A regulated investing and learning app
2026
Fractional Head of AI and Transformation
Diagnosis + operating model
Finding the activation bottleneck in a subscription funnel
An event-level read of a 192-day mobile export shows where installs stop becoming paying subscribers, and the operating model that follows from it.
At a glance
0.15%
of installs become paid
77%
uninstall before a second session
7%
day-1 retention vs a 25–30% benchmark
93×
lift in payment likelihood after setup
The problem
6× the subscribers in 11 months, with spend going into the top of the funnel.
The business needed to grow its paid subscriber base about 6× in 11 months. Spend was going into acquisition. So I started from the event data to find where the funnel breaks.
Required-run-rate math set the size of the gap. The event export then showed which stage was losing the most people.
The analysis
Chart A · Install to paid, % of installs
99.85% of installs never become paid, and most are lost before the second session.
Source: Mobile analytics event export, 192 days (Nov 2025 to May 2026), anonymised. Bars are true to scale; the smallest carry a minimum visible width.
14%
of users who complete setup go on to pay.
The lever is setup completion. Only 1.1% of installs get there today.
Chart B · Churn sensitivity
Cutting monthly churn from 7% to 5% adds about 14 retained subscribers every month.
Monthly churn today
Target
Source: Churn sensitivity on the subscriber base, anonymised.
Chart C · Day-1 retention
Day-1 retention sits at 7%, against a benchmark of 25–30%.
0%30%
Source: Mobile analytics event export, 192 days, anonymised; day-1 retention benchmark.
The operating model
5 stages, one owner each, and a target the owner can be held to.
AGrowth
Acquire
500+ qualified leads/day by Oct 2026
BPM + Growth
Onboard
Day-1 retention 7% → 20%
CPM
Activate
Setup completed 1.1% → 5% of installs
DPM
Convert
Trial-to-paid 35% by Q4 2026
EPM
Renew
Renewal rate 70% by Q1 2027
Foundation · Engineering
Engineering
Day 7
Event tracking + user-level joins
Day 14
Cohort dashboard (D1/D7/D30 retention)
Day 30
Performance (LCP < 2.5s, crash < 1%)
The measurement foundation is a prerequisite: product can't operate without it.
Risk register
4 of 5 risks score red, and the first one gates everything else.
Severity = likelihood × impact (High = 3, Med = 2). Red at 6 or above, amber at 3 to 5. Reviewed weekly.
ID
Risk
Likelihood / impact
Score
Mitigation
R1
Measurement foundation slips
High / High
9 RED
Lock engineering scope before the PM hire. Non-negotiable gate.
R2
Instructor incentive fails
Med / High
6 RED
Pay tied to setup completion and trial-to-paid per cohort.
R3
Trial baseline below 35%
Med / High
6 RED
Establish baseline in 30 days; iterate activation before assuming a structural miss.
R4
Acquisition stays flat
Med / High
6 RED
Direct channels and a graduate referral loop, instrumented from day one.
R5
Legacy-tier migration churn
Med / Med
4 AMBER
Loyalty discount phased over 3 months; CEO sign-off first.
What it changed
Fix activation first, and measure everything from day one.
The recommendation is 4 success factors in sequence: course-graduate trials, activation improvement, acquisition channel fix, measurement foundation. Activation moves first because it is where the funnel loses the most, and where a completed setup is worth the most.
02Activation improvement
Moves first.
03Acquisition channel fix
- Tie owners and dates to each stage target, reviewed weekly through the risk register.
- Gate the PM hire on locked engineering scope for the measurement foundation.
- Pay instructors on setup completion and trial-to-paid, per cohort.
- Phase any legacy-tier loyalty discount over 3 months, with CEO sign-off first.
Alongside the plan, a trilingual (Urdu, Roman Urdu, English) lead-qualification agent routed inbound leads.
Method notes
Event-level install-to-paid funnel built from a 192-day mobile analytics export (Nov 2025 to May 2026).
All funnel stages are shares of installs; the 14% figure is a share of users who completed setup.
Required-run-rate math sizes the 6× in 11 months gap; churn figures are a sensitivity, not a forecast.
Day-1 retention is compared with a 25–30% benchmark; the 93× lift is observational, from user behaviour in the export.
Stage targets and risk scores are a plan, and no realised outcome is claimed here.