Where this session sits
Sessions b2-b6 built your computing muscles. This session and the next build your shipping muscles: b7 turns the numbers into the weekly pack, b8 writes the spec sheet that keeps the numbers honest. On the leader side, a4 teaches the weekly business review ritual from the reader's chair - this session is the writer's chair for the exact same meeting.
Four pages, same order, every week 11 min live
A weekly pack works because it is boring in the right way: the same four pages in the same order every single week. Readers stop hunting for numbers and start noticing changes - which is the whole point. Page 1 is the scorecard (core four, WoW and YoY). Page 2 is the funnel gates (four gates, WoW). Page 3 is channels (GMV by channel, share, WoW). Page 4 is the narrative - one paragraph plus an actions table. Nothing else.
LiveWhy the order is fixed4 min▶
The order is a funnel of questions. Page 1 answers "how did we do" (the outcome). Page 2 answers "where in the journey did it happen" (the gates). Page 3 answers "which traffic did it come from" (the channels). Page 4 answers "so what" (the story and the actions). A reader who trusts pages 1-3 can skim straight to page 4; a reader who doubts page 4 can trace every claim back through 3, 2, 1. That two-way traceability is what makes a pack more than a screenshot dump.
Amazon's weekly business review runs on the same discipline: a fixed deck of metrics reviewed in the same order every week, so anomalies pop out by familiarity rather than by highlighting. You are building the small-store version - four pages instead of seventy, but the same principle: format stability is what makes change visible.
Self-studyWho reads which page3 min read▶
Each page has a natural first reader, and knowing who that is tells you how to write it:
- Page 1 (scorecard) - the CEO and finance. They read the WoW and YoY columns first and the levels second. Round consistently; never change decimal places between weeks.
- Page 2 (gates) - product and engineering. A gate that moved is their to-do list; a gate that held is their alibi.
- Page 3 (channels) - marketing. Share tells them the mix, WoW tells them this week's fires.
- Page 4 (narrative) - everyone, first. In practice the room reads page 4, then flips back to whichever page backs the claim they doubt. Write page 4 knowing it will be audited.
Assemble this week's pack ★ 15 min · everyone works
The setup: Mango Lane just closed a soft week. To keep every number checkable, we use the canonical baseline you memorized in b1 as "last week" and these actuals as "this week": sessions 356,400 (-1.0%), session CR 2.673% (-1.0%), AOV flat at $68, so GMV lands at $647,741 (-2.0%). Build pages 1-3 now; page 4 (the narrative) is Demo 2.
LivePage 1 - the scorecard, cross-footed6 min▶
| Metric | Last week | This week | WoW | YoY |
|---|---|---|---|---|
| GMV | $660,960 | $647,741 | -2.0% | +9.9% |
| Sessions | 360,000 | 356,400 | -1.0% | +7.0% |
| Session CR | 2.700% | 2.673% | -1.0% | +1.5% |
| AOV | $68.00 | $68.00 | 0.0% | +1.2% |
| Orders (check row) | 9,720 | 9,526 | -2.0% | +8.6% |
Orders = sessions x CR: 356,400 x 2.673% = 9,527 (9,526.6 - the pennies are rounding). Orders fell the same -2.0% as GMV because AOV held still.
GMV = orders x AOV: 9,526.6 x $68 = $647,741. Matches the headline. Page 1 agrees with itself.
Cross-foot the WoW: 0.99 (sessions) x 0.99 (CR) x 1.00 (AOV) = 0.9801, i.e. -2.0% GMV. Two small -1% dips multiply into the -2% headline. If your factor changes do not multiply back to the GMV change, a number on the page is wrong - find it before the room does.
YoY cross-foots too: 1.070 x 1.015 x 1.012 = 1.099, matching the +9.9% GMV YoY. Soft week, healthy year.
LivePage 2 - the funnel gates4 min▶
| Gate (b4 definitions) | Last week | This week | WoW |
|---|---|---|---|
| PDP penetration (of all sessions) | 40.00% | 40.00% | flat |
| Add-to-cart (of PDP sessions) | 25.00% | 24.75% | -1.0% |
| Checkout start (of cart sessions) | 45.00% | 45.00% | flat |
| Checkout completion (of checkout sessions) | 60.00% | 60.00% | flat |
Gates multiply to CR: 0.40 x 0.2475 x 0.45 x 0.60 = 0.02673 = 2.673% - exactly page 1's CR. The -1% CR dip lives entirely in one gate: add-to-cart.
Session counts confirm it: 356,400 x 40% = 142,560 PDP sessions; x 24.75% = 35,284 cart sessions; x 45% = 15,878 checkout sessions; x 60% = 9,527 orders. Same orders as page 1.
LivePage 3 - channels5 min▶
| Channel | Last week GMV | This week GMV | Share | WoW |
|---|---|---|---|---|
| Organic | $264,384 | $263,797 | 40.7% | -0.2% |
| Paid | $198,288 | $197,892 | 30.6% | -0.2% |
| CRM | $132,192 | $131,853 | 20.4% | -0.3% |
| Social | $66,096 | $54,199 | 8.4% | -18.0% |
| Total | $660,960 | $647,741 | 100% | -2.0% |
Channel rows sum to the page 1 total: 263,797 + 197,892 + 131,853 + 54,199 = $647,741. Never ship a channel page that does not sum to the scorecard.
The story jumps off the page: three channels roughly flat, social down $11,897 (-18%) - which is nearly all of the $13,219 total GMV drop. b6's segment-cut instinct did its job: the average hid it, the cut found it.
