Scorecard versions
Every weight, bin and threshold is a versioned setting, not code. A credit committee reviews a draft here, sees which decisions it would change, and approves it. Old versions stay available so past decisions can be replayed.
expert-0.2.0
championPhase A expert scorecard. Statement income limited to customer payments and crop sales; statement quality gates its weight.
Approval: Illustrative defaults, not yet approved by a lender's credit committee · Effective: 2026-10-01
expert-0.3.0-draft
draftExample draft for a credit-committee review: heavier betting penalty, a 45% debt-service cap, a bigger cut to the applicant's own income figure and a lower salon margin. Not in use.
Approval: not approved · Effective: —
Settings
Changed settings are highlighted.
| Setting | expert-0.2.0 | expert-0.3.0-draft |
|---|---|---|
| Debt-service cap | 50% | 45% |
| Sector margins | veg 0.2 · grocery 0.15 · clothing 0.3 · salon 0.5 · food 0.35 · hardware 0.2 · transport 0.35 | veg 0.2 · grocery 0.15 · clothing 0.3 · salon 0.4 · food 0.35 · hardware 0.2 · transport 0.35 |
| Income weights | statement 0.35/0.25 · officer 0.45 · applicant 0.2 | statement 0.35/0.25 · officer 0.45 · applicant 0.2 |
| Cut to applicant’s figure | 30% | 40% |
| Business tenure (months) | < 6: 0 · < 12: 10 · ≤ 36: 25 · else 35 | < 6: 0 · < 12: 10 · ≤ 36: 25 · else 35 |
| Premises | mobile 5 · stall 15 · rented 25 · owned 30 | mobile 5 · stall 15 · rented 25 · owned 30 |
| Buyer contract (farm) | yes 30 · no 10 | yes 30 · no 10 |
| Income stability (CV) | > 0.8: 0 · > 0.5: 15 · > 0.3: 30 · else 40 | > 0.8: 0 · > 0.5: 15 · > 0.3: 30 · else 40 |
| Wallet active days | < 5: 0 · ≤ 12: 15 · ≤ 20: 25 · else 35 | < 5: 0 · ≤ 12: 15 · ≤ 20: 25 · else 35 |
| Cash buffer (min balance ÷ income) | < 0.02: 0 · < 0.1: 10 · else 20 | < 0.02: 0 · < 0.1: 10 · else 20 |
| Betting share of spending | > 0.1: -30 · > 0.03: -10 · else 0 | > 0.1: -40 · > 0.03: -15 · else 0 |
| Other digital loans | reversed -20 · 3+ repayments 10 | reversed -20 · 3+ repayments 10 |
| History with lender | none 0 · one 30 · two 50 · past -40 · now -40 | none 0 · one 30 · two 50 · past -40 · now -40 |
| Credit bureau | nofile 0 · clean 20 · adverse -30 · writtenoff -30 | nofile 0 · clean 20 · adverse -30 · writtenoff -30 |
| Guarantor / group | none 0 · guarantor 10 · group 15 | none 0 · guarantor 10 · group 15 |
| Savings group | yes 10 · no 0 | yes 10 · no 0 |
| Officer site checks | 0 to 25 | 0 to 25 |
| Loan size ÷ income | > 0.6: -20 · > 0.3: 0 · else 10 | > 0.6: -20 · > 0.3: 0 · else 10 |
| Score scale | raw -135 to 300 → 300–850 | raw -145 to 300 → 300–850 |
| Bands | A ≥ 700 · B ≥ 640 · C ≥ 580 · D ≥ 520 | A ≥ 700 · B ≥ 640 · C ≥ 580 · D ≥ 520 |
| Lending by band | A 1 · B 0.9 · C 0.75 · D 0.5 · E 0 | A 1 · B 0.9 · C 0.75 · D 0.5 · E 0 |
| Product maximum | trader US$2,000 · farm US$1,500 | trader US$2,000 · farm US$1,500 |
| Refer above betting share | 20% | 20% |
Impact on decisions
The four sample borrowers and every decision in this browser’s log, re-scored on both versions. 1 of 4 would change.
| Applicant | expert-0.2.0 | expert-0.3.0-draft |
|---|---|---|
| Chiedza Mutasa (sample) | Approve 734 · US$300 | Approve 736 · US$300 |
| Tatenda Chikomo (sample) | Approve 691 · US$1,090 | Approve 688 · US$720 |
| Tendai Moyo (sample) | Approve 692 · US$600 | Approve 696 · US$600 |
| Farai Ncube (sample) | Decline 440 · — | Decline 437 · — |
To measure a draft properly, back-test it on the lender’s loan book on the Model validation page; this table only shows which known cases move.