DRAFT
DRAFT — Pending CFO Review

Finance · February 2026
Period: January 2026 · Ghana

Ghana January 2026 BVA

Disbursements & Repayments Deep Dive

Two things broke January: repeat borrower demand collapsed (seasonal), and the credit model tightened (deliberate). The budget assumed December's record application volume would carry into January — it never does. December salary payments created a temporary wave of "satisfied borrowers" who repaid but didn't reborrow, crashing the reborrow rate from 63% to 50%. Separately, gh-bm-v4 dropped the repeat approval rate by 3.4pp. Together: -GHS 57M disbursements, -GHS 50M repayments.

PAR30+ came in favorable (12.5% vs 14.8% budget), but this is partly artificial — 1,789 loans were refinanced on Jan 28 as covenant-protective actions, resetting their DPD to zero.

1. Headline Variances

Metric Budget Actual Gap %
Loans Disbursed 159,194 137,014 -22,180 -13.9%
Disbursements (GHS M) 330.3 273.4 -56.9 -17.2%
  ↳ L0 Customers 21,849 20,898 -951 -4.4%
  ↳ Repeat Loans 137,016 116,116 -20,900 -15.3%
Repayments (GHS M) 327.2 276.9 -50.3 -15.4%
Revenue (GHS M) 41.7 37.9 -3.8 -9.2%
Loanbook EOP (GHS M) 522.6 463.9 -58.7 -11.2%
PAR30+ 14.8% 12.5% -2.3pp Favorable
PAR90+ 6.8% 6.3% -0.5pp Favorable

2. The 5 Whys — Why Disbursements Missed by GHS 57M

The surface answer — "fewer repeat loans" — is circular. Below is the causal chain, each level backed by data.

Disbursements missed by GHS 57M because 20,900 fewer repeat loans were disbursed. Repeat loan count: 116,116 actual vs 137,016 budget (−15.3%). This drives 81% of the GHS gap. The remaining 19% is ticket size compression (GHS 1,995 avg vs GHS 2,076 budget).
↓
Fewer repeat loans because repeat applications crashed 30% AND the approval rate dropped 3.4pp. Dec had 187,491 repeat applications → Jan had only 130,475 (−30.4%). Repeat AR fell from 87.7% to 84.3%. At Dec-level application volume with Dec AR, we'd expect ~164K repeat disbursements — 48K more than actual. ~85% of the miss is demand-driven (fewer applications), ~15% supply-driven (lower AR).
↓
Applications crashed because December→January is a structural seasonal drop — every year. Jan 2025 had 123,517 repeat apps vs Dec 2024's 169,654 = −27.2% drop. Jan 2026 had 130,475 vs Dec 2025's 187,491 = −30.4% drop. Same pattern. The budget implied ~156K repeat applications (137K loans at 87.7% AR) — 20% above the seasonal norm of ~130K.
↓
The seasonal drop happens because December salary payments create "satisfied borrowers" who repay but don't reborrow. Dec 2025 had 140K unique client closures (vs 125K in Nov, +12%). But the reborrow rate crashed: 63.2% of Nov closures reborrowed in Dec, vs only 50.0% of Dec closures reborrowing in Jan (−13.2pp). That's 18,400 fewer reborrowing clients. Borrowers who repaid early with salary money are temporarily flush and don't need a new loan for weeks.
↓
The approval rate also dropped because gh-bm-v4 tightened scoring, plus AR floor overshoot and covenant-driven mass applications. gh-bm-v4 deployed Dec 17 contributed to repeat AR falling from 87.7% (6-month avg) to 84.3% in Jan. The Jan 1-3 AR constraint (38% vs 55%) blocked ~5-8M in approvals. On Jan 26-27, covenant-driven mass re-lending pushed 22K applications at 73-76% AR — low-quality applicants who would normally not apply. Mid-month AR (Jan 4-25) ran at ~86%, closer to historical norms.
Core finding: The budget model fundamentally mispriced the December salary season effect. It assumed December's record application volume would carry into January, ignoring the structural seasonal drop that happens every year. The actual January repeat applications (130K) are almost exactly what seasonal patterns predict.

3. Application Pipeline — Demand Collapsed, Not Just Supply

6-Month Repeat Application Trend

Month Repeat Apps Repeat Disbursed Repeat AR vs Avg
Aug 2025 156,976 138,501 88.2% baseline
Sep 2025 157,280 139,317 88.6% baseline
Oct 2025 174,095 149,766 86.0% +9%
Nov 2025 153,523 134,720 87.8% -4%
Dec 2025 187,491 164,363 87.7% +17%
Jan 2026 130,475 109,929 84.3% −18%

vs Avg = deviation from Aug-Nov baseline average of ~160K. Jan is 18% below baseline, not just below the Dec peak.

