Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Growth Marketing › Expert

Checkpoint: Growth Models, Growth Accounting and Leading Indicators

Scenario questions set at an invented meditation app with monthly and annual plans test the first expert growth module: what a growth model is for, how cohorts and returning subscribers feed it, how scenarios differ from sensitivity, why a bridge can balance and still mislead, what the revenue bridge adds and when a survey score has earned a place as an early warning. Each answer gives the reasoning and the guide behind it.
CheckpointExpert levelModule 1: Modelling the growth engine5 minUpdated

What this checkpoint covers

Ulquastra Meditation is an invented meditation app sold on monthly and annual plans; an annual subscriber stays active until the term ends. The questions apply the tutorial on building a growth model in a spreadsheet, the growth accounting tutorial and the tutorial on testing whether survey scores predict churn, and assume the guitar lesson app exercise.

Illustrative example — not a real result

Ulquastra Meditation: figures used in the questions

  • Question two: back-test gap against billing, last four months: −150, −310, −470, −620. Annual plans rose from 25% to 45% of new subscribers this year.
  • Question three: "Budget raised" adds 2,400 subscribers by next December, its extra spend divided by the average cost per new subscriber; "Onboarding lands" adds 1,100.
  • Question four, April, the month annual plans were promoted, as exported by subscription:
OpeningNewResurrectedChurnedRetainedClosing
30,5003,9006004,10026,40030,900
  • Question five, April recurring revenue in monthly equivalents: opening £252,000; new £19,000; resurrected £5,000; expansion £1,000; contraction £6,500, of which £6,000 from 1,500 switches to the annual plan; churned £23,500; closing £247,000.
  • Question seven, monthly-plan subscribers cancelling voluntarily within 3 months of a 1 to 5 rating from the in-app rating, whose invitation says answers are linked to the account; each cell gives answers, share cancelled and its 95% interval:
BandFirst half-yearSecond half-year
Low, 1 to 246, 30% (18% to 45%)38, 21% (11% to 37%)
Middle, 3410, 17% (13% to 21%)380, 15% (12% to 19%)
High, 4 to 51,620, 14% (12% to 16%)1,540, 13% (11% to 15%)
Invited, did not answer19%18%

Task

  • Note your call on each Ulquastra scenario and the reason for it, then open the explanations.
  • A right call with the wrong reason is a miss, and each miss sends you to the guide section it links.

Questions

  1. The chief executive wants next year's target set at last year's growth rate, and finance wants the paid social budget pasted into the model's acquisition row. What is the model for, and what goes in that row?
  2. Look at the back-test in the figures. The sheet uses one retention curve for both plans, taken from cohorts two years old, and has no resurrection row. What do you fix?
  3. The chief financial officer reads the two scenarios and concludes the budget should rise. What is missing?
  4. The April bridge balances on both checks, and the analyst wants to report record new subscribers. Do you agree?
  5. Subscribers rose in April, recurring revenue fell. Is the annual plan promotion failing?
  6. Product shows that subscribers who cancelled in May gave lower ratings in May than those who stayed, from the quarterly survey promised to be anonymous, a separate survey from the in-app rating. Does the rating warn of churn?
  7. Read the rating results in the figures. Does the score go on the retention dashboard?

Answers explained

  1. A check on what the inputs can deliver, with customers as its acquisition input. The model says where the base goes if the plan holds, which lever matters most and where gains meet losses, which a growth rate cannot show. Spend becomes subscribers along a flattening curve, converted outside the sheet. See what a spreadsheet growth model is for.
  2. Add resurrection, then split the layers by plan. A gap growing one way means a wrong input, and running low means missing customers: add returners as a rate on the dormant pool. Annual subscribers pay until renewal, so a blended curve from old cohorts misstates a changed plan mix. Record the gap left. See applying retention by cohort age.
  3. A sensitivity table, and a budget built on a response curve. Scenarios change several inputs together; equal one-input steps say which lever is strongest. Dividing spend by average cost ignores diminishing returns. See scenarios and a sensitivity table.
  4. No. Rows keyed to subscriptions turn every plan switch into a churned monthly plan and a new annual one, so new and churned both rise by the same amount and both checks still hold. Key rows to customers: a switch is retention, and April's new subscribers drop back. See one row per customer per period.
  5. Not on this evidence. A switch to the cheaper annual equivalent is retention in the customer bridge and contraction in the revenue bridge, so the two disagree by design. Whether it pays shows at renewal, if the switchers' Cohort layers outlast monthly ones. See the bridge for recurring revenue.
  6. No. The rating has to come before the cancellation, and the follow-up period is set before anyone looks; a same-month comparison shows neither. Answers promised anonymity cannot be joined to accounts at all, so the test belongs to the in-app rating, whose invitation discloses the link. Annual subscribers cannot leave before renewal, so compare within plan, with voluntary churn kept apart. See linking answers and fixing the window.
  7. No, not yet. The low band stood apart from the high band once; with few answers, its second Confidence Interval overlaps the others, and the middle and high bands overlap in both periods. A gap seen once is unproven: pool periods or collect longer. Non-responders leave more often than the middle and high bands, though less often than the low band, so compare those who answered with all invited subscribers on tenure and segment before trusting the bands. No Leading Indicator claim yet. See reading the result.

Where to go next

If the model and bridge questions tripped you, rebuild the guitar lesson app's sheet and bridge before moving on; if the rating questions did, work back through the survey tutorial's steps on linking, the window and non-responders. Then continue to the podcast hosting trial exercise, or revisit the advanced growth level.

Pages this lesson applies

Terms used on this page

Also related

Related