Post-Release & Learning · master area

Lagging Signal Tracking

The prediction metric, support volume, retention. The verdicts that arrive in weeks — and the data the prediction check draws on. Slow to move, hard to argue with, expensive to hear wrong.

Owners: PO, Data Phase it lives in: After We Build (Volume V) The corpus principle this enacts: A prediction not checked is indistinguishable from a guess.

Where it lives in the chain

• After We Build · Signal and the Prediction — the canon

The lagging signals

• The prediction metric itself — "Gal completes grading in under 15 minutes." Measured on the date the brief named, with the method the brief named. Not by survey. Not by impression.
• Support volume — total tickets, tickets by category. A spike in "how do I" tickets two weeks after release is content design feedback; a spike in "this doesn't work" is a scenario gap.
• Retention — for features users opt into, what fraction are still using them at 30/60/90 days? Drop-off at 7 days is novelty wearing off; drop-off at 30 days is the feature failing to fit the workflow.
• VRI (Value Realization Index) — fraction of the predicted value that actually reached the person. Predicted 30 min saved per session × actually 24 min × actually used 70% of sessions = VRI 0.56.
• Satisfaction (CSAT, NPS) — the relationship at the affective level. Slow to move, but the most expensive to hear wrong.

What good practice looks like

The lagging signals are wired at release time, watched monthly, and fed into the model update the cycle after they stabilise. Lagging signals do not produce immediate action — they produce learning. A lagging signal that contradicts the prediction is not an outage; it is a model correction.

The discipline: lagging signals are read alongside leading. A leading signal saying "adoption is high" combined with a lagging signal saying "VRI is 0.3" is a clear story — people are using it but it isn't producing the change we promised. That is the cycle's most important finding. The brief was wrong in a specific way; the next brief inherits that lesson.

Related crafts

• Leading Signal Tracking — what arrives in days
• Helpdesk Metrics — the support-volume detail
• Signal Reading — the artefact the signals feed