The door table, computed 12 min live
Two report ingredients build this table. Entrances count the sessions whose first event happened on each screen - that gives you the sessions and share columns. Then, for each entry group, you measure how many of its sessions ever fire view_item - the per-door reach-PDP rate. Multiply the two and each door tells you exactly how many PDP sessions it produces.
| Entry door | Sessions | Share | Reach-PDP rate | PDP sessions produced |
|---|---|---|---|---|
| Home | 180,000 | 50% | 20% | 36,000 |
| PDP-direct (deeplink, ads, push) | 54,000 | 15% | 100% | 54,000 |
| PLP - category grid | 54,000 | 15% | 50% | 27,000 |
| Search landing | 36,000 | 10% | 75% | 27,000 |
| Other (account, orders, support) | 36,000 | 10% | 0% | 0 |
| Total | 360,000 | 100% | 40% weighted | 144,000 |
LiveWhere every number comes from5 min▶
Nothing in this table is exotic - it is two standard reports joined:
- Sessions and share: the landing page / landing screen report. GA4 defines the landing page as the first page of a session, and entrances as the count of sessions whose first event happened on that screen. Sort by entrances, group into your five doors, done.
- Reach-PDP rate: segment sessions by landing screen, then ask what fraction of each segment fires
view_itemat least once. That is the door's quality - how good it is at getting a shopper in front of a product. - PDP sessions produced: sessions x reach rate. The column that turns a traffic report into a contribution report.
Read the picture the SVG paints: Home is the widest door and the palest - half the traffic, one fifth of it ever sees a product. PDP-direct is narrow and the deepest color - every one of its sessions is a PDP session by definition (they land on one).
Self-studyWhy "Other" earns a row of zeros2 min read▶
36,000 sessions land on account, order tracking, or support screens and never reach a PDP. Keeping them in the table (rather than filtering them out) is deliberate: they are 10% of the denominator of every session-level metric you report. When someone asks "why is our PDP penetration only 40%?", one honest tenth of the answer is "because a tenth of sessions were never shopping." b8's scorecard keeps this row visible for exactly that conversation.
The door quality drill ★ 12 min · everyone works
Three computations, in order. Do them by hand - the punchline only lands if you feel the weights doing the work.
LiveCompute along: three steps to the punchline10 min▶
Step 1 - each door's PDP contribution share. Divide each door's PDP sessions by the 144,000 total: Home 36,000 / 144,000 = 25% · PDP-direct 54,000 / 144,000 = 37.5% · PLP 27,000 / 144,000 = 18.75% · Search 27,000 / 144,000 = 18.75% · Other 0%. Home carries 50% of traffic but only 25% of PDP reach; PDP-direct carries 15% of traffic and 37.5% of PDP reach. Traffic share and contribution share are different rankings.
Step 2 - rebuild the 40% as a mix-weighted sum. Multiply each door's traffic share by its reach rate and add: 0.50 x 0.20 = 0.100 · 0.15 x 1.00 = 0.150 · 0.15 x 0.50 = 0.075 · 0.10 x 0.75 = 0.075 · 0.10 x 0 = 0. Sum: 0.100 + 0.150 + 0.075 + 0.075 + 0 = 0.400 exactly. The headline number falls out of the table with nothing left over.
Step 3 - say the punchline out loud. PDP penetration is not one number. It is a mix-weighted average of five door qualities. Any move in the headline is some blend of the weights moving (mix) and the qualities moving (doors getting better or worse) - and you cannot tell which from the headline alone.
Mix versus quality: the 44.4% that fooled the standup ★ 13 min · everyone works
A worked scenario. Next month a paid push campaign deep-links users straight to products. Home's reach-PDP rate holds at exactly 20% - no door gets better or worse. But the mix shifts: Home drops to 160,000 entries, PDP-direct rises to 74,000, and PLP, Search, and Other are unchanged. Compute what happens to penetration.
LiveCompute along: the full arithmetic10 min▶
Step 1 - new PDP sessions per door. Home 160,000 x 0.20 = 32,000 · PDP-direct 74,000 x 1.00 = 74,000 · PLP 54,000 x 0.50 = 27,000 · Search 36,000 x 0.75 = 27,000 · Other 36,000 x 0 = 0.
