How to turn customer reviews into claims
To turn reviews into claims, split each review into single statements, tag each with a theme and a date, keep the original wording, and count distinct reviewers instead of mentions. Done this way, twenty reviews become something you can compare against a vendor's own page.
Why “customers say…” is not enough
A sentence like “customers find Meridian hard to set up” hides how many people said it, when they said it and what they meant. A reader cannot check it, and a sales team cannot safely repeat it.
The method
- Collect reviews from public pages and record the date and the platform for each.
- Split each review into single statements. “Support was slow but the import tool worked” becomes two.
- Write each statement as a claim about that reviewer's experience: “A reviewer reports a three-week migration.”
- Tag each claim with a theme (onboarding, pricing, support, reporting) and a sentiment.
- Keep a short verbatim excerpt next to the claim.
- Count distinct reviewers per theme, not total mentions.
A worked example
Fieldnote's team reads 18 recent reviews of Meridian. They produce 31 claims. Eleven concern onboarding, from seven reviewers. Of those, four describe setup as quick, three describe it as slow, and the slow ones all mention more than 30 client accounts.
| Claim pattern | Distinct reviewers | Notes |
|---|---|---|
| Setup took under a week | 4 | All agencies under 20 clients |
| Setup took three weeks or more | 3 | All agencies over 30 clients |
| Support helped with import | 2 | Both also report slow setup |
The finding is narrower and more useful than “setup is slow”: the experience appears to depend on agency size. With seven reviewers, it is a lead to test, not a conclusion.
Where review analysis goes wrong
- Counting mentions instead of reviewers, so one long review dominates.
- Ignoring dates. A complaint fixed in a later release is context, not a current weakness.
- Treating reviews on a vendor-curated page as independent.
- Mixing reviewers of different company sizes into one average.
- Quoting a reviewer in a sales deck without checking that the excerpt is accurate.
Linking review claims to vendor claims
A review claim becomes useful when it is attached to a vendor claim as supporting, contradicting or contextual evidence. “Onboarding in a day” on Meridian's page is the vendor claim. The three-week review contradicts it for one segment. The changelog's bulk import adds context.
How many reviews are enough
There is no threshold that makes a review set representative. Reviews come from people motivated to write them, often at the extremes of satisfaction. Treat a theme backed by five or more distinct reviewers as a pattern worth testing in interviews, and report the count with every theme so the reader can weigh it.
Reviews and Sigblaze
In the private alpha, Sigblaze can retrieve public review pages found through research or added to a watchlist, extract claims and link them to evidence. Dedicated feeds for review sites and app stores are a later capability, and pages that need a login are not read.
This piece contains no external statistics. Fieldnote and Meridian are fictional companies, and the figures in the examples are synthetic.