Three x-rays for one customer journey
Every customer walks a path: they discover you, try you, pay you, and either return or quietly vanish. Funnels x-ray the path forward (where do they leak?), cohorts x-ray the path back (do they return?), and RFM x-rays the crowd (who deserves which message?). Master these three and you can interrogate almost any growth claim anyone puts in front of you - which is precisely the job of a data-literate leader.
Funnel analytics 5 min live
Every business is a funnel: many enter, few pay. The magic number is not the count at any stage - it is the conversion rate between stages, because that is where the money silently leaks out.
LiveThe leak-first principle3 min▶
Three rules of funnel reading:
- Conversion between stages beats counts at stages. "13,500 added to cart" is information; "only 30% of viewers add to cart, versus 45% and 65% elsewhere" is the insight.
- Fix the weakest stage, not the loudest one. The loudest stage is usually the top (marketing wants more traffic) or the bottom (finance stares at purchases). The weakest stage is wherever conversion sags - and fixes there multiply through everything downstream.
- Leaks compound. The funnel is a chain of multiplications, which you already know how to read: it is a KPI tree lying on its side.
Baymard Institute's running meta-study: average documented cart abandonment is roughly 70% - seven of ten loaded carts never check out. A top killer is disclosing shipping costs late: the price grows at the last step and trust snaps. The fix (show total cost early) is a product tweak, not an ad budget - a classic weakest-stage win that no amount of top-of-funnel spend could buy.
Self-studyAARRR: pirate metrics, seriously2 min read▶
Dave McClure's AARRR framework (yes, named for the pirate noise) is the funnel generalized for any product: Acquisition (they find you) → Activation (they experience first value) → Retention (they come back) → Referral (they bring friends) → Revenue (they pay). Its power is diagnostic: a struggling business is usually failing at exactly one letter, and each letter has a different fix. Dropbox's famous referral program (free space for both inviter and invitee) was an R-for-Referral play chosen only after activation and retention already worked - referral pours fuel, and fuel on a leaky funnel just burns faster.
Cohorts and retention 5 min live
A cohort is simply a group that started together - the January signups, the Black Friday buyers. Track each cohort separately over time and you get the single most honest chart in analytics: the retention curve.
LiveWhy cohorts beat averages3 min▶
Fast growth is a magician's cape: it hides churn. When signups grow 20% a month, the average user is always young, engaged and shiny - even while every cohort quietly bleeds out behind the curtain. Total active users climbs; the business is dying. Only a cohort view separates the two stories, because it refuses to let new users answer for old ones.
- Read the shape, not the level: a curve that flattens (even at 25%) means a durable habit; a curve that slides to zero means a leaky bucket, whatever this month's topline says.
- Compare cohorts to each other: newer cohorts flattening higher = the product is improving. Flattening lower = growth is buying worse customers.
Facebook's early growth team mined cohort data for the difference between users who stayed and users who vanished, and landed on a famous rule: get a new user to 7 friends in 10 days and they stick. That single activation insight reorganized the entire onboarding flow around friend-finding - not profile completion, not feature tours. One cohort analysis became the company's growth strategy.
RFM: segmentation you can run in a spreadsheet 5 min live
Recency, Frequency, Monetary - Arthur Hughes built it in the direct-mail era, and it is still the workhorse of e-commerce CRM because it needs exactly three columns you already have: last purchase date, purchase count, total spend.
LiveScore, segment, act - the whole method3 min▶
Score every customer 1-5 on each letter: Recency (how recently they bought - 5 is most recent), Frequency (how often), Monetary (how much). Segments fall out of the combinations:
| Segment | Profile (R / F / M) | The move |
|---|---|---|
| Champions | 5 / 5 / 5 - recent, frequent, big | reward, early access, ask for reviews - never discount |
| Loyal | 4 / 4 / 3 - steady regulars | upsell gently, build the habit loop |
| New | 5 / 1 / low - just arrived | onboard toward the second purchase fast |
| At-risk champions | 2 / 5 / 5 - valuable but slipping away | win-back NOW, personal touch, highest priority |
| Hibernating | 1 / 1 / low - long gone, barely knew you | cheap reactivation at most, or let go gracefully |
The whole point: stop sending the same email to everyone. A discount that delights a hibernating customer insults a champion (who would have paid full price) - one blast, two opposite effects, both invisible in the average open rate.
Sephora's Beauty Insider and Starbucks Rewards are RFM-style machines at loyalty-program scale: tiers keyed to spend and visit frequency, with different perks, nudges and win-back offers per tier. And when win-back budgets are tight, the play is always the same: target the at-risk champions first - they were recently valuable, so recency says they are the most likely to answer and the most expensive to lose.
