Turning a Pile of Feedback Into a Build Order
BuildFR reads everything your customers said and ranks it. The hard part is that people don't write tickets.
BuildFR takes customer feedback from wherever it lives and ranks it into what to build next. Support tickets, app store reviews, sales call notes, a Slack thread someone pasted in.
The input is not clean and never will be.
People don’t describe problems, they describe moods
Real feedback looks like this:
- “the export thing is broken again”
- “love the app but it’s slow”
- “why can’t I just do X”
- three paragraphs about something unrelated, containing one sentence that matters
None of that is a feature request. All of it contains one.
The pipeline
Extract. Pull the actual claim out of the noise. A single piece of feedback often carries two complaints and a compliment, and they belong in different places.
Classify. Bug, request, confusion, praise. Confusion is the interesting bucket, because it usually means the feature already exists and nobody could find it. That is a much cheaper fix than building it again.
Cluster. Most of the value lives here. Forty people saying the same thing in forty different ways is one item, and the count is the signal. Get clustering wrong and everything downstream is wrong with it.
Rank. Frequency, severity, and who said it, weighted together.
The mistake I made first
I clustered on surface similarity, and it grouped by vocabulary instead of by problem. “Export is slow” and “the download takes forever” landed in different buckets, while “export is slow” and “export is broken” landed in the same one. One of those pairs is the same issue and the other is two issues, and I had it exactly backwards.
Clustering has to run on the extracted claim, after normalisation, not on the raw text. Obvious in hindsight. Was not obvious at the time.
What generalises
Every one of these projects is the same four moves. Get the messy thing in. Pull out what it actually says. Normalise it into a shape. Then do the useful work on the shape.
The useful work is the part people talk about. The three steps in front of it decide whether it works.