Where you are
Session eight of eight. You mapped the store (b1), measured its doors (b2) and paths (b3), ran the machine (b4), audited the decision room (b5) and the last mile (b6), and diagnosed a leak end to end (b7). This session turns all of it into the document that survives without you - and closes the course.
The spec fields, plus one 10 min live
The metrics course's spec sheet (its b8) fixed seven fields per metric: name, definition, formula, denominator, source events, owner, alert threshold. Journey metrics need one more: SCREEN/STEP - the address a breach points to. A journey KPI without a screen is just another number; with one, every alert lands on someone's actual work.
LiveWhy the screen field changes behavior4 min▶
When PDP penetration breaches its threshold, the old spec pages "the analytics owner" - who then spends a day finding where to look. The screen field pre-computes that day: the spec itself says the breach lives on search results, the PLPs, or the Home modules, and the b7 drill-down starts there. Two rules make it work:
- Screens, plural, ranked. Most journey KPIs have two or three candidate addresses; list them in drill-down order.
- The owner owns the screens. If the metric's owner cannot change its screens, the spec names the wrong owner - fix the row, not the org chart.
Mango Lane's full journey scorecard ★ 15 min · everyone builds
The course's eight rows, filled. Thresholds are simple percentage bands - the statistical version of "is this move real" belongs to learn-anomaly-detection-with-phoebe.
| Row | Baseline | Formula / events | Owner | Alert | Screen / step |
|---|---|---|---|---|---|
| Instrumentation health | all screens firing | event volume per screen within expected band | Analytics | any screen ±40% vs 4-week norm | whichever screen went quiet |
| PDP penetration | 40% | sessions with ≥1 view_item ÷ sessions | Search + nav | ±10% WoW | Search results · PLP · Home modules |
| Door mix | 50 / 15 / 15 / 10 / 10 | entrances per landing screen ÷ sessions | Growth | any door ±5 pts | the door's landing screen |
| Home reach-PDP | 20% | Home entries reaching ≥1 view_item ÷ Home entries | Homepage team | ±10% WoW | Home modules · nav menu |
| Search path share + CR pair | 30% · 12% vs 4.5% | view_search_results sessions of PDP sessions; orders ÷ path sessions | Search team | share ±5 pts; CRs ±10% | Search results · autocomplete |
| Engaged-PDP rate + CR pair | 50% · 9% vs 4.5% | engaged rule (2+ images / reviews / size guide) ÷ PDP sessions | PDP team | ±8% WoW | Gallery · reviews · size guide |
| Checkout step completions | 90 / 83.3 / 80 | GA4 checkout journey per-step rates | Payments + cart UX | any step ±8% WoW | Shipping form · payment form |
| Zero-result search rate | instrument first | zero-result view_search_results ÷ all searches | Search team | set after 4-week baseline | Search results (empty state) |
LiveWhy instrumentation health is row one4 min▶
Every other row assumes the events are firing. b1's traps (screens that never log, search that skips the URL) and b7's false-positive checklist both taught the same lesson: the most common journey "incident" is a measurement outage wearing a business costume. The health row - each screen's event volume within a band of its 4-week norm - catches it before anyone panics. It goes first on the scorecard and first in every diagnosis.
Self-studyWeekly rows vs monthly rows3 min read▶
Not every row earns weekly attention (a6 set the leader's view; this is the analyst's split):
- Weekly (the WBR rows): instrumentation health, the four gates WoW, door mix, search share + zero-result rate, per-step checkout. Fast-moving, incident-shaped.
- Monthly: path CR pair, engaged-PDP pair, and the reach-PDP rates per door. Slow-moving, mix-sensitive - weekly reads on them are mostly noise (the thin-slice rule from b3 applies).
- Changelog discipline (inherited from the metrics course): definitions change only with a dated entry, and affected dashboards carry a footnote for one quarter.
What you can now do ★ 8 min · say it out loud
Eight sessions, one skill chain - walk it once, out loud:
Map (b1): six screens, five doors, the events that draw them - and the audit that keeps the map honest.
Doors (b2): entrances, per-door reach, and the mix-vs-quality decomposition that stops false credit.
Paths (b3): search 12% vs browse 4.5%, forward and backward pathing, and the what-if arithmetic of shifting sessions between roads.
Machine (b4): four gate dials, fixes as gate deltas, the +47.2% ladder - every fix priced.
Decision room (b5): engaged sessions convert 2x plain - stated as a diagnostic, never a causal promise.
Last mile (b6): three checkout steps, Baymard's killers mapped to each, drops located and priced from the chain.
Diagnosis (b7): from a top-line drop to a single screen with an owner - the finding template that ends war rooms.
Scorecard (b8): all of it as a standing instrument, instrumentation health first.
Try it yourself - this week ◐ 30-40 min total
- Write your own 8-row journey scorecard. Start from Mango Lane's rows and swap in your store's baselines - "instrument first" is an honest cell where you have no data.
- Put the instrumentation-health row first and wire the cheapest version: a weekly glance at event volume per screen.
- Get every owner named in a row to agree in writing - a spec nobody accepted is a wish with a table around it.
- Ship it where your team actually reads (wiki, dashboard header, the WBR deck appendix) - not a drawer.
Sources this session draws on
The scorecard consolidates the whole course; its rows cite the sessions that verified them. This page draws on:
Three questions before you go 🎯 ◐ 90 seconds
1 · What does the SCREEN field add that the metrics course's spec sheet lacked?
Journey metrics live on screens. The field turns "PDP penetration breached" into "start at search results, then PLP, then Home modules" - b7's drill-down, pre-loaded.
2 · Why is instrumentation health row one?
Every other row assumes events are firing. A screen that went quiet fakes a business drop perfectly - the health row catches it before the war room convenes.
3 · The zero-result search row's baseline cell says "instrument first". Why not estimate one?
A spec row is a promise about reality. Where no data exists yet, the honest state is "instrument first" - the same discipline as the course's flagged benchmarks.