a price that swings before a "discount," becomes a documented change rather than a snapshot.
What this check does not claim
- A high rating is not evidence of quality; a fast-growing one is not proof of fraud.
- Patterns are recorded as observations with dates — the reader weighs them.
- Reviewer-level analysis (verified-purchase clusters, reviewer overlap) needs data sources we name honestly when we don't yet have them.
See this check on a real product.
Every LookCloser record runs this check with named, dated public sources — and shows the gaps too.