GH Waiver Campaign Performance

Mar 17 – Apr 10, 2026 · Source: ml.loan_info_tbl, LOAN_ACCOUNT_HISTORY, COLLECTION_DAILY, INFOBIP_SMS

Overall Summary

Total Loans

7,930

Fully Paid (discount applied)
481
6.1% conversion

Total Collected

GHS 4.68M

all 7,930 loans incl. partial

Original Debt Exposed

GHS 44.5M

Due After Discount

GHS 21.4M

min repayment to qualify

Yield vs. Original Debt

10.5%

collected / total_due_amount

Structural problem: 4,934 loans (62% of campaign) had never paid on the loan before the campaign. They generated only GHS 366K on GHS 26.4M original debt — 1.4% yield. The 830 loans that paid within 30 days before the campaign generated GHS 2.05M — 5.6× more from 17% of the volume.

Performance by DPD

DPD Band	Loans	Fully Paid	Conversion	Collected	Original Debt	Yield
30–59	2,939	372	12.7%	GHS 2,688K	GHS 17,121K	15.7%
60–89	1,721	66	3.8%	GHS 745K	GHS 9,350K	8.0%
90–119	1,903	33	1.7%	GHS 831K	GHS 10,575K	7.9%
120+	1,367	10	0.7%	GHS 421K	GHS 7,434K	5.7%

DPD 120+: GHS 4.55M discount offered, GHS 421K collected. For every GHS 1 collected, ~GHS 10.8 in discount was ignored.

As expected, the conversion rate for DPD 30–59 was the highest. This aligns with expectations, as this group was exposed to the discount for the first time during this campaign.

Performance by Segment (Prior Payment × DPD)

Payment History	DPD	Loans	Paid	Conversion	Collected	Original Debt	Yield
Paid on campaign day	30–59	82	42	51.2%	GHS 257K	GHS 538K	47.7%
Paid 1–7 days before	30–59	304	89	29.3%	GHS 734K	GHS 1,992K	36.9%
Paid 1–7 days before	60–89	59	14	23.7%	GHS 134K	GHS 345K	38.8%
Paid 1–7 days before	90–119	35	4	11.4%	GHS 56K	GHS 187K	29.8%
Paid 8–30 days before	30–59	526	140	26.6%	GHS 1,061K	GHS 3,225K	32.9%
Paid 8–30 days before	60–89	196	15	7.7%	GHS 255K	GHS 1,162K	22.0%
Paid 8–30 days before	90–119	114	7	6.1%	GHS 195K	GHS 709K	27.5%
Paid 31d+ before	30–59	385	30	7.8%	GHS 390K	GHS 2,424K	16.1%
Paid 31d+ before	60–89	329	19	5.8%	GHS 283K	GHS 1,914K	14.8%
Paid 31d+ before	90–119	537	10	1.9%	GHS 532K	GHS 3,111K	17.1%
Never paid	30–59	1,642	71	4.3%	GHS 247K	GHS 8,941K	2.8%
Never paid	60–89	1,133	16	1.4%	GHS 62K	GHS 5,900K	1.0%
Never paid	90–119	1,212	12	1.0%	GHS 39K	GHS 6,549K	0.6%
Never paid	120+	947	2	0.2%	GHS 18K	GHS 4,992K	0.4%

The conversion rate was highest across all DPD groups for those who had paid within the last 1–7 days.

On the other hand, 62% of the campaign (almost 5,000 loans) had never paid before. They generated GHS 366K — we can assume that without the campaign, these loans would not pay.

Additional Signals

By Balance Size (Due After Discount)

Balance Band (GHS)	Loans	Fully Paid	Conversion
<500	33	11	33.3%
500–1K	344	55	16.0%
1K–2K	2,848	206	7.2%
2K–5K	4,067	202	5.0%
5K+	638	7	1.1%

Achievable amount drives action. Loans under GHS 1K converted at 16–33% — 3–6× the campaign average. Loans over GHS 2K (71% of volume) converted at ≤5%. High balance is the primary friction point: borrowers willing but unable to pay lump sum. Strongest case for tiered instalment offers at 2K+.