Step 2 - new totals. Sessions: 160,000 + 74,000 + 54,000 + 36,000 + 36,000 = 360,000 - same month-sized store, only the doors reshuffled. PDP sessions: 32,000 + 74,000 + 27,000 + 27,000 + 0 = 160,000.
Step 3 - new penetration. 160,000 / 360,000 = 44.4%. Up 4.4 points from 40%.
Step 4 - name what happened. Every reach rate is identical to last month. Not one door improved. Penetration "improved" 4.4 points purely because traffic moved from the palest door (Home, 20%) toward the deepest one (PDP-direct, 100%). This is a pure mix effect - the campaign bought different sessions, it did not build a better journey.
This is the most common false win in growth reviews. Paid campaigns that deep-link to products mechanically raise PDP penetration and often raise session CR - while the organic journey quietly stays the same or degrades. Teams that report the decomposition alongside the headline catch it in the same meeting instead of three months later when the campaign budget ends and the "improvement" evaporates with it.
Weak landings, and who sent them 6 min live
The door table measures how far entering sessions get. Its dark twin measures the sessions that never get anywhere: single-screen sessions that land and leave. They do not spread evenly - they concentrate on weak landings, and finding which door leaks them is the cheapest diagnostic in this course.
LiveBounce, by door3 min▶
Segment single-screen sessions by landing screen, the same cut as the door table. The usual pattern, and Mango Lane follows it: PDP-direct landings bounce hardest when the ad promised something the product page does not deliver (price, availability, the exact item); Home bounces when it loads slowly or buries the way in; Search landings bounce when the query lands on an irrelevant results page. A door with big entrances and a heavy single-screen share is a door paying for traffic it immediately loses - fix the landing before buying more of it.
Self-studyDoors are WHERE - sources are WHO SENT3 min read▶
Keep the two dimensions straight: the door table is about WHERE sessions begin (landing screen). Source scopes are about WHO SENT them, and GA4 records that at three scopes:
| Scope | Answers | Against the door table |
|---|---|---|
| First user source (user-scoped) | Which channel found this person first, ever | Why a returning shopper walks in the Home door out of habit |
| Session source (session-scoped) | Which channel drove this session | Why 54,000 sessions land straight on a PDP - ads and pushes aim there |
| Event scope (key events) | Which channel gets conversion credit (data-driven attribution default) | Which sender gets paid for the orders a door produces |
Cross the two and reports get sharp: "PDP-direct door, session source = push campaign X, reach 100%, bounce 31%" is actionable. "Traffic is up" is not. The event-scope story runs deep - it is its own course, learn-marketing-attribution-with-phoebe.
Try it yourself - this week ◐ 20-25 min total
- Pull your landing-screen report, sorted by entrances. Group the long tail until you have your own five doors.
- Build your five-door table: sessions, share, reach-PDP rate (share of each door's sessions that fire view_item), PDP sessions produced.
- Compute your penetration as the mix-weighted sum of door qualities - then check it against the reported PDP penetration number. If they disagree, your door grouping or your segment definition is leaking sessions; find which.
- Bonus: rerun last month's table next to this month's. Hold rates constant to separate mix effect from quality effect before your next traffic review.
Sources this session draws on
Everything here is standard-report material applied to one store, plus one piece of analyst craft shown with full arithmetic:
Three questions before you go 🎯 ◐ 90 seconds
1 · Mango Lane's 40% PDP penetration is best described as...
The weighted sum rebuilds 0.40 exactly. It is not one number - any move in it blends mix shifts and quality shifts, and only the decomposition tells you which.
2 · In the campaign scenario (Home 160,000, PDP-direct 74,000, all reach rates unchanged), penetration hits 44.4%. What improved?
32,000 + 74,000 + 27,000 + 27,000 = 160,000 PDP sessions over the same 360,000 = 44.4%. Zero doors got better - the weights moved, the qualities did not.
3 · What exactly does "entrances" count for a screen?
Entrances are session-scoped and first-event-scoped: the session began there. Any-visit counts are screen reach; first-ever-visit is the user-scoped acquisition lens.