Closing the gap to plan
A quarter into the year, monthly active personal-banking customers were behind the growth plan, and sign-ups were further behind still. The Head of Growth and the CMO asked me where the gap came from and which levers could close it. I broke the target into its drivers, sized each one, and found that the lever we controlled best was a new customer's first two weeks.
Situation
The year's plan set monthly and quarterly targets for active customers. At the end of the first quarter, actives were 14% behind plan and sign-ups 26% behind, and the year-end target meant growing about 7% a month.
Task
The Head of Growth and the CMO asked me to find where the gap came from and which levers could close it by the end of the year.
Action
Broke monthly actives into sign-ups, conversion to a first transaction and early retention; sized each lever; then measured how fast customers of different eventual value make their first transaction, by product and channel.
Result
The analysis ran in Q2 and a rebuilt onboarding flow went live in August. Monthly active users grew 8–9% a month in August and September, up from 2–4% before, mostly from new customers; sign-up conversion rose 3–4pp and new-user VAS adoption 8–9pp.
Situation: behind plan after one quarter
The growth plan for the year set a target for monthly active users (MAU) in personal banking: customers who make at least one transaction in the month. The base itself was in reasonable shape. The share of registered customers who were active was in line with what banks usually see. But the plan needed the active base to grow fast, and growth has to come from somewhere: either more new customers who start transacting and keep going, or fewer existing ones drifting away.
The plan broke the year-end target into monthly and quarterly milestones. By the end of the first quarter the numbers had fallen short on both counts that mattered. Monthly actives were 14% below the plan's figure for that point, and new sign-ups were 26% below. To land the year-end target from where the quarter finished, monthly actives had to grow by about 80% in nine months: roughly 7% every month. They were growing at 2–4%.
End of the first quarter, each metric as a share of the plan's figure for that point (plan = 100). Shares only; the plan's absolute figures stay with the company.
Task: find the levers
The Head of Growth and the CMO asked me to work out where the gap came from and what could close it by the end of the year. The tempting answer is another acquisition campaign. Sign-ups were the metric furthest behind, after all. But every sign-up has an acquisition cost, and a sign-up who never transacts adds that cost without adding an active customer. Before recommending more spend at the top, I wanted to know how much of the gap sat in what happened to customers we had already paid for.
Breaking monthly actives into levers
Monthly actives in any month are the existing actives who stay, plus the new customers who start transacting and keep going. The second part splits into three numbers that marketing and product can each move:
Written this way, the levers multiply. Ten points on conversion alone means roughly a sixth more new active customers every month from the same sign-ups, with no extra acquisition spend. Add five points of early retention and a fifth more sign-ups, and each month's intake of lasting new actives is about half as large again. The calculator below lets you move each lever and see how much of the change each one contributes.
Sign-ups are indexed so the baseline month = 100; conversion and early retention are the plan's starting rates and targets. The bars split the change in lasting new actives between the three levers on a log scale, so the contributions add up exactly. A simplification: it holds the existing base fixed and ignores reactivations.
That reframed the problem. Conversion and early retention work on customers the business had already paid to acquire, so they were the cheapest levers and the ones onboarding could move directly. The question became: what makes a new sign-up transact, and keep transacting? Two pieces of analysis answered it. Case study 01 followed thirteen monthly cohorts and found that an everyday payment (airtime, data or a bill) in the first 30 days marked the customers who stayed. The second, below, asked how early that is decided.
Action: how fast do valuable customers start?
On day one a customer has no history, only speed. So I ranked a large population of customers onboarded in the first half of 2025 into five tiers by their twelve-month value, then measured how quickly each tier made its first transfer, airtime purchase and data purchase, day by day from account opening, separately for customers who signed up on their own (digital) and through an agent.