By LN Group

LN	Loans	Paid	Rate
4–6	316	9	2.8%
7–10	1,994	74	3.7%
11–15	2,381	126	5.3%
16–20	1,325	85	6.4%
21+	1,914	187	9.8%

By Collection Call Response

Segment	Loans	Paid	Rate
Queue answered	2,333	259	11.1%
IVR only	1,146	89	7.8%
Attempted, no answer	4,450	132	3.0%

All IVR-only loans also had agent attempts — no loans received purely IVR.

SMS Response (INFOBIP_SMS — 6 discount batches)

SMS Batch	SMS Sent	Loans Paid Within 1 Day
Mar 18 — "Second chance"	7,916	100
Mar 20 — "Discount active"	7,814	32
Mar 27 — "Discount reminder"	7,694	37
Apr 1	7,044	36
Apr 6 — Easter SMS	6,311	23
Apr 10 — Last day	6,228	57
Total unique (any batch)	—	285 of 481 (59%)

59% of fully paid loans paid within 1 day of an SMS. First batch (Mar 18) drove the most — borrowers waiting for the offer. Last-day urgency (Apr 10) was second. The first SMS is the highest-leverage send.

What Worked & What Didn't

✓ Worked

DPD 30–59 + paid recently: 27–51% conversion — discount caught active repayers and closed loans
LN 21+: 9.8% overall, 18% at DPD 30-59 — loyal borrowers respond to relationship + offer
First SMS batch: 100 loans paid within 1 day — strongest single channel action
Agent calls: queue-answered loans converted at 11.1% — 3.7× unanswered
Small balances (<GHS 1K due): 16–33% conversion — achievable amount drives action; 3–6× campaign average

✗ Didn't Work

Never paid + DPD 60+: 0.4–1.0% yield — discount wasted on 3,292 loans
DPD 120+: 0.7% conversion — GHS 10.8 discount ignored per GHS 1 collected
LN 4–6 + DPD 60+: near-zero — no relationship equity, no response
4,450 unanswered loans: 56% of campaign never got agent contact — massive unrealised upside
Flat 50% discount for high-intent borrowers: recent payers don't need the full 50% to close — recovering GHS 400–500K more per campaign

Strategy for Next Campaign

Tier 1 — Active repayers

Paid ≤7 days before campaign — all DPD bands

20–35%

Recent payment signal dominates DPD — strong conversion across all bands:
DPD 30–59: 29.3% → 20% discount · DPD 60–89: 23.7% → 25% discount · DPD 90–119: 11.4% → 35% discount
Current rule misses 94 high-intent loans (DPD 60–119) who receive 50% unnecessarily.

Tier 2 — Recent payers

Paid 8–30 days before — DPD 30–59 and DPD 60–89 only

30–40%

Recency signal holds through DPD 60–89 but fades at 90+:
DPD 30–59: 26.6% → 30–35% discount · DPD 60–89: 7.7% → 40% discount
LN 21+ with last payment 31d+ ago = 8.5% / 7.0% — not enough to pull out of Tier 3.

Tier 3 — Everyone else

Never paid · paid 8–30d + DPD 90–119 · paid 31d+ · DPD 120+

50%

Stay in campaign with current 50% discount. Covers cold starters (4–8% conversion), deep delinquent (1–2%), and segments where payment recency signal is lost. Low conversion expected — 50% is the floor offer needed to extract any recovery.

Financial impact estimate: Tiers 1+2 cover ~1,120 loans (Tier 1: 398 · Tier 2: 722). Reducing average discount from 50% to ~30% with conservative 15% conversion loss → +GHS 400–500K incremental recovery per campaign. Tier 3 unchanged at 50%. No exclusions — everyone stays in campaign.

