Marketing & GrowthFree
Competitor Review Mining and Theme Analysis
Each complaint theme faces three gates: has it already decayed, do your own reviewers say it too, and can you point at a fix.
Marleigh read 1,284 reviews across three competitors and 412 of its own. 731 of the competitor reviews carried at least one complaint, which produced 1,123 complaint mentions and 14 themes at ten mentions or more. Three themes survived. The loudest theme by volume, reporting being slow on large sites at 187 mentions, died because Marleigh's own reviewers raise it at 12.9% against the competitors' 14.6%. A 1.7 point spread is not a claim.
Every theme passes three gates in order. Gate one compares its share of recent mentions against its share of lifetime mentions, so a theme the vendor already shipped a fix for drops out. Gate two reads the same theme in your own reviews and keeps only the ones where the rate gap is real. Gate three needs a specific product mechanism you can name, or the theme dies whatever its volume. Five died at the first gate, four at the second and two at the third.
Written for whoever hands a competitive claim to a sales team and then has to defend it. The competitor teardown covers one rival's positioning, pricing and disclosures, while the monitoring pack treats review volume as a change signal rather than a corpus to read. Surviving themes belong in the battlecard pack before a rep repeats them, and the ad and messaging teardown is where a claim becomes copy that has to hold up.
Why the corpus is not the population
A review corpus is shaped by how each vendor asked for it. The FTC's endorsement guides answer this directly: asking for reviews only from customers you think are likelier to be happy would be misleading if it substantially skews the favorability of the reviews. A general solicitation is fine, so a vendor running a review drive gets volume that dilutes its complaint rate. That is why every theme is a rate per review rather than a count, and why the counts are held next to yours.
Read the third-party platform, not the vendor's own review page. The FTC's rule on consumer reviews makes it unlawful to misrepresent that a site's displayed reviews represent most or all the reviews submitted when reviews are being suppressed for their rating or sentiment. It also lists the sentiment-neutral grounds on which a vendor may legitimately withhold one. Both readings point the same way: a vendor-hosted corpus is a subset by design, so the counting has to happen where the vendor does not control display.
The gates are worth running because volume and claimability point in opposite directions. On the Marleigh set the 14 themes covered 999 of the 1,123 complaint mentions, 89.0%, and the three survivors carried 222 of those, 22.2%. The other 777 mentions sat behind themes that were decaying, shared or unaddressable. A frequency-ranked list would have led with the two largest themes. Marleigh has one of them at nearly the same rate, and the other fell from 12.6% of lifetime mentions to 2.8% of recent ones.
How it works
Read both corpora
Their public reviews and yours, with a recent window marked off inside each.
Cluster the complaints
Every complaint becomes a theme, counted per review rather than as a raw total.
Run the three gates
Decay, then your own reviewers, then whether a named product mechanism actually addresses it.
Write only survivors
The claims that cleared all three gates, each with the reviews and the mechanism.
What you get
- Every complaint theme with its mention count, its severity read and the reviews behind it.
- A decay ratio per theme, comparing its recent mention share against its lifetime share.
- The same themes counted in your own reviews, with the rate gap in points.
- Each surviving theme tied to a named product mechanism, or dropped for lacking one.
- The gate that killed every dead theme, so nobody relitigates a dropped one.
- A Sheet of themes by competitor and a Doc of the claims you can defend.
Common questions
Why not just rank complaints by frequency?
Because the ranking inverts under the gates. Marleigh's two largest themes, at 187 and 142 mentions, both died. The first went because Marleigh's own reviewers raise it at 12.9% against 14.6%. The second went because its share of recent mentions is 2.8% against a lifetime 12.6%. The three survivors ranked third, sixth and ninth. Loss reasons reorder the same way in the lost deal patterns.
How do you know a theme has already been fixed?
You do not know, and the gate does not claim to. It measures whether the complaint is still arriving. A theme at 12.6% of lifetime mentions and 2.8% of the last two quarters has stopped being current, whatever the cause. Claiming it invites a prospect to check and find it untrue.
Why read our own reviews at all?
Because it is the check that stops an indefensible claim. Four of Marleigh's nine surviving themes died here, including implementation running long, which its own reviewers raise at 10.0% against the competitors' 6.9%. Attacking a rival on something your reviewers say more often is a claim that turns around on the call.
What counts as mechanically addressable?
A specific thing in your product that changes what the reviewer complained about, named and checkable. Two of Marleigh's five remaining themes failed this: no sandbox for testing changes, and search not covering custom fields. Both are real complaints about the competitor. Neither is something Marleigh could point at and demonstrate.
Do vendors game their own review profiles?
They shape them legally, which is why the counting is a rate rather than a total and why it happens on third-party platforms. The FTC treats soliciting reviews only from likely-happy customers as misleading when it skews the ratings, and its rule polices presenting a filtered set as the full one. Both distort volume more than sentiment.
How many reviews do you need?
Enough that a theme at ten mentions means something. Marleigh's set was 1,284 competitor reviews across three rivals and 412 of its own, and 56.9% of the competitor reviews carried a complaint. Below roughly 150 reviews per competitor the recent window gets too thin to read a decay ratio from.
What actually gets delivered?
A Sheet with one row per theme and competitor, carrying mentions, rate, recent share, decay ratio, your own rate and the gate outcome. A Doc with the surviving claims, the evidence under each, and the themes to leave alone with the reason. Copy for reps belongs in the battlecard pack.
Competitor Review Mining and Theme Analysis
Fill in the form and your workspace opens with the work already underway.