A 50/50 randomised A/B test to evaluate whether framing SMS collection messages around loss aversion (sanctions, credit bureau reporting, reputation damage) drives higher repayment than generic messaging referencing the Lenders & Borrowers Act.
Users split by Client ID into two groups. Messages sent at DPD 10, DPD 14, and DPD 21.
% of delivered SMS recipients who made at least one repayment within 24 hours of receiving the message.
Calculated as: Paid Loans (1d) / SMS Delivered
| CAMPAIGN | SMS SENT | SMS DELIVERED | DELIVERY RATE | DATE RANGE | BALANCE |
|---|---|---|---|---|---|
| Group A — Step 1 (DPD 10) | 10,104 | 7,763 | 76.8% | Mar 5 – Mar 20 | Balanced |
| Group B — Step 1 (DPD 10) | 10,183 | 7,714 | 75.8% | Mar 5 – Mar 20 | Balanced |
| Group A — Step 2 (DPD 14) | 3,410 | 2,624 | 76.9% | Mar 10 – Mar 24 | — |
| Group B — Step 2 (DPD 14) | 1,002 | 785 | 78.3% | Mar 10 – Mar 24 | 3.4x fewer |
| Group A — Step 3 (DPD 21) | 3,236 | 2,485 | 76.8% | Mar 16 – Mar 31 | — |
| Group B — Step 3 (DPD 21) | 928 | 723 | 77.9% | Mar 16 – Mar 31 | 3.5x fewer |
If the collection system only sends Steps 2/3 to users who haven't repaid, then the lower Group B volumes at Steps 2/3 could indicate more Group B users repaid after Step 1 and were therefore excluded from subsequent steps. This would be a positive signal for loss aversion messaging that is hidden by the raw aggregate comparison. However, this needs to be confirmed by checking the actual selection logic.
| SMS Sent | 16,750 |
| SMS Delivered | 12,872 (76.8%) |
| Paid 1d | 439 (3.41%) |
| Paid 2d | 441 (3.43%) |
| Paid 7d | 442 (3.43%) |
| Total Repaid (1d) | GHS 159,780 |
| Avg per Payer (1d) | GHS 364 |
| SMS Sent | 12,113 was 16,851 |
| SMS Delivered | 9,222 was 12,823 |
| Paid 1d | 293 (3.18%) was 396 |
| Paid 2d | 297 (3.22%) |
| Paid 7d | 298 (3.23%) |
| Total Repaid (1d) | GHS 146,101 was 118,404 |
| Avg per Payer (1d) | GHS 499 was 299 |
| METRIC | GROUP A | GROUP B | DELTA |
|---|---|---|---|
| Delivered | 7,763 | 7,714 | -0.6% |
| Paid 1d | 284 (3.66%) | 250 (3.24%) | A +0.42pp |
| Paid 2d | 286 (3.68%) | 254 (3.29%) | A +0.39pp |
| Paid 7d | 287 (3.70%) | 255 (3.31%) | A +0.39pp |
| Total Repaid (1d) | GHS 116,242 | GHS 124,476 | B +7.1% |
| Avg per Payer (1d) | GHS 409 | GHS 498 | B +21.7% |
While Group A has a higher repayment count (+34 more payers), Group B payers repay GHS 89 more on average (GHS 498 vs 409). Group B generated GHS 8,234 more total repayment despite fewer payers. The loss aversion framing may not increase the number of repayments but increases the amount repaid per transaction.
| DPD STEP | GROUP | DELIVERED | PAID 1D | RATE 1D | PAID 2D | RATE 2D | PAID 7D | RATE 7D | AVG/PAYER (1D) |
|---|---|---|---|---|---|---|---|---|---|
| DPD 10 | Group A | 7,763 | 284 | 3.66% | 286 | 3.68% | 287 | 3.70% | GHS 409 |
| DPD 10 | Group B | 7,714 | 250 | 3.24% | 254 | 3.29% | 255 | 3.31% | GHS 498 |
| DPD 14 | Group A | 2,624 | 70 | 2.67% | 70 | 2.67% | 70 | 2.67% | GHS 263 |
| DPD 14 | Group B | 785 | 18 | 2.29% | 18 | 2.29% | 18 | 2.29% | GHS 319 |
| DPD 21 | Group A | 2,485 | 85 | 3.42% | 85 | 3.42% | 85 | 3.42% | GHS 295 |
| DPD 21 | Group B | 723 | 25 | 3.46% | 25 | 3.46% | 25 | 3.46% | GHS 636 |
Moving from 1-day to 2-day or 7-day windows adds almost no additional payers (A: 439 → 441 → 442; B: 293 → 297 → 298). Virtually all repayments triggered by SMS happen within 24 hours.