Decision logic for next campaign

      -- Tier 1: paid within 7 days (all DPD)

      IF last_pmt ≤ 7d AND dpd 30–59  → 20% discount

      IF last_pmt ≤ 7d AND dpd 60–89  → 25% discount

      IF last_pmt ≤ 7d AND dpd 90–119 → 35% discount

      -- Tier 2: recent payers (DPD ≤89)

      IF last_pmt 8–30d AND dpd 30–59 → 30–35% discount

      IF last_pmt 8–30d AND dpd 60–89 → 40% discount

      -- Tier 3: everyone else

      ELSE → 50% discount (current campaign standard)

A/B Test Setup — Campaign Starting 2026-05-06

Loans in Test (T1+T2)
5,092
2,544 treatment · 2,548 control

Tier 3 (no test)

54,530

all get 50% flat

Incremental GHS if strategy wins
+GHS 464K
vs. flat 50% for all T1+T2

Total campaign GHS (projected)

GHS 2.36M

treatment + control + T3

Hypothesis: Tier 1 and Tier 2 borrowers convert at the same rate whether they receive a tiered discount (20–40%) or the flat 50%. If true, the tiered discount recovers +GHS 232K more from the A/B group alone, and +GHS 464K if rolled out to all T1+T2 loans in future campaigns.

Test Structure

Tier 1 — Treatment

1,255 loans · paid ≤7d ago

20–35%

DPD 30–59: 20% · DPD 60–89: 25% · DPD 90+: 35%
Hypothesis: recent payment means these borrowers don't need 50% to close.

Tier 2 — Treatment

1,289 loans · paid 8–30d ago, DPD <90

30–40%

DPD 30–59: 32.5% · DPD 60–89: 40%
Recency signal still warm but weakening — partial discount should hold conversion.

T1 & T2 — Control

2,548 loans · same populations

50%

Flat 50% — same discount as Tier 3. Establishes the baseline conversion rate for active payers receiving the full offer. Direct comparator for treatment groups.

Loan Counts by A/B Group, Tier & DPD

Tier	A/B Group	Discount	DPD 30–59	DPD 60–89	DPD 90–119	DPD 120+	Total
Tier 1	Treatment	20 / 25 / 35%	727	280	127	121	1,255
Tier 1	Control	50% flat	728	280	128	121	1,257
Tier 2	Treatment	32.5 / 40%	710	579	—	—	1,289
Tier 2	Control	50% flat	711	580	—	—	1,291
Tier 3	No test	50% flat	11,478	11,648	13,812	17,592	54,530
Total			14,354	13,367	14,067	17,834	59,622

50/50 random split within each Tier × DPD cell. Source: LOAN_ACCOUNT_HISTORY PERIOD = 2026-05-06, ACTIVE_IN_ARREARS, loan_discount_amount IS NULL.

Expected Repayment Rate — If Strategy Works

Conversion rates from Mar 2026 actuals. Assumption: tiered discount does not reduce conversion vs flat 50%. Repayment rate = conv% × (1 − discount%) — what % of total debt exposed the campaign recovers.

