MVPSTUDIOSIDEAS · PRODUCTS · PEOPLE Start a conversationLet's talk
← Week 07: Map at scale: why people stay, leave and pay SESSION 01 · TUESDAY

From “is it real?” to “why do they stay?”

Loop 2 starts with the people already using your product: who came back, who left, and what that tells you to work on next.

Retention cohortsAARRRVanity vs actionableOpportunity treeEdge interviews
Present slides Full screen, one slide at a time · 12 slides
TONIGHT 01

By the end of tonight you can…

  • Explain how loop 2’s questions differ from loop 1’s: stay, leave and pay
  • Build a weekly retention cohort table from your events, every cell shown as a count with its base
  • Use the AARRR stages to locate your biggest leak, and tell vanity metrics from actionable ones
  • Plan interviews with churned users, power users and people who nearly left
  • Map one outcome, its opportunities, solutions and experiments on an opportunity solution tree

OUTPUT · Retention table + metrics tree + opportunity solution tree

MVPAI · BACK TO MAP 02

Back to Map, with evidence

Loop 1 at Map: guesses

  • Does this problem exist, for whom, and how often? Asked in a handful of conversations
  • Will anyone try the first slice? A guess until the first users arrived
  • Is £8 a month the right price? An assumption, written down to test later
  • Evidence: conversations, tests and a first funnel of a few people

Loop 2 at Map: numbers

  • Who comes back week after week, and who drifts away? A cohort table says
  • Where in the journey do we lose the most people? The biggest leak, with counts
  • Who would pay, and what makes them value it? Real tutors’ answers, with a base
  • Evidence: behaviour over time, edge interviews, payment

Loop 1 asked “is it real?” Loop 2 asks why people stay, leave and pay, and answers with numbers.

THE IDEA 03

A retention cohort table, read in plain English

  1. ACTIVEDecide first what counts as active: the value event, e.g. a booking confirmed that week
  2. COHORTOne row per start week: everyone whose first use fell in that week
  3. COHORT SIZEHow many started that week. It is the base for every cell in the row
  4. WEEKS 1–4Columns: how many of the cohort were active 1, 2, 3 and 4 weeks after starting
  5. READ ITAcross a row: where the drop flattens. Down a column: are newer cohorts doing better?

Every cell is “x of n”. Write down what “active” means before you count, then don’t change it.

WORKED EXAMPLE · TUTORDESK 04

TutorDesk’s first cohorts: who came back

Starting point

“We have 20 tutors and most of them still log in. Retention looks great.”

Reframed

Illustrative. Active = confirmed a booking that week. Launch cohort: 5, 5, 4, 4 of 10 in weeks 1–4. Next: 5, 4, 4 of 6. Newest: 3 of 4. Busier tutors: 4 of 5 at week 4; others 0 of 5.

Ask

  • Where does the launch cohort’s line flatten, and what might that mean?
  • Is the second cohort doing better, or are 6 tutors too few to say?
  • Why is the newest cohort’s row mostly empty?
  • Half the launch cohort never confirmed a booking. Why might that be?
THE IDEA 05

AARRR: five stages, one biggest leak

  1. ACQUISITIONPeople arrive and sign up. TutorDesk: 20 sign-ups from about 60 visitors
  2. ACTIVATIONThey reach first value. TutorDesk: 13 of 20 confirmed a first booking
  3. RETENTIONThey come back. TutorDesk: 4 of the 5 activated launch tutors active in week 4
  4. REFERRALThey bring others. TutorDesk: 2 of 20 tutors came via another tutor
  5. REVENUEThey pay. TutorDesk: none yet; £8 a month is under test

Fix the leakiest stage first. Pouring more people in at the top of a leaky funnel wastes them. (Illustrative.)

