Do not trust the average alone. Check how many reviews exist, who wrote them and in what role, how recent they are, whether incentives were offered, and what the written detail says. Treat stars as a prompt to read text, not as a verdict.
Start with the shape, not the mean
An average of 4.5 can come from fifty detailed reviews or from six short ones. Look at the distribution. A cluster of five star and one star reviews with little in between often signals polarized experiences or a review campaign. A broad spread with specific comments usually signals real use.
Look at who is reviewing
The most useful reviews come from people with your role and your scale. A developer in a small team and a security lead in a regulated enterprise want different things. Filter by company size and role if the site allows it.
Check recency
Products change quickly. A glowing review from several versions ago may describe a different product. Sort by newest and read the last twelve months.
Know the incentive rules
The FTC final rule says businesses cannot offer compensation or incentives conditioned on reviews expressing a particular sentiment, positive or negative. It also treats undisclosed insider reviews as a problem and bars suppressing negative reviews. Many sites offer gift cards for any review, which is different from a payment for a positive one, but it can still skew who writes. Check whether the site discloses its incentives.
The FTC page on reviews also notes that platforms that feature reviews should have processes to confirm reviews reflect genuine customer feedback. See whether the site describes such a process.
Read the text for specifics
Strong reviews name versions, languages, integration points and outcomes. Weak ones use words like "great tool" or "easy to use." Mark claims you can verify in a test, and collect them.
Ignore comparison grids built from stars
A grid that ranks products by average stars confuses popularity and fit. Products with larger user bases get more reviews. Products sold to a narrow audience may have few. Neither says anything about your needs.
Use ratings to build a test, not to decide
Turn recurring praise and complaints into test criteria, then run them. Our post on running a bake-off shows the steps.
A short worked check
Open a product page and sort by newest. Read ten reviews from people whose company size and role match yours. Highlight every sentence with a number, a version or a named integration. Then read the three lowest rated reviews, and look for a vendor reply. If the replies are specific and respectful, that tells you something about support. If there are no low reviews at all on a product with many reviews, ask why.
Where Safeguard fits
Safeguard offers SBOM and dependency scanning, reachability analysis and policy gates. Instead of reading stars, try the free scan at gold.safeguard.sh/scan.
Quick check
Is a product with few reviews worse?
Not necessarily. It may be newer or aimed at a narrower audience. A small sample just gives you less to go on.
Sources
- FTC announces final rule banning fake reviews and testimonials
- FTC: endorsements, influencers and reviews
Last reviewed: 19 October 2026.