Group	DPD Band	Loans	Conv Rate	Avg Due (GHS)	Discount	Converters	Repayment Rate
T1 Treatment tiered discount	30–59	727	29.3%	1,901	20%	213	23.4%
	60–89	280	23.7%	2,149	25%	66	17.8%
	90–119	127	11.4%	2,417	35%	14	7.4%
	120+	121	8.0%	2,311	35%	10	5.2%
T1 Treatment subtotal		1,255			avg ~21%	303	18.2%
T1 Control 50% flat	30–59	728	29.3%	1,901	50%	213	14.7%
	60–89	280	23.7%	2,149	50%	66	11.9%
	90–119	128	11.4%	2,417	50%	15	5.7%
	120+	121	8.0%	2,311	50%	10	4.0%
T1 Control subtotal		1,257			50%	304	11.8%
T2 Treatment tiered discount	30–59	710	26.6%	1,734	32.5%	189	18.0%
T2 Treatment tiered discount	60–89	579	7.7%	2,055	40%	45	4.6%
T2 Treatment subtotal		1,289			avg ~36%	234	11.4%
T2 Control 50% flat	30–59	711	26.6%	1,734	50%	189	13.3%
T2 Control 50% flat	60–89	580	7.7%	2,055	50%	45	3.9%
T2 Control subtotal		1,291			50%	234	8.7%
Tier 3 no test · 50%	30–59	11,478	6.0%	1,406	50%	689	3.0%
	60–89	11,648	3.5%	1,567	50%	408	1.8%
	90–119	13,812	1.5%	1,832	50%	207	0.8%
	120+	17,592	0.7%	1,831	50%	123	0.4%
Tier 3 subtotal		54,530			50%	1,427	1.2%
TOTAL CAMPAIGN		59,622				2,502	2.3%

Tier 3 conversion rates: blended from Mar 2026 never-paid + 31d+ segments by DPD. T1/T2 conversion rates from Mar 2026 payment-recency actuals. Tier 1 DPD 120+ rate (8.0%) estimated — no exact data point.

Loans

Half the tier's population — 50/50 random split per Tier × DPD cell. Treatment and control are equal in size by design.

Conv Rate

Observed conversion from Mar 2026 for that exact payment-recency × DPD segment. Same rate applied to treatment and control — the A/B test tells us if treatment diverges.

Avg Due (GHS)

Average TOTAL_DUE per loan at snapshot 2026-05-06 — principal + interest + fees + penalties. Full amount before any discount.

Repayment Rate

conv% × (1 − discount%) — what % of total debt exposed gets recovered. Treatment vs control have same converters; gap comes entirely from the smaller discount. E.g. T1 DPD 30–59: treatment 23.4% vs control 14.7%.

Treatment vs. Control — What We Expect to See

Scenario	T1+T2 Treatment GHS	T1+T2 Control GHS	T1+T2 Gap	Full rollout gain	Verdict
Strategy works Treatment conv ≈ control conv	GHS 745K	GHS 513K	+GHS 232K	+GHS 464K/campaign	Roll out tiered discounts to all T1+T2 in next campaign
Partial win Treatment conv drops ≤15% relative	GHS ~630K	GHS 513K	+GHS 117K	+GHS 234K/campaign	Roll out Tier 1 only; keep Tier 2 at 50% for next test
Strategy fails Treatment conv drops >15% relative	GHS <513K	GHS 513K	negative	no gain	Keep flat 50% — these borrowers need the full offer

T1+T2 Treatment GHS

GHS collected from the treatment half of T1+T2 (2,544 loans). Formula per cell: converters × avg_due × (1 − discount%). Summed across all T1 and T2 DPD bands.

Example (T1 DPD 30–59): 213 converters × GHS 1,901 × (1 − 20%) = GHS 324K

T1+T2 Control GHS

GHS collected from the control half (2,548 loans) — same populations, same conversion rates, but 50% flat discount. Lower GHS per converter because borrower pays only 50% of avg_due.

Example (T1 DPD 30–59): 213 converters × GHS 1,901 × (1 − 50%) = GHS 203K vs treatment GHS 324K

T1+T2 Gap

Treatment GHS minus Control GHS from the A/B group only. This is the incremental GHS earned by giving a smaller discount — assuming conversion held.

GHS 745K − GHS 513K = +GHS 232K from 5,092 test loans

Full Rollout Gain

If the test wins, apply tiered discounts to all T1+T2 loans (not just the test half). Gap doubles because the control group also switches to tiered.

GHS 232K × 2 = +GHS 464K per future campaign

Scenario

Defined by whether treatment conversion rate stays within 15% of control. Works: conv gap ≤15% → tiered discount is free money. Fails: conv gap >15% → borrowers needed the full 50% offer to close.