Self-studyThe pre-ship checklist - three reconciliations3 min read▶
Before any pack leaves your hands, run the same three checks in the same order. They take two minutes and they catch almost everything:
Cross-foot the scorecard. Sessions change x CR change x AOV change must reproduce the GMV change (0.99 x 0.99 x 1.00 = -2.0% this week). Rounding pennies are fine; a tenth of a percent is not.
Multiply the gates. The four gate rates on page 2 must multiply to page 1's CR (0.40 x 0.2475 x 0.45 x 0.60 = 2.673%). If they do not, one gate is stale or one denominator drifted (b8 territory).
Sum the channels. Page 3's rows must total page 1's GMV to the dollar. A gap means untagged traffic is hiding somewhere - find it or footnote it, never silently absorb it.
A pack that fails a reconciliation and ships anyway teaches the room to double-check you forever. A pack that reconciles for ten straight weeks earns the most valuable thing an analyst owns: numbers nobody re-derives.
Page 4 - the narrative ★ 8 min · everyone works
The narrative is one paragraph in a fixed template: "What moved: [metric, size]. Why: [lever + slice + hypothesis]. What we are doing: [action + owner]. When we will know: [re-check date]." No adjectives, no throat-clearing, every claim traceable to pages 1-3.
LiveThe filled narrative for this week5 min▶
What moved: GMV fell 2.0% WoW to $647,741 (-$13.2k). Why: the drop is concentrated in the social channel, down 18% (-$11.9k) while other channels held roughly flat; the add-to-cart gate dipped 1% with it. Working hypothesis: a platform algorithm change cut our social reach - impressions on our top posts fell midweek. What we are doing: social lead is rebalancing creative toward formats the feed currently favors and shifting this week's boost budget to the two best-performing posts. When we will know: re-check social GMV and add-to-cart in next Friday's pack.
| Action | Owner | Check date |
|---|---|---|
| Rebalance social creative to favored formats | Social lead | Next Friday's pack |
| Shift boost budget to top 2 posts | Social lead | Next Friday's pack |
| Confirm ATC dip is social-only (b6 cut by channel) | Analytics | Wednesday |
Self-studyNarrative anti-patterns - four ways page 4 goes wrong3 min read▶
- The weather report. "GMV was down slightly this week amid softer traffic." Describes, explains nothing, commits to nothing. If a sentence could have been written without opening pages 2-3, delete it.
- The victory lap with no denominator. "Social engagement was strong!" Strong against what? The narrative only speaks in the pack's specced metrics - engagement is not one of them.
- The hypothesis stated as fact. Write "working hypothesis: platform algorithm change", not "the platform changed its algorithm". The re-check date is what upgrades a hypothesis to a finding.
- The orphan action. An action with no owner or no check date is a wish. The actions table has three columns for a reason; a row missing any of them does not ship.
Weekly vs monthly - what the bigger pack adds 5 min live
The weekly pack manages the run-rate; the monthly pack manages the direction. A monthly pack keeps the same four pages and adds four things: benchmarks (a5's tables - where does our CR sit against IRP and Littledata this month), the returning-customer trend (Shopify's returning customer rate over the last 6 months), the guardrail metrics (margin and return rate - the numbers that stop you celebrating GMV bought with discounts), and one deep dive (a single question worked properly, b5-style). And where the weekly compares WoW first, the monthly compares YoY primarily - month-over-month is too seasonal to trust (a5).
Self-studyPack hygiene - three rules that keep it alive3 min read▶
- Freeze the format for a quarter. The pack's power is familiarity; a page that moves around resets everyone's pattern-matching to zero. Change format at quarter boundaries, announced.
- Footnote any definition change. If a metric's definition changes mid-quarter (a6 governance), the pack carries a dated footnote on every affected page until the comparison periods are clean.
- Never add a page without removing one. Packs die by accretion - every stakeholder wants their pet chart added and nobody ever asks to remove one. Four pages weekly is the budget. A new page must evict an old one.
Try it yourself - this week ◐ 20-30 min total
- Build page 1 (the scorecard) for your own store, this week: GMV, sessions, session CR, AOV - each with WoW and YoY.
- Cross-foot it: multiply your sessions, CR, and AOV changes and confirm they reproduce your GMV change. If they do not, find the leak before you show anyone.
- Bonus: write the four-sentence narrative for your week using the template, even if the answer is "nothing moved, no action". That sentence is also information.
Sources this session draws on
The weekly pack is a craft pattern, not a certified standard - it is this course's small-store distillation of how disciplined ecommerce teams review their numbers. This page draws on:
Three questions before you go 🎯 ◐ 90 seconds
1 · What is the weekly pack's page order, and why is it frozen?
Outcome, then where in the journey, then which traffic, then the story. Frozen format is the feature: familiarity is what makes an anomaly pop. A reshuffled pack resets everyone to hunting.
2 · Sessions fell 1%, CR fell 1%, AOV was flat. What GMV change should the scorecard show, and why?
GMV = sessions x CR x AOV, so changes multiply: 0.99 x 0.99 = 0.9801, about -2.0%. That is the cross-foot: if the scorecard's GMV change does not equal the product of its factor changes, a number is wrong.
3 · What must the page 4 narrative contain?
Four sentences, fixed template. It answers the room's four questions before they are asked; the pages behind it carry the numbers, the paragraph carries the meaning and the commitment to re-check.