4-Year Dec→Jan Seasonal Pattern (Repeat Applications)

Cycle Dec Repeat Apps Jan Repeat Apps Dec→Jan Drop Jan Repeat AR
2022–23 103,924 94,231 −9.3% 84.3%
2023–24 131,344 109,200 −16.9% 81.7%
2024–25 169,654 123,517 −27.2% 86.1%
2025–26 187,491 137,406 −26.7% 84.5%

The Dec→Jan application drop has stabilized at ~27% for the last two years — well before gh-bm-v4 existed. In early years (2022-23), December was not yet a peak month and the drop was smaller. As the portfolio matured and the December salary-season peak grew, the January correction became structural. Note: Jan 2026 repeat AR (84.5%) is essentially identical to Jan 2023 (84.3%) — the model is not uniquely restrictive.

Is gh-bm-v4 the Main Driver? No — The Data Says Otherwise

The risk team attributes the disbursement miss primarily to lower AR from the new BM V4 scorecard. The BM rejection tier data shows this is a contributing factor, not the primary one.

Period B Tier (Rejected) Monthly Δ Total BM Pool B Rate
Jun 20250—191,6570.00%
Sep 2025190+190215,9040.09%
Oct 20252,226+2,036224,7100.99%
Nov 20253,739+1,513232,9711.60%
Dec 20257,124+3,385244,4102.91%
Jan 2026 9,420 +2,296 251,786 3.74%
Feb 202610,929+1,509259,4054.21%

In January, 2,296 clients were newly blocked by the BM model (incremental B-tier growth). Even if every single one would have taken a loan, that's 2,296 out of 20,900 missing repeat loans = 11%.

Three independent data points converge:
1. Application pipeline: 85% of the repeat miss is demand-driven (fewer applications), 15% supply-driven (lower AR).
2. BM rejection tiers: only 2,296 newly blocked clients in Jan = 11% of the repeat loan gap.
3. 4-year seasonal pattern: the same ~27% Dec→Jan drop existed in 2024-25 (pre-V4). Jan 2026 AR (84.5%) matches Jan 2023 AR (84.3%).

December Duration Extension — A Contributing Factor

Fido temporarily extends loan terms in December. This delays when Dec loans mature, pushing some closures (and hence reborrowing) from January into February.

Month Avg Term — Repeat (days) vs Baseline
Mar-Sep 2025 avg46.1baseline
Oct 202546.2+0.1
Nov 202548.1+2.0
Dec 202548.3+2.2
Jan 202648.4+2.3
Feb 202646.7+0.6

Dec loans are 2.2 days longer than baseline. On a 46-day loan, +2 days shifts ~4% of maturities from January into February — roughly 6,500 delayed closures. This is a real but secondary contributor. The same pattern occurred in Dec 2024 (47.6 days vs 44.7 baseline), confirming it is structural, not V4-related.

4. Reborrow Rate & Cycle Dynamics

In Fido's short-duration lending model, repeat loans depend on prior loans maturing. The reborrow rate measures what fraction of borrowers who close a loan in month M re-apply in month M+1.

Flow Unique Client Closures Reborrowed Next Month Reborrow Rate Δ
Nov → Dec 124,624 78,721 63.2% baseline
Dec → Jan 139,739 69,887 50.0% −13.2pp

If the Nov→Dec reborrow rate (63.2%) had held, Dec's 139,739 closures would have yielded 88,315 reborrowing clients — not 69,887. That's 18,428 missing reborrowers, worth approximately GHS 37M in foregone disbursements.

Why did the reborrow rate drop? December salary season (13th month, bonuses) caused many borrowers to repay early — not because they needed to reborrow, but because they were temporarily flush. These "satisfied borrowers" don't need credit for several weeks. The effect dissipates by mid-February as post-holiday cash reserves deplete. This is a natural, recurring cycle — not a sign of portfolio deterioration.

5. Disbursement Decomposition

Volume × Size Decomposition

Component GHS M Impact % of Gap Driver
Volume effect (fewer loans) −46.1 81% 22,180 fewer loans × GHS 2,076 budget avg ticket
Size effect (smaller tickets) −11.1 19% 137,014 loans × GHS 81 ticket gap (1,995 vs 2,076)
Total Disbursement Gap −57.2 100%

Volume Effect Drill-Down

Segment Budget Actual Gap GHS M Impact
Repeat loans (LN > 1) 137,016 116,116 −20,900 ~−43.4
L0 loans (LN = 1) 21,849 20,898 −951 ~−0.2
Total Loan Count Gap 159,194 137,014 −22,180 ~−43.6

Repeat Application Decomposition

The 20,900 repeat loan miss decomposes into demand-side (fewer applications) and supply-side (lower approval rate):

Effect Mechanism Est. Loans Lost % of Repeat Miss
Demand drop (fewer applications) ~26K fewer apps × budget AR ~17,700 ~85%
AR tightening (lower approval rate) 130K apps × 3.4pp AR drop ~4,400 ~15%

Policy Timeline — Credit Actions Dec–Jan

Dec 8 AR cut 50% → 38% ("restore in Jan")
Dec 10 Max loan amount cut ~13-14% (Dec only)
Dec 17 gh-bm-v4 scorecard deployed
Jan 4 AR restored 38% → 55% (3-day delay into Jan)
Jan 14 Max loan amount +4% (reactive: loan count below budget)
Jan 18 Processing fee +4% (revenue recovery attempt)
Jan 26 Max loan amount +5% UCLL 4-11, 22 (covenant-driven)
Jan 28 1,789 loans refinanced + due date postponements (covenant-protective)