Worked Example — "Strategy Works" Row

T1 DPD 30–59 treatment: 213 × 1,901 × 0.80 = GHS 324K
T1 DPD 60–89 treatment: 66 × 2,149 × 0.75 = GHS 106K
T1 DPD 90–119: 14 × 2,417 × 0.65 = GHS 22K
T1 DPD 120+: 10 × 2,311 × 0.65 = GHS 15K
T2 DPD 30–59: 189 × 1,734 × 0.675 = GHS 221K
T2 DPD 60–89: 45 × 2,055 × 0.60 = GHS 55K
Total treatment = GHS 743K ≈ 745K
Same cells at 50% flat → GHS 513K
Gap = +GHS 232K

Comparison — New Strategy vs. Old (Flat 50% for All)

Old strategy — flat 50% all T1+T2

GHS 2,132K

T1 GHS 606K + T2 GHS 420K + T3 GHS 1,106K

New strategy — tiered T1+T2 (if wins)
GHS 2,596K
T1 GHS 938K + T2 GHS 552K + T3 GHS 1,106K

Incremental gain per campaign
+GHS 464K
same converters · lower discount · more GHS per loan

Win condition for A/B test: Treatment conversion rate ≥ 85% of control conversion rate within each Tier × DPD cell. If this holds, the tiered strategy earns +GHS 232K more from the test group alone — same number of payers, less discount given away. Primary metric: GHS collected per loan in campaign. Secondary: conversion rate gap between treatment and control.

Database: GHANA_PROD · Campaign cohort filter: loan_discount_start_date::date = '2026-03-17' · Fully paid: loan_discount_amount IS NOT NULL

Methodology notes: Use loan_discount_start_date (not end_date) for cohort — gives 7,930 loans matching LAH exactly. · Prior payment recency must use LAST_REPAYMENT_DATE from LAH at PERIOD = '2026-03-17', not from loan_info_tbl (post-campaign field updated on settlement). · Do not join INFOBIP_SMS + MANUAL_CALLS_PRIORITY_DAILY in same query — triggers MCP false write-detection. Use date-based matching instead.

1 — Overall Summary

SELECT COUNT(DISTINCT loan_id) AS total_loans, COUNT(DISTINCT CASE WHEN loan_discount_amount IS NOT NULL THEN loan_id END) AS fully_paid_loans, ROUND(100.0 * COUNT(DISTINCT CASE WHEN loan_discount_amount IS NOT NULL THEN loan_id END) / COUNT(DISTINCT loan_id), 1) AS conversion_pct, ROUND(SUM(total_due_amount), 0) AS due_before_discount, ROUND(SUM(loan_discount_min_repayment_amount), 0) AS due_after_discount, ROUND(SUM(total_repayment_amount), 0) AS total_collected, ROUND(100.0 * SUM(total_repayment_amount) / NULLIF(SUM(total_due_amount), 0), 1) AS yield_pct FROM GHANA_PROD.ml.loan_info_tbl WHERE loan_discount_start_date::date = '2026-03-17'

2 — Performance by DPD Band

SELECT CASE WHEN lah.DAYS_LATE < 60 THEN '30-59' WHEN lah.DAYS_LATE < 90 THEN '60-89' WHEN lah.DAYS_LATE < 120 THEN '90-119' ELSE '120+' END AS dpd_band, COUNT(*) AS total_loans, COUNT(CASE WHEN lit.loan_discount_amount IS NOT NULL THEN 1 END) AS fully_paid, ROUND(100.0 * COUNT(CASE WHEN lit.loan_discount_amount IS NOT NULL THEN 1 END) / COUNT(*), 1) AS conversion_pct, ROUND(SUM(lit.total_repayment_amount), 0) AS total_collected, ROUND(SUM(lit.total_due_amount), 0) AS original_debt, ROUND(100.0 * SUM(lit.total_repayment_amount) / NULLIF(SUM(lit.total_due_amount), 0), 1) AS yield_pct FROM GHANA_PROD.ml.loan_info_tbl lit LEFT JOIN GHANA_PROD.SEMANTIC_LAYER.LOAN_ACCOUNT_HISTORY lah ON lit.loan_id = lah.LOAN_ID AND lah.PERIOD = '2026-03-17' AND lah.LOAN_DISCOUNT_START_DATE = '2026-03-17' WHERE lit.loan_discount_start_date::date = '2026-03-17' GROUP BY 1 ORDER BY MIN(lah.DAYS_LATE)