THE IDEA 06

Vanity metrics vs actionable metrics

Vanity metrics

  • Total sign-ups since launch: only ever goes up
  • Page views and app opens, whatever happened next
  • Percentages with no base: “60% retention”
  • Numbers that feel good but change no decision

Actionable metrics

  • Tutors confirming a booking in week 4, by cohort
  • Requests confirmed within 24 hours, of requests made
  • Counts with bases, split by a segment that matters
  • Numbers that tell you what to do differently

Test any metric with one question: if it moved, would we do something different? If not, stop reporting it.

THE IDEA 07

The opportunity solution tree

  1. OUTCOMEOne measurable result: more new tutors confirm a booking within 14 days
  2. OPPORTUNITIESNeeds and pains from evidence: “I miss requests buried in WhatsApp”
  3. SOLUTIONSSeveral ideas per opportunity: instant alerts, a slot proposal, a reschedule link
  4. EXPERIMENTSSmall tests of a solution’s riskiest assumption, before building it

Solutions hang from opportunities, and opportunities from evidence. An idea with nothing above it is a guess.

THE IDEA 08

Talk to the edges, not the middle

  1. CHURNEDStopped using it. Ask about the last week they used it and what replaced it
  2. NEARLY LEFTStalled, then came back. What happened, and what pulled them back?
  3. POWER USERSUse it most. What do they use it for, and what would they do without it?
  4. HOW TO ASKPast behaviour, not opinions: “Tell me about the last time…”, as in week 1

Churned users tell you what is broken; power users tell you what is valuable. You need both.

CLAUDE STUDIO · CLAUDE.AI 09

Analyse an event export into cohorts, honestly

Prompt for Claude
I’m looking at whether people keep using [MY PRODUCT], so that I can choose what to fix first. Attached is an event export, one row per event: a user ID (no names), an event name and a date. I’ll put your table in my Product Book, so it must not claim more than the data shows.
<data>[ATTACH CSV]</data>
Group users by the week of their first [START EVENT]. Count someone as active in a week if they did [VALUE EVENT] that week. Show a table: cohort, cohort size, then weeks 1–4, each cell as “7 of 10”. Below it, say how you defined weeks, which rows you dropped and why, and which cells rest on fewer than 10 people. If you’re unsure what a column means, ask before calculating. Done means the table, those notes, and one sentence on what it suggests, with your confidence.

Check the response

  • Every cell shows a count with its base, not a percentage
  • Cohort sizes and small samples are stated plainly
  • One cohort is checked by hand against the raw export
WORKSHOP · YOUR PRODUCT 10

Retention table, metrics tree and opportunity tree

Find out who stays and who leaves your product, where the biggest leak is, and which opportunity to work on first.

  1. Export events, build the retention table with Claude, check one cohort by hand (18 min)
  2. Draw the metrics tree: North Star on top, AARRR stages beneath with counts and bases (12 min)
  3. Mark your biggest leak and name the segment that leaks most, with its counts (5 min)
  4. Build the opportunity tree: one outcome, 3+ evidenced opportunities, solutions, one test (17 min)
  5. Name three edge users (churned, nearly left, power), and write the invite and date (8 min)
Deliverable

Table, metrics tree and opportunity tree in product/07-insights.md

GATE 11

Retention and opportunity rubric

  • The table groups users by start week and shows every cell as a count with its base
  • It states what counts as “active” and flags cohorts too small to trust
  • The metrics tree links the North Star to AARRR stages, each counted, with one biggest leak marked
  • The opportunity tree has one measurable outcome, 3+ opportunities, solutions and one experiment
  • Every opportunity cites evidence: an event count, a retention cell or a quote
  • Three edge users (churned, nearly left, power) are named, with an invitation date for each

HOMEWORK · Hold two edge interviews and gather your feedback (messages, reviews, notes) into one file for Thursday.

© 2026 MVP Studios Ltd. MVPAI™ and the MVP Founders Programme materials are original works of MVP Studios Ltd. All rights reserved. No copying, redistribution or commercial use without written permission. Course terms