6. Repayment Shortfall — Smaller Pool, Not Worse Behavior

Pool × Rate Decomposition

Component Budget Actual Gap
Opening loanbook (Dec EOP) ~497 448.6 −48.4
+ Jan disbursements 330.3 273.4 −56.9
Repayment pool ~827 722.0 −105.3
Effective repayment rate 39.6% 38.4% −1.2pp
Total repayments 327.2 276.9 −50.3

Variance Attribution

Effect GHS M % of Gap Mechanism
Volume effect −41.7 83% Pool smaller by 105M × 39.6% budget rate
Rate effect −8.7 17% Actual pool 722M × 1.2pp rate gap

83% of the repayment shortfall is mechanical — a smaller loanbook and fewer January disbursements produced a smaller repayment pool. Credit quality (rate effect) is a minor contributor.

Why Was the Dec Closing Loanbook 48M Below Budget?

The budget implied a Dec 31 loanbook of ~GHS 497M. Actual was GHS 448.6M. This gap was NOT caused by lower December lending — December actually had record disbursements (181K loans, GHS 373.5M).

The loanbook shrank because December repayments (GHS 412M) massively exceeded disbursements (GHS 374M). Salary season collections were so strong that the loanbook actually contracted in December. The Nov 30 loanbook was GHS 461M → Dec 31 was GHS 449M (−GHS 13M net). The budget likely assumed December lending would grow the loanbook, not shrink it.
Month Repayments (GHS M) Disbursements (GHS M) Net Flow
Nov 2025 300.7 290.2 −10.5
Dec 2025 412.0 373.5 −38.5
Jan 2026 276.9 273.4 −3.5
Jan 28 covenant actions: 1,789 loans were refinanced and due dates were postponed on Jan 28-31 as covenant-protective measures (to push performing loanbook above the 16% covenant threshold). These actions shifted some January repayments into future months. The actual underlying repayment pressure for January is slightly higher than reported — the rate effect (−1.2pp) is partially artificial.

7. Portfolio Quality — Favorable, but Caveated

Metric Budget Actual Δ Assessment
PAR30+ 14.8% 12.5% −2.3pp Favorable
PAR90+ 6.8% 6.3% −0.5pp Favorable
Provision ratio 7.98% 7.04% −0.94pp Favorable
Caveat — artificial PAR compression: The Jan 28 refinancings reset DPD to zero on 1,789 delinquent loans. Without these covenant-protective actions, PAR30+ would be higher. February PAR may reverse as these refinanced loans re-age. Monitor closely.

8. Root Cause Attribution — Ranked by Impact

# Root Cause Mechanism Est. GHS M
1 Budget mispriced seasonal demand drop Budget assumed ~156K repeat apps; seasonal norm is ~130K. The 26K application shortfall accounts for ~85% of the repeat loan miss ~−37
2 Reborrow rate collapse (salary season) 63.2% → 50.0% reborrow rate = 18,428 fewer reborrowing clients. Dec salary payments created temporarily satisfied borrowers ~−37
3 Dec loanbook shrinkage (repayments > disbursements) Dec repayments GHS 412M exceeded disbursements GHS 374M → loanbook shrank → Jan opened with 48M smaller pool ~−19
4 Ticket size compression (Jan 1-13) Max loan amounts held at Dec-era caps until Jan 14 intervention. Avg ticket GHS 1,995 vs GHS 2,076 budget ~−11
5 Dec duration extension (+2.2 days) Longer Dec loan terms push ~4% of maturities from Jan to Feb, delaying ~6,500 closures and their reborrowing cycle ~−13
6 gh-bm-v4 + AR tightening 2,296 newly blocked clients in Jan (BM rejection tiers). Repeat AR fell 3.4pp — but Jan 2026 AR (84.5%) matches Jan 2023 (84.3%) ~−5
7 Repayment rate softness 38.4% effective vs 39.6% budget on 722M pool. Partly distorted by Jan 28 covenant deferrals ~−9
8 Covenant actions (Jan 28-31) 1,789 refinancings + postponements shifted repayments out of January, compressed PAR artificially distortionary

Note: Root causes 1 and 2 overlap significantly — the seasonal demand drop and the reborrow rate collapse are two facets of the same salary-season phenomenon. They are listed separately because they enter the analysis through different data (application pipeline vs closure-to-reborrow tracking).

Implication for budgeting: The topline model should incorporate a December-to-January seasonal deflator of ~28% on repeat applications. Without this adjustment, every January will underperform its budget on the disbursement line. The actual Jan 2026 repeat loan count (116K) is roughly what a seasonally-adjusted budget would predict: 187K Dec apps × 0.70 seasonal factor × 84% AR ≈ 110K loans.

Methodology & Data Sources

Key Definitions

"Chale" — the numbers tell the story. Let the data lead.