3 — Performance by Segment (Prior Payment Recency × DPD)

SELECT CASE WHEN DATEDIFF('day', lah.LAST_REPAYMENT_DATE::DATE, '2026-03-17'::DATE) IS NULL THEN 'D_never_paid' WHEN DATEDIFF('day', lah.LAST_REPAYMENT_DATE::DATE, '2026-03-17'::DATE) <= 0 THEN 'A_campaign_day' WHEN DATEDIFF('day', lah.LAST_REPAYMENT_DATE::DATE, '2026-03-17'::DATE) <= 7 THEN 'B_1-7d_before' WHEN DATEDIFF('day', lah.LAST_REPAYMENT_DATE::DATE, '2026-03-17'::DATE) <= 30 THEN 'C_8-30d_before' ELSE 'E_31d+' END AS pmt_seg, CASE WHEN lah.DAYS_LATE < 60 THEN '30-59' WHEN lah.DAYS_LATE < 90 THEN '60-89' WHEN lah.DAYS_LATE < 120 THEN '90-119' ELSE '120+' END AS dpd_band, COUNT(*) AS total_loans, COUNT(CASE WHEN lit.loan_discount_amount IS NOT NULL THEN 1 END) AS fully_paid, ROUND(100.0 * COUNT(CASE WHEN lit.loan_discount_amount IS NOT NULL THEN 1 END) / COUNT(*), 1) AS conversion_pct, ROUND(SUM(lit.total_repayment_amount), 0) AS total_collected, ROUND(SUM(lit.total_due_amount), 0) AS original_debt, ROUND(SUM(lit.loan_discount_min_repayment_amount), 0) AS due_after_discount, ROUND(100.0 * SUM(lit.total_repayment_amount) / NULLIF(SUM(lit.total_due_amount), 0), 1) AS yield_pct FROM GHANA_PROD.ml.loan_info_tbl lit LEFT JOIN GHANA_PROD.SEMANTIC_LAYER.LOAN_ACCOUNT_HISTORY lah ON lit.loan_id = lah.LOAN_ID AND lah.PERIOD = '2026-03-17' AND lah.LOAN_DISCOUNT_START_DATE = '2026-03-17' WHERE lit.loan_discount_start_date::date = '2026-03-17' GROUP BY 1, 2 ORDER BY 1, 2

4 — Performance by LN Group

SELECT CASE WHEN lit.LN BETWEEN 4 AND 6 THEN 'LN_4-6' WHEN lit.LN BETWEEN 7 AND 10 THEN 'LN_7-10' WHEN lit.LN BETWEEN 11 AND 15 THEN 'LN_11-15' WHEN lit.LN BETWEEN 16 AND 20 THEN 'LN_16-20' ELSE 'LN_21+' END AS ln_band, COUNT(*) AS total_loans, SUM(CASE WHEN lit.loan_discount_amount IS NOT NULL THEN 1 ELSE 0 END) AS fully_paid, ROUND(100.0 * SUM(CASE WHEN lit.loan_discount_amount IS NOT NULL THEN 1 ELSE 0 END) / COUNT(*), 1) AS conversion_pct, ROUND(SUM(lit.total_repayment_amount), 0) AS total_collected, ROUND(SUM(lit.total_due_amount), 0) AS due_before_discount, ROUND(100.0 * SUM(lit.total_repayment_amount) / NULLIF(SUM(lit.total_due_amount), 0), 1) AS yield_pct FROM GHANA_PROD.ml.loan_info_tbl lit WHERE lit.loan_discount_start_date::date = '2026-03-17' GROUP BY 1 ORDER BY MIN(lit.LN)