The top tiers move almost immediately. Their curves rise steeply in the first few days and bend within the first week or two; after day 14 they keep rising, but slowly. The gap between the top and bottom tiers opens in the first 48 hours and never closes. Customers who have not done anything meaningful by day 14 rarely reach the top tiers later without help. That gave onboarding a deadline, not just a goal.
Transfer
Airtime
Mobile data
Tiers are ranked on twelve-month value. The dashed line marks day 14. Hover any chart to read every tier on that day across all three products. Switch to agent-referred to see the slower, flatter start: those customers get a card first and fund the account later, and far fewer of them buy mobile data in the first month. Curves traced from the original velocity charts and lightly perturbed.
What separates the tiers at day 7
Day 7 was the most telling point. Every product separates the tiers to some degree, but by very different amounts. Card adoption barely does: the bottom tier is not far behind the others, because a card is often handed out as part of signing up. Airtime does, by about ten times between the top and bottom tiers. Whether a new customer has bought airtime in their first week says far more about where they will end up than whether they have a card.
Both onboarding channels combined. Bars are grouped by product; within each group, tiers run from the top 5% (left) to the bottom 30% (right). Values from the report's day-7 table, lightly perturbed.
Result: onboarding rebuilt around the first two weeks
What changed
I took three recommendations to the Head of Growth and the CMO, drawing on this analysis and the cohort research in case study 01:
- Get every new customer to an everyday payment inside 14 days. Airtime first, because it is common everywhere; data where adoption is already high.
- Reward the second purchase, not just the first, inside a short window, because the habit forms in days.
- Run two journeys, not one. Agent-referred customers start with a card and fund the account later; digital customers start with a transfer. One flow for both serves neither well.
The analysis ran through the second quarter. The company rebuilt its onboarding flow to steer new users towards an everyday payment early, and the new flow went live in August. Measured before and after the change, the share of sign-ups that went on to open an account rose by 3–4 percentage points and VAS adoption among new users rose by 8–9 points.
The headline number moved too. Monthly active users had been growing at 2–4% a month through the spring and early summer, short of the roughly 7% a month the year-end target needed. In August and September they grew 8–9% a month, faster than the plan's required pace for the first time in the year. Most of the extra growth came from new customers, which is where the changes were aimed.
I also designed a set of controlled experiments to separate cause from selection. Each had three arms (no contact, a message alone, and a message with a reward), so the effect of the message could be told apart from the effect of the money. Not all of them ran. That means the gains above come from a before-and-after comparison, not a controlled test, and I can't attribute them to any single change in the flow.
How I did it
- Restate the target as drivers. Monthly actives as retained existing actives plus new customers who start and keep going, with the second term split into sign-ups, conversion and early retention.
- Size each lever. Compare the plan's lever targets with the first quarter's actuals, and work out how much each would add if it moved on its own and together.
- Find what drives the cheapest levers. Cohort personas (case study 01) for what valuable customers do; time-to-first-transaction curves for when they do it.
- Rank by outcome, then look back. Five tiers by twelve-month value, then cumulative first-transaction curves by product, tier and channel from day 0 to 30.
- Turn it into a deadline. The day the curves flatten (about day 14) became the onboarding window the recommendations were built around.
Caveats and what I'd do differently
- Before and after is not a controlled test. Seasonality, campaigns and pricing changes all overlapped with the new flow, and two months of faster growth is a short run. A holdout group kept on the old journey would have made the 3–4pp, the 8–9pp and the step-up in growth far easier to defend.
- The driver tree is a simplification. It multiplies average rates and holds the existing base fixed. In practice churn and reactivation in the base moved too, and a full monthly stock-and-flow model would show how much of the year-end gap each lever could realistically close.
- Tiers are defined by the outcome. Ranking customers by twelve-month value shows what valuable customers did early, but it can't score a new customer on day one. For live targeting I'd train a model on the first fourteen days' behaviour and validate it on later cohorts.
- Card adoption in the agent channel is procedural. Cards are issued as part of agent sign-up, so they aren't a behavioural signal for those customers.