Self-studyYour industry's metric pack3 min read▶
Every industry has a standard dashboard. Knowing yours - and your investors' - is table stakes:
| Industry | The pack | The one to watch |
|---|---|---|
| Internet / consumer apps | DAU, MAU, DAU/MAU stickiness, session length | DAU/MAU - above ~20% means a habit, not an app |
| SaaS | MRR/ARR, NRR, churn rate, CAC, LTV, CAC payback months, magic number | NRR - above 100% means existing customers grow even with zero sales |
| E-commerce | GMV, AOV, conversion rate, repeat purchase rate, cart abandonment | repeat purchase rate - the difference between a brand and a transaction |
Notice the family resemblance: every pack is a funnel metric + a retention metric + a money metric. Same three x-rays, industry costume.
Find the leak worth fixing ★ 9 min · everyone simulates
The Part 1 funnel, live. Two executives propose two fixes; you have the simulator they did not.
Read the baseline: 100,000 visitors trickle down to purchases at $80 average order value. Say the end-to-end conversion out loud - it is smaller than most people guess.
Find the flagged weakest stage. (It is the add-to-cart step at 30% - the Baymard story from Part 1 predicts exactly this kind of sag.)
The boardroom duel: the CMO proposes doubling ad spend - set visitors to 200,000 and note the revenue. Reset. The product lead proposes fixing add-to-cart from 30 to 40 - try that and note the revenue. Now compare the two outcomes per dollar of effort: media budgets rent traffic monthly; a checkout fix is built once.
Free play: find the smallest single change that adds $1M of annual revenue. Compare answers across the room - there are several paths, and they disagree in interesting ways.
Score three customers yourself ★ 8 min · everyone scores
RFM only sticks once you have scored someone. Three imaginary customers, three verdicts, three different emails.
Customer 1 - the friend of the brand: bought last week, buys monthly, spends big. Score them. (Expect a Champion - note the recommended move is a reward, not a discount.)
Customer 2 - the big spender who vanished: two large purchases, but the last one was a year ago. Score them and read the segment carefully - this is the at-risk profile the win-back budget exists for.
Customer 3 - the newcomer: one small purchase last week. Score them. The move is onboarding toward purchase two, not loyalty perks they have not earned.
Discuss: which segment does your current marketing actually talk to? In most companies the honest answer is "everyone gets the same blast" - which means champions are subsidizing the discounts and hibernators are ignoring them.
Draw YOUR funnel ★ 5 min · everyone draws
Sketch your own business as a funnel: 4-6 stages from first touch to money. B2B works too - lead → qualified → demo → proposal → closed.
Guess the conversion rate at each stage. Write the guesses down - next week, ask your analyst for the real numbers and enjoy the humility.
Mark the stage nobody owns. Almost every room finds one - typically the handoff between two departments (marketing to sales, or purchase to second purchase).
Circle your best guess at the weakest stage. That circle plus Session 4's tree tells you where next quarter's effort belongs.
This week ◐ 30 min total
- Get the real funnel numbers for the stages you sketched in Apply-along 3 and compare against your guesses. Bring the biggest surprise to next session.
- Ask your analyst for one cohort chart: monthly signup (or first-purchase) cohorts, with retention read along each row. Check the shape first: flattening or diving?
- Run baby RFM in a spreadsheet: your top 20 customers, three columns (last purchase, count, total spend), score each 1-5, and find your at-risk champions. It genuinely takes 15 minutes.
- Optional deep end: Croll & Yoskovitz, Lean Analytics on funnel and retention metrics, plus Dave McClure's original "Startup Metrics for Pirates" deck - still funny, still right.
Three questions before you go 🎯 ◐ 90 seconds
1 · Your funnel converts: visit → view 50%, view → cart 30%, cart → checkout 45%, checkout → purchase 90%. Where do you put the team first?
The leak-first principle: 90% is already excellent, and traffic poured into a leaky mid-funnel mostly leaks. Lifting view-to-cart from 30% to 36% raises purchases 20% with zero extra media spend - the compounding works for you.
2 · Active users grew 20% this quarter. The CMO says retention is clearly fine. What does the data-literate leader say?
Growth is a magician's cape. Total actives can climb for quarters while every cohort dives to zero - the leaky bucket refilled faster. Only cohort curves separate a habit from a treadmill.
3 · RFM stands for - and why does R lead the win-back list?
Hughes' direct-mail finding still holds: how recently someone bought predicts response better than how much or how often. That is why the at-risk champion (high F, high M, fading R) is the most urgent - and most winnable - save in the file.
Concepts covered & their origins
This session teaches the working 80% of each idea and credits the canon - the books and decks stay the deep end for anyone who wants more.