You run product at a marketplace matching blue-collar workers with employers — drivers, warehouse staff, delivery riders. The team tracks applications submitted, up 60% this year. Employers post 12,000 roles a month and fill 4,100. Workers submit a median of 9 applications and hear back on 2. Median time from application to a decision is 11 days. Thirty per cent of filled roles are vacant again within 45 days.
Propose a north-star metric with an input tree and guardrails, and say what you would do if two metrics pointed in different directions.
100 points, 60% to pass.
Applications submitted is actively misleading: it goes up when matching is bad, because a worker who hears nothing applies again. The marketplace's real output is a filled role that lasts, and the 30% re-vacancy inside 45 days says fills alone are not it. A north star of roles filled and still filled at 45 days captures both sides. Inputs: employer response rate, time to decision, applications per fill, match quality. Guardrails: worker applications per hire — rising means wasted effort — and employer posting quality. Strong answers notice the two-sided tension explicitly: raising fill speed by loosening matching would raise fills and worsen retention, and say which they would protect.