Direction metrics are leading indicators measured on the cadence of the work itself. They exist because outcomes are too slow to steer with. For AI-native products, outcome metrics like NRR or CSAT lag four to twelve weeks behind your changes, and the data does not move faster than that no matter how fast you ship. With teams iterating ten to twenty times a week, by the time an outcome attributes back you have shipped forty to eighty more changes. Outcome accountability becomes a lagging measurement that cannot drive the decision you have to make today.
The seven leading indicators
Each predicts a specific outcome on a lag.
- Eval pass rate over the last seven days. Predicts customer CSAT four weeks out.
- Agent quality score from sampled outputs. Predicts NPS six weeks out.
- Iteration count and shipped changes. Predicts feature adoption eight weeks out.
- Design coherence, whether agent outputs match the brief. Predicts customer trust six weeks out.
- Customer escalation rate. Predicts churn twelve weeks out.
- Dispute rate on outcome billing. Predicts NRR eight weeks out.
- Latency at p95 and p99. Predicts retention six weeks out.
The two-layer system
Layer one, direction, runs daily and weekly on the seven indicators and drives day-to-day decisions: what to ship, what to roll back, what to evaluate further. Layer two, outcomes, runs monthly and quarterly on NRR, CSAT, NPS, expansion revenue, and churn, and drives strategic decisions: do we keep investing, are we pricing right, is the buyer changing. Both layers are deliberate and reviewed in different meetings with different audiences. Most teams run only one.
The Goodhart audit
Goodhart's law is real, so verify constantly. Every quarter, for each indicator, plot it against the outcome it predicts with the lag and compute correlation. Above 0.7 is healthy. Between 0.5 and 0.7 it is weakening. Below 0.5 it is dead, so replace it. Treat leading indicators as derivatives of outcomes, not substitutes.
This changes how PMs are measured too. Outcome accountability moves from monthly to annual. Direction accountability becomes the day-to-day, and PMs are evaluated on whether their leading indicators are well-chosen and actually predicting outcomes.
This week, build the indicator registry. A YAML file listing your seven indicators, where the data lives, the healthy band, and what each predicts.