Brief
Glovaskin Vintage is an invented app that connects people selling vintage clothing with buyers. Sellers pay a monthly plan after a free trial; buyers pay per order, and because orders are collected or delivered by local courier, the app opens area by area. Growth has stalled. Seller sign-ups rise every quarter but active sellers barely move. Buyers in the newest area find too few listings. A seller satisfaction score is reported every week without anyone knowing whether it predicts anything. And the team runs tests without a meeting that turns their results into decisions.
The advanced growth level audited experiments, read incremental growth and set prices. This plan is about running growth as a system: knowing where the base is heading, which signals deserve attention, which experiments to run in what order, which side of the market to grow where, and how the team decides each week.
Inputs and constraints
Glovaskin Vintage: the seller base over the last four quarters
| Quarter | Trial starts | Active sellers at start | New | Resurrected | Churned | Active sellers at end |
|---|---|---|---|---|---|---|
| First | 1,900 | 1,800 | 420 | 60 | 400 | 1,880 |
| Second | 2,300 | 1,880 | 510 | 55 | 480 | 1,965 |
| Third | 2,700 | 1,965 | 600 | 50 | 590 | 2,025 |
| Fourth | 3,100 | 2,025 | 680 | 45 | 690 | 2,060 |
- Seller plan £12 a month after a 14-day trial with no card at sign-up; about 22% of trials pay at the trial's end. Most sellers who later pay list a first item within three days.
- Paying sellers still paying, older cohorts against the last two quarters' cohorts: second month 74% and 66%, third month 63% and 55%; sixth month 49% and twelfth month 41% for older cohorts only.
- Active buyers by quarter: 9,400, 10,200, 10,900 and 11,300. Buyers pay a £1.50 fee per order, shown from the first screen. Monthly cohorts exist for both sides.
Glovaskin Vintage: listings, orders and spend by area last quarter
| Area | Active sellers | Searches with no matching listing | Listings sold within 30 days | Sellers with a sale | Buyer ads | Seller recruitment |
|---|---|---|---|---|---|---|
| Area A, opened first | 1,240 | 5% | 18% | 52% | £3,000 | £4,200 |
| Area B | 610 | 11% | 31% | 70% | £6,000 | £300 |
| Area C, opened last year | 210 | 38% | 57% | 88% | £18,000 | £0 |
Glovaskin Vintage: the survey, the cancellation flow and the test log
- Survey: sellers score satisfaction from 1 to 5; 1,140 answers last year. Since March the invitation has said answers are linked to the account (780 answers); earlier answers were anonymous.
- Cancellation: about 260 sellers start cancelling each month. Every one sees half price for three months, with the cancel option as a grey text link leading to a second confirmation screen. 34% accept, reported weekly as "saves". The offer was never tested.
| Test last half-year | How it was read | Call |
|---|---|---|
| Seller onboarding checklist | First look on day five | Win, shipped |
| Trial-ending reminder email | Trial-to-paid rate at the trial's end | Win, shipped |
| £5 first-order credit for buyers in Area C | Orders before and after, no comparison area | "Worked" |
- Weekly meeting: ninety minutes on 22 dashboard charts, no decision log.
Constraints:
- The team is a growth lead, a part-time analyst, a product manager, a community manager who recruits sellers and engineers shared with the product roadmap, who can build one experiment in the trial or cancellation flows at a time.
- The year's budget is fixed; money can move between sides and areas, not grow.
- Cancelling must be easy, with no obstacle added to the exit; take qualified advice on the flow.
- Survey answers are linked to accounts only where the invitation said so.
- State expected effects as a direction with a way to check them; invent no uplifts.
Task
Write one operating system document with these sections, in this order.
- The growth model. Build a model for each side with the tutorial on building a growth model in a spreadsheet: seller layers from monthly trial cohortsA group of customers who share a starting event and period, tracked together over time. with retention by paid month, recent cohorts for the early ages, resurrection from the bridge, a back-test, scenarios, a sensitivity table and the ceiling. Balance every quarter for both sides, by area, with the growth accounting tutorial, and tie the north star to its inputs with the KPI tree framework.
