A quiet churn problem hiding in plain sight — and the discovery process that finally surfaced it.
Our largest enterprise segment had started churning quietly. Not with support tickets or angry emails — just silence. Usage would drop off slowly over 60–90 days, and by the time the account reached renewal, the decision was already made.
The retention team flagged it. The CS team had theories. But nobody had a clear answer for why.
I was brought in as PM to lead the investigation and own the resulting workstream. I partnered with a senior designer, two engineers, and our data analyst. No existing spec. No assumed solution.
I started where I always start: with the data. Our analyst pulled session recordings and funnel drop-off data for the segment. The pattern was immediately visible — a specific 4-step workflow had a 34% abandonment rate on step 3. Nobody had flagged it because the completion rate still looked okay in aggregate.
"Users weren't confused. They were bored. The workflow made them repeat the same action four times."
We ran 12 JTBD-structured interviews with churned and at-risk customers. The insight that kept coming up: the workflow felt designed for an admin, not for the actual user doing the job. Every session required the same repetitive input regardless of context.
Key insight: The problem wasn't complexity — it was unnecessary repetition. Users had developed workarounds (copy-pasting from spreadsheets, keeping the tab open in another window) that added 3+ minutes per session and left them feeling like the product didn't respect their time.
Rather than redesigning the whole workflow, I scoped the intervention tightly: collapse steps 2–4 into a single context-aware step that pre-filled known values. This required a backend change (surfacing user context at the right moment) and a frontend change (conditional form logic).
Figma prototype
In Figma: Share → Copy link → paste into src above.
Change embed_host=share&url= to your link.
Before: the original 4-step workflow. After: the redesigned single-step version with context pre-filled. Use Figma's prototype view to show the interaction.
Measured at 90 days post-launch for affected accounts.
Average time-on-task dropped from 5.2 to 2.1 minutes.
Up from 66% on the previous workflow.
The change also had an unexpected secondary effect: support tickets related to this workflow dropped 58% in the following quarter. Users weren't confused — they'd just been frustrated.
The most important lesson here wasn't about the workflow — it was about where to look. Churn is often a trailing indicator. By the time it shows up in your metrics, the user has already made their decision. The signal was buried in session data, not in the renewal conversation.
This case reinforced my belief that discovery isn't a phase — it's a habit. The team that noticed this pattern earliest was the one that had been watching session recordings regularly, not waiting for a problem to be escalated.