What didn't work · demonstration

Stopping ineffective programmes is responsible governance

This shows how the register of discontinued programmes would be kept: original hypothesis, cost, actual result, why it failed, evidence, lessons and decision. The entries below are simulated examples — none of these programmes has run.

STOPPEDY1 Q2 (sample)Health

◌ Simulated

SMS reminder for PHC appointments

Original hypothesis: Automated SMS reminders would reduce missed appointments and waiting-time variance.
Expected outcome: Missed appointments down 20% in 25 PHCs.
Intervention: Bulk SMS 24h before scheduled visit in local language.
Cost: ₹0.6 cr
Actual result: Missed appointments fell 3% (not statistically significant). Waiting-time variance unchanged.
Why it failed: Most PHC visits are walk-in; scheduled appointments were a small minority. The intervention targeted a problem that barely existed at the facility level.
Evidence: Facility logs from 25 PHCs, 6 months, compared with 25 control PHCs. Primary government data
Labs evaluation memo, peer reviewed. Expert analysis
Lessons: Validate that the problem exists at the point of intervention before building the fix.
Cheap pilots are valuable precisely because they can fail cheaply.

REDESIGNEDY1 Q2 (sample)Education

◌ Simulated

Tablet-based learning app, Grade 3 (single-district pilot)

Original hypothesis: Adaptive learning software would raise reading scores faster than teacher-led remediation.
Expected outcome: 0.2 standard deviation gain in reading in one academic year.
Intervention: Tablets and adaptive app in 40 schools; 40 comparison schools received teacher-led remediation.
Cost: ₹2.1 cr
Actual result: No significant difference versus teacher-led remediation; device downtime averaged 30% of school days.
Why it failed: Device maintenance and connectivity were underestimated; teachers used the app as a substitute rather than a supplement.
Evidence: Independent assessment of 3,200 children at baseline and endline. Independent study
Lessons: Technology pilots must budget for maintenance as a first-class cost.
Compare against the best cheap alternative, not against doing nothing.

REPLACEDY1 Q3 (sample)Climate · Energy · Water

◌ Simulated

Recharge-structure count as a groundwater KPI

Original hypothesis: Counting recharge structures built would track groundwater improvement.
Expected outcome: Structures built would correlate with water-table recovery.
Intervention: KPI adopted for first two quarters of PS-014.
Cost: ₹0 cr
Actual result: Structure counts rose; water-table readings showed no relationship because most structures were sited without hydrogeological assessment.
Why it failed: Output measured instead of outcome. The KPI incentivised construction, not recharge.
Evidence: Piezometer readings vs. structure inventory across 8 blocks. Primary government data
Lessons: Never adopt an output KPI as a proxy for an outcome without evidence of correlation.
Site selection needs hydrogeology, not just budget.