Context
Marketing and product teams needed to know whether acquisition led to real customers and real loans. That depends on identity links across several systems.
Problem
Events were duplicated, identifiers were missing, iOS and Android disagreed, and clicks could not be tied reliably to loan requests.
System and insight
Ad clickDevice or sessionUserLoan requestApplication outcomeFunnel model and reporting
Product decisions are only as good as the information architecture underneath them. A funnel is a model of identity and state changes, not a chart.
Trade-offs
- A single identity bridge is simpler to reason about, but any break in an identifier loses the chain beyond that point.
- Showing missing links openly makes the numbers less impressive and more trustworthy.
Intervention
- Built an identity bridge from ad click through device, user and loan-request IDs, with funnel data modelled in dbt.
- Defined what each event means at every transition.
- Kept known gaps, such as missing identifiers and platform differences, visible.
My role
I built the attribution and funnel analytics infrastructure: requirements, identity design and funnel definitions, working with data engineers and analysts.
Result
- Full-funnel conversion measurement, from ad click to disbursement, with drop-off recovery.
Learning
Measurement is a product, not a report.