The third column matters more than the second. A ministry that knows what drives its own timeline can compress it; one working from a vendor's Gantt chart cannot.
Data cleansing of the legacy registry
Routinely underestimated by an order of magnitude. Profile before you budget, not after.
Enrolment centre operations
Staffing, rent, power and consumables usually exceed the platform cost over the programme's life. The software is the cheap part, and almost every business case gets this backwards.
The exception queue
People without breeder documents need adjudication, which is human, slow and unavoidable.
Communications
A programme nobody understands generates a queue of avoidable questions at every counter.
Reissue
Credentials expire. The second cycle needs to be in the model from the start, or the programme books a cliff-edge cost into a future administration's budget.
The identifier itself
Length, check digit, whether it encodes anything. Encoding date of birth or region into the number leaks personal data forever and cannot be undone once issued.
Biometric modalities at enrolment
Recapturing a national population is not realistic. Capture more than you currently need.
Storing images alongside templates
Templates from one vendor's algorithm expire with that algorithm. Without raw images, migration means re-enrolling the country.
Card body and chip
A fifteen-year credential commits you to a chip's end-of-life and to its manufacturer's roadmap.
Where the keys live
Root key custody defines sovereignty in practice, whatever the contract says in principle.
Every one of these is cheap to decide correctly in month three and extremely expensive to revisit in year five.
- Enrolment throughput modelled on a demonstration rather than a queue with real people in it
- Deduplication thresholds tuned on a vendor's test set rather than your population, producing false matches concentrated in specific groups
- A pilot in a well-connected district, followed by national rollout into districts that aren't
- No grievance mechanism, so a wrongly rejected citizen has nowhere to go and the story reaches the press
- Coverage declared on enrolment numbers rather than credentials actually in hands
Failure-to-enrol rate, broken down by province, age, gender and occupation. A national average conceals precisely the failure it should reveal — and the concealed group is always one that was already marginal.
A programme reporting 97% coverage with no disaggregation is not reporting success. It is declining to look.