Instrumentation converts the map into a measurement system. The Intermediate layer's journey — stages, touchpoints, handoffs — becomes an event schema: named actions (searched, shortlisted, payment_started, payment_failed) with properties (device, channel, segment) and, crucially, a resolved identity connecting them. Identity is the hard part and the honest prerequisite: without cross-device and cross-channel stitching, analytics sees fragments and calls them customers — five anonymous sessions instead of one person deciding over five days. Get the schema and identity right, and the whiteboard's hypotheses ('we think people drop off after search') become queryable facts with denominators.
Diagnosis has one iron rule: segment before you conclude. An aggregate funnel is an average of stories, and averages launder every insight — the overall 3% conversion hides that returning desktop users convert at 11% while first-time mobile visitors die at the payment step behind one failing wallet integration. The working toolkit: funnels cut by cohort, source, segment, and device; path analysis (the sequences customers actually walk, versus the designed route — repeated loops between two screens are a confusion signal no survey will volunteer); and velocity (time-between-stages — journeys often stall before they die, which makes the stall the earlier, cheaper alarm). Watch one instrumented journey: flight searched at lunch on a phone, price alert set, evening desktop comparison, wallet failure at payment, completion next morning after a retry nudge. Channel dashboards file that as four sessions and an ad conversion. Journey analytics reads one story with one fixable villain.
Journey metrics answer questions touchpoint metrics can't ask. Every touchpoint can be individually optimised — search fast, app rated highly, support courteous — while the journey fails between them, because handoffs have no owner and no scoreboard of their own. Completion rate (of those who started, who finished?), time-to-complete, and customer effort (steps, re-entries, repeated information) instrument the whole story. This is the quantitative sequel to the Intermediate page's graveyard warning: the map said handoffs are where journeys die, and the funnel-by-cohort tells you which handoff, for whom, costing how much this quarter — turning an empathy exercise into a prioritised engineering queue.
Orchestration closes the loop: measurement that doesn't change the next customer's journey is just a report. The mechanics: triggers (state changes fire actions — payment_failed at hour two sends the retry link), next-best-action (the customer's state and history pick the one most useful message, replacing the campaign blast), and — as load-bearing as any message — suppression and frequency rules (the purchased are excluded from retargeting, the fatigued are rested, the mid-complaint are spared the cross-sell). One boundary keeps this page honest: journey analytics diagnoses and repairs experiences — where journeys break and what to do next. Which channel deserves credit for the conversion is a different question with different instruments, and that's attribution's page, next door.