Average payment per payer is consistently higher for Group B: DPD 10 (+22%), DPD 14 (+21%), DPD 21 (+115%). Loss aversion messaging appears to drive larger repayments among those who act.
| Group A rate | 3.41% (439 / 12,872) |
| Group B rate | 3.18% (293 / 9,222) |
| Absolute diff | -0.23 pp (B worse) |
| Pooled proportion | 3.31% |
| Standard error | 0.00244 |
| Z-score | 0.94 |
| P-value (two-tailed) | 0.35 |
| Result | NOT SIGNIFICANT |
| Group A rate | 3.66% (284 / 7,763) |
| Group B rate | 3.24% (250 / 7,714) |
| Absolute diff | -0.42 pp (B worse) |
| Pooled proportion | 3.45% |
| Standard error | 0.00293 |
| Z-score | 1.43 |
| P-value (two-tailed) | 0.15 |
| Result | NOT SIGNIFICANT |
| WINDOW | GROUP A PAID | GROUP A RATE | GROUP B PAID | GROUP B RATE | DELTA |
|---|---|---|---|---|---|
| 1 Day | 439 | 3.41% | 293 | 3.18% | A +0.23 pp |
| 2 Days | 441 | 3.43% | 297 | 3.22% | A +0.21 pp |
| 7 Days | 442 | 3.43% | 298 | 3.23% | A +0.20 pp |
| WINDOW | GROUP A TOTAL (GHS) | GROUP A AVG/PAYER | GROUP B TOTAL (GHS) | GROUP B AVG/PAYER |
|---|---|---|---|---|
| 1 Day | 159,780 | GHS 364 | 146,101 | GHS 499 |
| 2 Days | 159,807 | GHS 362 | 146,454 | GHS 493 |
| 7 Days | 159,907 | GHS 362 | 149,554 | GHS 502 |
Group B payers consistently pay GHS 135 more per transaction across all windows. Despite 33% fewer payers in Group B (due to volume imbalance), total repayment is only 8.6% less. This suggests loss aversion messaging motivates larger individual repayments.
| LN SEGMENT | GROUP | DELIVERED | PAID 1D | RATE 1D | PAID 7D | RATE 7D | TOTAL (1D GHS) | AVG/PAYER (1D) |
|---|---|---|---|---|---|---|---|---|
| LN 0-3 New Borrowers |
A | 4,891 | 107 | 2.19% | 110 | 2.25% | 16,993 | GHS 159 |
| B | 3,425 | 74 | 2.16% | 77 | 2.25% | 12,493 | GHS 169 | |
| LN 4-9 Mid-Experience |
A | 3,041 | 98 | 3.22% | 98 | 3.22% | 32,067 | GHS 327 |
| B | 1,815 | 43 | 2.37% | 43 | 2.37% | 12,026 | GHS 280 | |
| LN 10+ Repeat Borrowers |
A | 4,940 | 234 | 4.74% | 234 | 4.74% | 110,721 | GHS 473 |
| B | 3,982 | 176 | 4.42% | 178 | 4.47% | 121,581 | GHS 691 | |
Repayment rates are nearly identical (2.19% vs 2.16%, diff = 0.03pp). New borrowers respond similarly to both message types. Group B slightly higher avg payment (GHS 169 vs 159).
Mid-experience borrowers show the biggest rate gap: 3.22% vs 2.37%. However, note Group B had only 1,815 delivered (vs 3,041 for A) — the volume imbalance makes this comparison less reliable.
Rate gap is modest (4.74% vs 4.42%). But Group B repeat borrowers pay GHS 691 vs 473 avg — 46% more per payer. Total repaid: B = GHS 121,581 vs A = GHS 110,721 despite fewer payers.
The Loss Aversion messaging (Group B) did not outperform the generic Control (Group A) on repayment rate. Group A had a higher 1-day repayment rate at DPD 10 (3.66% vs 3.24%) and DPD 14 (2.67% vs 2.29%). Group B led marginally at DPD 21 (3.46% vs 3.42%). The Step 1 p-value of 0.15 (z = 1.43) falls short of the 95% confidence threshold.
Caveat: The severe volume imbalance at Steps 2 and 3 (Group B received 70-71% fewer SMS) makes the aggregate comparison unreliable. This imbalance needs root-cause investigation — it could indicate that more Group B users repaid after Step 1 and exited the collection sequence.
No statistically significant lift from sanctions framing on repayment rate. Continue current generic messaging. However, explore whether amount-based KPIs (total GHS recovered) tell a different story.
Group B received 3.4x fewer SMS at Steps 2 & 3. If this is because more Group B users repaid after Step 1 (and thus exited the sequence), it reverses the narrative entirely. Check the collection system's selection logic.
Group B payers pay 37% more on average (GHS 499 vs 364). For LN 10+ it's 46% more. Consider re-framing the KPI to include total GHS recovered, not just payer count. A message that drives fewer but larger payments may still be optimal.