Where users leave before KYC and what to test
Vladimir RossausContext
A Telegram-wallet analysis model: explain identity verification and follow the path to first and repeat transactions.
- Project
- Fintech / Crypto
- Period
- Anonymised scenario
- Area
- Product funnel / Analysis model
Task
Where to look for the loss
In a Telegram wallet, a user arrives for a financial task, then encounters identity verification. Unexpected requirements can cause abandonment. This analysis shows where to look for a cause and which change to test.
Solution
How I would organise the work
Separate acquisition, product start, KYC start and completion, first transaction and repeat transaction. Retain the source alongside an internal user identifier. Compare losses by source and segment, then agree a test with product and the team responsible for KYC.
The change being tested
Starting path
- Document requirements appear unexpectedly
- Required data and next steps are unclear
- The user decides whether to continue
Test hypothesis
- Explain the purpose and required data before verification
- Give realistic timing from the product’s process
- Show the action available after verification
Match the message to the user’s task
Traveller
Explain a supported everyday-payment scenario during a trip. Conditions must match the country and the product’s actual capabilities.
Relocation
Show the specific task the tool can solve in a new country and the steps needed for access. Do not imply an absence of restrictions.
A user holding cryptoassets
Explain the path from existing assets to a supported operation, including actual product conditions and limitations.
How I would test the effect
Record the baseline
Validate events and filtering. Set observation windows, the primary metric and acceptable changes in other metrics.
Assign comparable groups
Randomly assign eligible users to the current and new explanation, retaining a comparable input.
Check the whole journey
Follow KYC starts and completions, timing, errors, first transactions and repeat use. Assess beyond the successful local stage.
Make the decision
Account for sample size and uncertainty. Continue, revise or stop using the agreed criteria.
Metrics
Calculating the effect of a change
- Started KYC completed
- 45 → 60%
135 / 300 and 198 / 330 users
- KYC conversion change
- +15 pp
Not a 15% relative increase
- First transactions
- 81 → 119
1,000 qualified visits per cohort
Compare the starting path and scenario
Without explanation
- 01Qualified visit1,000
100% of this cohort · After bot filtering
- 02Wallet started600
60% of the previous stage · Actual product start
- 03KYC started300
50% of the previous stage · Verification started
- 04KYC completed135
45% of the previous stage · Successful verification
- 05First transaction81
60% of the previous stage · First useful action
- 06Repeat transaction32
39.5% of the previous stage · Within 30 days
Reached the last action: 3.2% of entrants
Explanation before KYC
- 01Qualified visit1,000
100% of this cohort · After bot filtering
- 02Wallet started600
60% of the previous stage · Actual product start
- 03KYC started330
55% of the previous stage · Verification started
- 04KYC completed198
60% of the previous stage · Successful verification
- 05First transaction119
60.1% of the previous stage · First useful action
- 06Repeat transaction48
40.3% of the previous stage · Within 30 days
Reached the last action: 4.8% of entrants
Conclusions
What would count as evidence
The calculated scenario shows more KYC completions and transactions. A causal conclusion requires valid group assignment and sufficient data. If only verification improves, examine the next part of the journey.
The approach at a glance
- Connect advertising sources to product events
- Explain verification before asking for documents
- Check transactions after improving a local conversion
Working on a similar task?
Share the product and the change you need. We can discuss my scope, the available data and a next step.
Discuss the task