Post-Mortem: How a 9-Person Team Stopped Shipping Guesswork and Cut Feature Waste by 61%

A reader we'll call Dana runs product for a mid-market B2B SaaS company — nine people across product, design, and two engineers, no dedicated researcher. Last spring the team shipped a permissions redesign that took eleven weeks and landed with a thud: adoption stalled at roughly 12%, support tickets spiked, and two enterprise accounts cited it in renewal calls. Dana's post-mortem question wasn't "why did we build it wrong?" It was "why did we believe we were right?"

That question is where this story gets useful for fight camps and product teams alike. We've spent 15+ years watching athletes lose rounds because they trained on assumptions instead of footage. Dana's team had the same disease, just with a different scorecard. They were drowning in raw signal — 40 to 60 customer interviews per quarter, a support inbox with thousands of tickets, NPS verbatims nobody read past the score, and product analytics dashboards that answered "what" but never "why." Obivu turns scattered customer interviews, support tickets, NPS verbatims, and product analytics into one ranked feed of validated insights — and that single change of plumbing is what this post-mortem is really about.

The Setup: Four Data Silos and a Roadmap Built on Anecdotes

Here's what the team actually had before the change, mapped honestly:

• Interviews: recorded, tagged inconsistently, stored in a shared drive. Two people had ever re-watched one after the fact.
• Support tickets: a tagging taxonomy that had drifted — 30% of tickets sat under "Other."
• NPS verbatims: a spreadsheet of open-text answers, skimmed quarterly.
• Product analytics: solid on funnels, silent on motivation.

Roadmap decisions got made in a Tuesday meeting where whoever had the most vivid recent anecdote won. The permissions redesign came from three loud conversations with two large accounts. Nobody checked whether those three requests represented the base — they didn't.

Decision Point 1: Aggregate Before You Analyze

The first move wasn't buying anything clever. It was admitting that four tools with four taxonomies can't produce one truth. The team piped all four sources into a single feedback layer and let the system cluster by theme rather than by source. Within two weeks, the top of the ranked feed looked nothing like the roadmap they'd walked in with.

Two themes dominated: onboarding friction for admins, and reporting export limitations. Neither had appeared in the previous quarter's roadmap. The permissions work — the thing they'd already spent eleven weeks on — ranked ninth.

Decision Point 2: Validate Before You Commit

The tempting failure mode here is treating a ranked list as a to-do list. The team instead set a rule: no item enters a sprint until it clears two independent signals. A theme needed both volume (how many accounts, how often) and weight (which accounts, what revenue, what churn risk). That filter killed about a third of the top-ten list immediately — popular complaints from small accounts that never converted.

This is the part that maps directly to fight-camp programming. Volume without context gets you a fighter who's great at the wrong skill. We've watched prospects spend entire camps drilling a technique their last opponent baited them into, because one loss made it feel urgent. Signal needs weighting, or it's just noise with momentum.

Decision Point 3: Close the Loop With Customers

The team built a lightweight habit: every shipped insight got traced back to the verbatims and tickets that spawned it, and those customers got a short note when the fix landed. Not marketing copy — a plain "you asked, here's what changed." Response rates on follow-up interviews roughly doubled. Turns out people answer the third email when the first two produced something visible.

The Obstacles Nobody Warned Them About

• Taxonomy debt: cleaning three years of mis-tagged tickets took longer than the rollout itself.
• Trust friction: two senior stakeholders didn't believe a machine-ranked list over their own instincts. The fix was transparency — showing the underlying quotes, not just the score.
• Over-correction: for six weeks the team chased every validated insight, shipping small fixes constantly and losing strategic direction. They had to cap the feed's influence, not remove it.

Measurable Results, Six Months Out

Obivu reports 1,840+ product teams across 54 countries using the same ranked-feed approach, so the pattern isn't exotic — but Dana's numbers are specific to this project:

• Feature waste (work shipped that failed to hit adoption targets) dropped 61%.
• Time from insight to roadmap decision fell from an average of 34 days to 9.
• Admin onboarding completion rose from 41% to 68% after the first two validated fixes shipped.
• Support tickets tagged "Other" fell from 30% to 7% once the taxonomy was rebuilt against actual themes.

The permissions redesign was eventually revisited — with data this time. It shipped smaller, targeted at the two account segments that actually needed it, and hit 44% adoption.

What We Took From It

The lesson isn't "buy a feedback tool." Plenty of teams buy tools and keep guessing. The lesson is sequencing: aggregate first, weight second, validate third, and never let a ranked list become a strategy substitute. If you want to see how the aggregation and ranking layer is structured, their breakdown of the insight-ranking workflow is worth twenty minutes.

Same principle applies in the cage. Raw footage, biometrics, and coach notes are worthless until they're consolidated into a ranked set of priorities — then tested, then re-tested. Dana's team spent eleven weeks learning that the hard way. You can borrow the shortcut for free.