Problems I've solved

The problem

A self-serve trial funnel converting at 4%. The onboarding tour measured success by completion (9 steps, ticked off dutifully) but completion didn't predict whether anyone stuck around. The team was optimising a metric that flattered us and told us nothing.

What I did

• 01Interviewed 30 trial users (joiners and leavers) to find the moment the product actually clicked for the ones who stayed.
• 02Redefined activation around that moment: the first shared report, not the last tour step.
• 03Cut the tour from 9 steps to 3 and moved setup into the product's real job, so learning and doing stopped being separate activities.

The problem

Buyers searched, scrolled, and left. Results were ranked by recency, which is a ranking only a database could love. The team assumed the fix was a search rewrite, months of work, new infrastructure, a proper project.

What I did

• 01Read the query logs. Actually read them, the top 20 queries covered a huge share of traffic and half returned nonsense.
• 02Shipped intent-aware ranking for just those top queries before touching any infrastructure.
• 03Rebuilt the empty state so a failed search still offered the nearest useful thing instead of a shrug.

The problem

An ops team of six ran the business from a 4,000-row spreadsheet. The company had built them a Proper Internal Tool. Adoption sat around 10%, training budgets had been spent twice, and the spreadsheet quietly remained the real product.

What I did

• 01Shadowed the ops team for a week and discovered the tool wasn't wrong about what the job was, it was wrong about how they thought about the job.
• 02Rebuilt the tool around the spreadsheet's actual columns and mental model, matching the order they worked in, warts and all.
• 03Deleted everything else. Ruthlessly. Every feature nobody had asked for by name.