Narrowed a broad social-listening product around one job: finding public buying intent.

I started Buska in November 2023, alongside the lempire role, as a general social listening tool: anyone with a keyword to watch, from brands to agencies.

Technically that was the easy version to build: every one of those users wants the same thing from the engine. Commercially it was the hard version to sell. Among our users, none of those segments wanted it badly enough to pay much for it.

Product, engineering, deployment and go-to-market, with my co-founder. I designed it, built it, shipped it and sold it.

The accounts that kept coming back were not watching their own brand. They were watching competitors, their category, and phrases that amount to someone raising a hand. They were prospecting by hand, with a monitoring tool.

Commit to lead generation and rebuild around intent rather than coverage. The same mention is a notification to a brand manager and a qualified lead to a sales team, and only one of those is a line in somebody's budget.

Rejected: adding more platforms, which makes the tool better at a job nobody paid much for. Rejected: serving both buyers, which need opposite defaults. Rejected: moving upmarket, which is the wrong fight for a small team with no sales org.

Every mention scored 0 to 100 on how close it is to a buying decision, and the interface became a queue of people to contact rather than a feed of things that happened.

Deliberately not done: no creator or personal-brand segment, no sentiment dashboard, no share-of-voice charts. During the repositioning we paused new platforms until the new use case was validated.

The full rebuild started in January 2026 and shipped in April 2026. MRR then grew by roughly 40% month over month over the five months that followed.

The cost was real: we lost users who had been with us from the start and were not doing anything wrong. They simply were not the business.

Decide earlier. The signal was in the usage data months before we acted on it; we were reading it as a curiosity rather than as an instruction.

Tell the abandoned users directly instead of letting the product drift away from them.

The 40% figure is month over month across five months from a small base after a relaunch. I would want to see it hold as the base grows before treating it as a growth rate rather than a recovery.