- The leading indicator. Test the survey score against later churn with the tutorial on testing whether survey scores predict churn, written up in the survey analysis plan template. Say what the score is for and where it sits until the answer is in.
- Trial experiments. Design the trial length and card-at-sign-up tests from the free trial length test and the card-required trial test: primary metric, horizon, guardrails, sample, duration and decision rule for each, in the experiment hypothesis template, and the order they run in.
- The save offer. Decide what changes in the cancellation flow before any test, then design the test with the cancellation save offer test and the guide to designing the cancellation experience.
- Two-sided growth. Read liquidity by area, name the side that limits orders in each, move acquisition and channel credit to match, and set the rule for opening the next area, following the guide to growing a two-sided marketplace and the guide to measurement for marketplaces.
- The weekly rhythm. Redesign the meeting with the tutorial on running a weekly growth meeting: owner, attendees, standing agenda and pre-read, capacity and owners for the next tests, and learning-log entries. Re-call every logged test on the experiment results report template, and keep decisions in the marketing decision log.
- The year on one page. The order of tests given one build at a time, each read date, and what is decided weekly and each quarter.
Evaluation criteria
| Criterion | What a strong answer shows | Common gap |
|---|---|---|
| Growth model | Builds both sides' models from cohorts with resurrection, scenarios and sensitivity, and balances each quarter with growth accounting | A forecast that multiplies last month's growth rate |
| Leading indicator | Tests the survey score against later churn in the app's own data before using it as an early warning | A weekly score reported because it exists |
| Trial experiments | Designs trial length and card-at-sign-up tests on paying sellers at a fixed horizon with refund and churn guardrails | Trial-to-paid rate used to declare a winner |
| Save offers | Tests the save offer on retained revenue net of the discount with an easy exit | A discount that hides the way to cancel |
| Two-sided growth | Finds the side that limits matches by area, plans acquisition per side, measures liquidity and expands area by area | Buyer acquisition raised while sellers are the constraint |
| Weekly rhythm | Runs a weekly meeting with a pre-read, test read-outs, decisions, capacity and a learning log | A status update with no decisions |
A strong plan is mostly a matter of order. The bridge shows churn rising faster than new sellers, and recent cohorts leaving sooner, so the model's sensitivity table, not the sign-up count, decides where effort goes. The Leading IndicatorA measure that moves before the business outcome it is expected to predict, paired with a lagging measure that confirms the outcome later. question is settled before the score is allowed into the pre-read as a KPIA metric tied directly to an objective.. The cancellation exit is fixed first, because ease of leaving is a guardrail, never a variable, and an offer can be tested only on a fair flow. The trial tests run one after the other, judged on paying sellers rather than the Trial-to-Paid Conversion RateShare of free trials that became paying subscriptions., which is also why the logged reminder email has to be re-called. And the area table is read before any budget moves: buyer money is concentrated where sellers are short, and seller recruitment where sellers already wait.
Reference figures for self-checking
- Gains over losses, new plus resurrected divided by churned: 1.20, 1.18, 1.10 and 1.05 by quarter. Net change: +80, +85, +60 and +35.
- Trial starts rose by about 63% across the year (1,900 to 3,100); active sellers rose by about 14% (1,800 to 2,060).
- Linked survey answers that can enter the test: at most 780 of 1,140, fewer once answers whose window has not closed are excluded.
- Reported "saves": about 88 a month (34% of 260), before anyone knows how many cancel when the discount ends.
- Buyer ads in Area C: £18,000 of £27,000, two thirds; seller recruitment in Area A: £4,200 of £4,500.
Submission checklist
When you have finished, compare your model with the guitar lesson app cohort model exercise, your trial tests with the podcast hosting trial exercise and your area plan with the dog walking marketplace exercise.