5 — Performance by Collection Call Response

WITH base AS ( SELECT lit.loan_id, CASE WHEN lit.loan_discount_amount IS NOT NULL THEN 1 ELSE 0 END AS fully_paid, lit.total_repayment_amount AS collected, lit.total_due_amount AS due_before_discount FROM GHANA_PROD.ml.loan_info_tbl lit WHERE lit.loan_discount_start_date::date = '2026-03-17' ), calls AS ( SELECT LOAN_ID, SUM(NUM_QUEUE_ANSWERED) AS queue_ans, SUM(NUM_IVR_ANSWERED) AS ivr_ans, SUM(NUM_QUEUE_ATTEMPTS) AS queue_att, SUM(NUM_IVR_ATTEMPTS) AS ivr_att FROM GHANA_PROD.SEMANTIC_LAYER.COLLECTION_DAILY WHERE "Date" BETWEEN '2026-03-17' AND '2026-04-10' GROUP BY LOAN_ID ) SELECT CASE WHEN COALESCE(c.queue_ans, 0) > 0 THEN 'QUEUE_ANSWERED' WHEN COALESCE(c.ivr_ans, 0) > 0 THEN 'IVR_ONLY' WHEN (COALESCE(c.queue_att, 0) + COALESCE(c.ivr_att, 0)) > 0 THEN 'ATTEMPTED_NO_ANSWER' ELSE 'NO_CONTACT' END AS call_segment, COUNT(*) AS total_loans, SUM(b.fully_paid) AS fully_paid, ROUND(100.0 * SUM(b.fully_paid) / COUNT(*), 1) AS conversion_pct, ROUND(SUM(b.collected), 0) AS total_collected, ROUND(100.0 * SUM(b.collected) / NULLIF(SUM(b.due_before_discount), 0), 1) AS yield_pct FROM base b LEFT JOIN calls c ON b.loan_id = c.LOAN_ID GROUP BY 1 ORDER BY conversion_pct DESC

6 — SMS Batch Dates

SELECT SENT_AT::date AS sms_date, COUNT(*) AS sms_sent FROM GHANA_PROD.DATA.INFOBIP_SMS WHERE SENT_AT::date BETWEEN '2026-03-17' AND '2026-04-10' AND (LOWER(TEXT) LIKE '%discount%' OR LOWER(TEXT) LIKE '%waiver%') GROUP BY 1 ORDER BY 1

7 — SMS Attribution (Paid Within 1 Day of Any Batch)

WITH sms_dates AS ( SELECT SENT_AT::date AS sms_date FROM GHANA_PROD.DATA.INFOBIP_SMS WHERE SENT_AT::date BETWEEN '2026-03-17' AND '2026-04-10' AND (LOWER(TEXT) LIKE '%discount%' OR LOWER(TEXT) LIKE '%waiver%') GROUP BY 1 ), paid_loans AS ( SELECT loan_id, LAST_REPAYMENT_DATE::date AS repaid_dt FROM GHANA_PROD.ml.loan_info_tbl WHERE loan_discount_start_date::date = '2026-03-17' AND loan_discount_amount IS NOT NULL AND LAST_REPAYMENT_DATE IS NOT NULL ) SELECT COUNT(DISTINCT p.loan_id) AS paid_within_1d_any_sms, (SELECT COUNT(*) FROM paid_loans) AS total_fully_paid FROM paid_loans p JOIN sms_dates s ON DATEDIFF('day', s.sms_date, p.repaid_dt) BETWEEN 0 AND 1