DATA NOTE: visualizations use reconstructed datasets shaped to match the real, reported outcomes of this work. Raw user data is confidential. The generator script and seed are public in this site's repository.
CONTEXT
I founded Distance Connect and ran it for 18 months. No growth budget, no data team. Every product decision had to come from instrumentation I built myself, and every wrong call cost weeks.
QUESTION
Which levers actually grow this product, and which ones just feel like growth?
WORK
The first bottleneck was onboarding. Session recordings showed users stalling, so I ran an A/B test: the original 5-step flow against a stripped 3-step flow.
FIG 1.1 — ONBOARDING A/B FUNNEL
5-STEP (CONTROL)3-STEP (VARIANT)
Cutting onboarding from 5 steps to 3 lifted activation from 34% to 52%.
Activation fixed, the question became retention. I cut every weekly cohort by acquisition channel. One segment refused to churn.
FIG 1.2 — COHORT RETENTION HEATMAP
W0W1W2W3W4W5W6W7
C01
C02
C03
C04
C05
C06
C07
C08
C09
C10
C11
C12
Retention decays as expected overall. The signal only appears when you split by channel.
That finding restructured the growth model. I rebuilt the KPI tree around the channels and behaviors that actually compounded.
FIG 1.3 — NORTH-STAR KPI TREE
NORTH STAR · ACTIVE USERS
5,000+
▲ 35% MoM
ACTIVATION
34% → 52%
3-step onboarding, FIG 1.1
30-DAY RETENTION
68%
WhatsApp-led referral loop
LTV : CAC
Organic-led
Referral kept CAC near zero
One tree, three levers. Growth compounded at 35% month over month.
DECISION
Ship the 3-step onboarding and rebuild acquisition around the WhatsApp referral loop. Both calls came from the data above, argued in writing, before a line of code changed.
RESULT
7-day retention rose from 42% to 58% after the onboarding A/B. The product passed 5,000 users in 18 months with no paid acquisition, growing 35% month over month at peak.
WHAT I'D DO DIFFERENTLY
Instrument the referral loop from day one. I found the WhatsApp cohort effect six months later than I could have, because channel attribution was an afterthought in the first schema.
METHODOLOGY ▸
Weekly cohorts, retention measured as return-visit within week window. Onboarding A/B split 50/50 at signup, evaluated on 7-day retention and step completion. Heatmap datasets on this page are reconstructed to match reported outcomes; the generator script and seed are public in the site repository.
The 2.4x day-30 delta compares WhatsApp-referred cohorts against all other channels, same weeks.