The Complaint Ladder
Most complaints decay. One kind does not, and it is the only one worth building against.
Customer complaints sort into four rungs. Three of them decay within a release cycle. The fourth survives, because fixing it would require the company to disagree with itself, and companies rarely do.
The four rungs
1. Expectation. The buyer misunderstood what they were getting. Tells you about the marketing, not the product. Decays as positioning tightens.
2. Execution. It broke. Tells you about that quarter. Decays, usually within weeks, because it is embarrassing and cheap to fix.
3. Design. It works exactly as intended and the intention is wrong. Does not decay. Fixing it requires reversing a decision someone senior defended, and often requires giving up the thing that makes the product good for its core user.
4. Identity. The complainer is not who the product is for. Tells you where the real segment boundary sits, which is usually not where the pricing page says it is. Decays only if the company changes who it serves.
The rule
Only rung three compounds, and only across time. A design complaint that appears in three consecutive release cycles is not feedback. It is a strategic commitment, disclosed accidentally, by the only party who cannot afford to disclose it deliberately.
That is the entire signal: not the complaint, the persistence of the complaint.
How to use it
On a competitor. Read two years of their worst reviews. Sort onto the ladder. Discard rungs one, two, and four. What remains, if anything appears repeatedly, is a list of things they have chosen not to build. Some of those are wise choices. One or two are usually a market they have decided to concede, and they will keep conceding it, because the decision is load-bearing for something else they care about more.
On yourself. The same sort, run honestly, tells you which of your complaints you are actually going to fix. Anything on rung three that you have carried for three cycles, you are not fixing. Stop putting it on the roadmap and either defend it out loud or change the strategy.
The failure mode
Sentiment scoring flattens all four rungs into one number and destroys the only useful distinction in the data. A product with a rising score and an unchanged rung-three complaint is in worse shape than the number suggests, because it is getting better at the things that were never the problem.
This came out of Weakspot, where it is the operating core rather than an aside.