Meet Alpine Roast
"Alpine Roast" is a subscription-plus-shop online coffee brand, and it is having a very bad quarter. Monthly income has collapsed from $23,000 to $4,250 over two quarters - an 82% drop - while marketing spend held perfectly steady. The founder's dashboard shows one red number and zero explanations. Your job today: run the full course toolkit on this one case, from panic to a signed one-page action memo, in 45 minutes. Every move you make is a move you learned in Sessions 1-7.
Walk the tree, not the org chart 6 min live
When income drops 82%, the untrained reflex is to summon marketing. The trained reflex is Session 4's KPI tree: income = revenue - cost; revenue = buyers × basket size × item value; buyers = new + returning + resurrected - churned. Fill in Alpine Roast's numbers and the panic resolves into exactly two moved branches.
LiveFrom one red number to two named problems4 min▶
Walk it top-down, exactly like Session 4 taught. Last healthy month: 1,000 buyers (400 new + 700 returning + 100 resurrected - 200 churned) × 3 items × $25 = $75,000 revenue, minus $52,000 cost = $23,000 income. Today: churned doubled to 400 and new buyers slipped to 350, so buyers = 350 + 700 + 100 - 400 = 750, revenue = 750 × 3 × $25 = $56,250, income = $4,250.
- Cost is flat. Cross that whole side of the tree off in ten seconds - no cost-cutting panic needed yet.
- Basket size and item value are unchanged. The product and pricing are fine. Revenue fell purely because buyers fell.
- Two branches moved: churned doubled (200 → 400) and new buyers slipped (400 → 350). Different problems, different owners - retention owns the first, growth owns the second.
- The leverage lesson: income sits at the top of a multiplication, so a 25% buyer drop became an 82% income drop. Small branch moves compound violently at the root - which is why the tree, not the dashboard's one red number, is where diagnosis starts.
Self-studyWhy the tree beats the org chart every time2 min read▶
When a topline number breaks, most companies route the question by org chart: revenue is down, so marketing presents next Monday. But the org chart tells you who owns activities, not which arithmetic branch moved. Alpine Roast's marketing spend never changed - summoning the CMO would have produced a defensive deck and zero diagnosis.
| Reflex | What happens | Time to truth |
|---|---|---|
| Org chart first | each team defends its own metrics | weeks, sometimes never |
| Tree first | arithmetic isolates the moved branches, then you invite the right owners | one meeting |
The tree is also a fairness device: nobody gets blamed for a branch that did not move. Returning buyers held at 700 - somebody is doing their job well, and the tree says so out loud.
One tool per red branch 5 min live
The tree names the problems; Sessions 5-7 gave you a drill for each. Churn questions go to cohorts. Acquisition questions go to the funnel. "Who do we save first?" goes to RFM. "What if we do nothing?" goes to the forecast. Four analyses, four one-line findings - this is the war room wall.
LiveThe four findings, one per drill4 min▶
- Churn branch → cohort view (Session 5): plot retention curves by signup month. Alpine Roast's curves broke two quarters ago - every cohort acquired since retains visibly worse. Something changed then (a price move, a quality dip, a competitor) and the cohorts date-stamp it for you.
- New-buyer branch → funnel view (Session 5): traffic is fine at 90,000 visits, product views are fine - but add-to-cart collapsed after the site redesign shipped. The redesign looked prettier and converted worse. The funnel finds the exact broken step in minutes.
- WHO to win back → RFM (Sessions 5 and 7): among the 400 churned, the RFM grid flags the at-risk champions - recent big-basket, high-frequency customers who suddenly went quiet. They carry the most recoverable dollars per win-back email, so they go first.
- WHY act this week → forecast (Session 6): project the trend forward. If churn stays at 400 and new buyers keep sliding, income crosses $0 within two quarters. A forecast is how you price inaction - and $0 is a very motivating price.
The one-page memo and the weekly review 5 min live
Analysis that ends in a meeting is entertainment. Analysis that ends in a memo with owners, dates and dollar figures is management. The template is one page, answer first - Minto's Pyramid Principle from the strategic thinking course, applied to data: lead with THE ANSWER, then support it.
LiveThe 1 + 3 + 3 + 1 memo template3 min▶
One page, four blocks, in this order. Line one is the answer - not the background, not the methodology, the answer. Busy readers decide in the first sentence whether the page deserves their attention; earn it immediately.
| Block | Rule | Alpine Roast version |
|---|---|---|
| 1 · THE ANSWER | one sentence, verdict first | "Income fell 82% from two fixable causes, not the market." |
| 3 findings | each carries a number | churn doubled 200 → 400/mo · new buyers slipped 400 → 350 after the redesign broke add-to-cart · do nothing and income hits $0 within two quarters |
| 3 actions | each has owner + date + expected $ recovery | win-back campaign to at-risk champions (retention lead, 2 weeks, ~$7,500/mo) · fix the add-to-cart flow (e-commerce lead, 3 weeks, ~$3,750/mo) · stand up the weekly metric review (founder, this Monday) |
| 1 measurement line | what we re-read, when | "Re-read churned count, new buyers and add-to-cart rate in 4 weeks." |
That is the whole artifact. If it does not fit on one page, the analysis is not finished - compression is comprehension.
LiveMetric review hygiene: the lightweight WBR2 min▶
The memo fixes this quarter. The weekly metric review prevents the next one - Alpine Roast bled for two quarters because nobody re-read the same numbers regularly enough to see churn bend. The hygiene rules are boring on purpose:
- Same metrics, same order, every week. Familiarity is the anomaly detector - you only notice a bent curve on a chart you have seen 40 times.
