Success Metrics
A week after LunchLoop's Skip feature launches, three people bring three numbers. Skips are up 300%. App sessions are up 20%. Revenue is down 3%. Is the feature working? If nobody decided in advance which number means success, each person will quote the one that tells their favourite story.
The rule
Pick the number that proves the goal, plus the numbers it must not break.
Every piece of work needs a small, agreed set of metrics, chosen before launch:
- Primary metric: measures the goal directly. LunchLoop's goal is fewer uneaten lunches, and riders collect the boxes each evening, so the metric is the untouched-box rate: boxes returned untouched ÷ boxes delivered. Baseline 12%, target 5%.
- Leading indicator: moves earlier and predicts the primary metric. The share of active subscribers who have skipped at least once shows up within days, while waste figures settle over weeks.
- Guardrail metrics: things that must not get worse. Skipped meals become credit, so revenue per subscriber should hold. Cook earnings must hold too.
Beware vanity metrics: numbers that rise and feel good but do not show whether the goal is closer. Downloads and page views are classics. "Total skips" is subtler: it goes up both when people skip a lunch they would have wasted (good) and when they simply buy fewer lunches (bad). On its own, it cannot tell the two apart.
A good metric is tied to the goal, has a baseline and a target, responds to the change you made, and is measured the same way before and after.
The shortcut
For each proposed number ask: "If only this went up, would the goal be closer?" "Yes" makes it a candidate. "It depends" means it is vanity or needs a partner. Then ask "What could we break?" to find the guardrails.
Start from the goal
Cut untouched tiffins from 12% to 5% of deliveries by the end of March.
Drills
Answer fast — trust the rule.
Which is a guardrail metric for LunchLoop's Skip feature?
Which of these is a vanity metric for the waste goal?
Why agree on the primary metric before launch?
The untouched-box rate falls from 12% to 6%, but revenue per subscriber drops 15%. What is the right reading?
Remember
- Pick the number that proves the goal, plus the numbers it must not break.
- Shortcut: "if only this went up, would the goal be closer?", then "what could we break?"
- Trap: vanity numbers like downloads or total skips can rise while the goal stands still.