- Anomalies get owners in the room. Every surprise leaves the meeting attached to a name and a date, or it was just theater.
- Lightweight. 30 minutes, one deck that updates itself, no new slides. The moment preparing the review costs more than an hour, it dies.
Amazon's Weekly Business Review is the famous industrial version: hundreds of metrics, same order every week, reviewed out loud, anomalies interrogated on the spot and assigned. You do not need hundreds - Alpine Roast needs about twelve. The discipline, not the scale, is the technology.
Domino's ran the culture-level version of today's loop: publicly admit the number is bad (their own pizza scored worst-in-class), instrument everything from order to doorstep, review relentlessly, ship fixes, re-measure in public. A decade later the stock had outperformed the tech giants. Same loop you just ran - diagnosis, action, re-read - held for ten years.
Self-studyThe honest boundary: what this course left out on purpose2 min read▶
You now read metrics, funnels, cohorts, RFM grids, forecasts and churn models fluently. Here is what we deliberately did NOT cover: deep learning and neural networks, recommendation engines ("customers also bought"), search ranking, and generative AI. Those systems run on different machinery, raise different questions (what data trained this? when does it hallucinate? what does it cost to be wrong?), and deserve their own literacy course - the sibling AI literacy course picks up exactly where this page ends, now live.
The good news: everything transfers. The ladder, the "so what?" test, evaluation literacy, the memo, the weekly re-read - AI systems are judged with the same habits you built here. You are not starting over; you are adding a wing to a house that now has foundations.
Re-diagnose Alpine Roast yourself ★ 9 min · everyone drives the tree
The tree below loads with the last healthy month. Reproduce the collapse with your own hands, then try to engineer the recovery two different ways - one of them will feel expensive, and that feeling is the lesson.
Reproduce the damage: set churned 200 → 400 and new buyers 400 → 350. Confirm income lands at exactly $4,250. Two quiet slider moves, 82% of income gone - feel how silently it happened.
Now engineer the recovery WITHOUT touching churn: raise basket size, raise item value, cut cost - whatever it takes to get income back near $23,000. Notice what it demands: roughly a full extra item in every basket, or a third of the cost base gone.
Reset, and instead halve the churn damage: set churned to 300. Compare how far one retention fix moves income versus all that compensating.
Debrief: fixing the leak beats compensating for it, every time. Squeezing basket, price or cost to paper over churn strains the whole business; plugging the churn branch just puts the money back.
Fix the funnel that broke new buyers ★ 8 min · everyone finds the leak
The second red branch: new buyers slipped after the site redesign. Here is Alpine Roast's acquisition funnel as it looks today. One stage is quietly on fire.
Find the flagged leak: add-to-cart is converting at 17%. Before the redesign it ran at 30%. Traffic never moved - the redesign broke the step where browsing becomes buying.
Restore add-to-cart to 30 and read the monthly revenue recovered off the widget. Say the number out loud - that is what the prettier product page has been costing every month.
Reset add-to-cart to 17 and try doubling visitors instead - the classic "just buy more traffic" reflex. Compare the price tags: pouring traffic into a leaking funnel versus fixing the leak.
Write the memo's action line 1 from this, in the full format: fix the add-to-cart flow - owner, date, expected monthly $ recovery. You just turned a widget into a commitment.
Write the memo ★ 5 min · everyone drafts
Using the 1 + 3 + 3 + 1 template from Part 3, draft the one-page memo for Alpine Roast - or for your own business if you brought numbers. Line one is THE ANSWER; no warming up.
Check every finding carries a number, and every action carries an owner, a date, and an expected $ recovery. A finding without a number is an opinion; an action without an owner is a wish.
Finish with the measurement line: what exactly gets re-read in 4 weeks, by whom.
One volunteer reads theirs aloud, answer-first. Notice how different the room feels when the verdict lands in the first ten seconds instead of the last slide.
This month ◐ the loop, for real
- Run the full loop on YOUR business once this month: tree → drill → memo → 4-week re-read. Pick your ugliest metric, walk it down the tree, drill the moved branch, write the 1+3+3+1 page, and put the re-read date in your calendar today.
- Start the lightweight weekly metric review: pick ~12 metrics, same order every week, 30 minutes, anomalies leave with owners. Give it eight weeks before judging - familiarity is what makes it work.
- Revisit any session page whenever you need it - the widgets, cheat sheets and cases are yours now. This site does not expire when the course does.
Three questions before you go 🎯 ◐ 90 seconds
1 · Monday's dashboard shows income down 40%. Your first move?
Walk the tree, not the org chart. Alpine Roast's collapse had nothing to do with marketing spend or cost - the arithmetic isolated churn and add-to-cart in minutes, and named the right two owners.
2 · Why does the memo lead with the answer instead of building up to it?
Answer-first is about the reader, not the writer. The findings and actions still do all the supporting work - they just support a verdict the reader already holds, instead of a mystery they may never finish.
3 · What actually makes a weekly metric review work?
Familiarity is the anomaly detector: you spot the bent curve because you have seen the healthy one 40 times. And a surprise without an owner attached is just theater - Alpine Roast's two quiet quarters prove what happens without both.
Concepts covered & their origins
The capstone credits the whole course's canon - and draws its honest boundary out loud.