Group lending works because the people who decide whether you are a good credit risk are the people who will lose money if you are not — and they live on your street. That is the entire mechanism. Everything else is operational detail, and the operational detail is where group lending programmes succeed or quietly fall apart.
Chamas, groups and the thing in the middle
A chama is an informal savings and investment group: members contribute regularly, and the pot is either lent out among members, rotated (the merry-go-round), or invested. It is a social institution first and a financial one second, and it long predates anybody lending to it.
A lending group, in the sense a microfinance institution means, is a structure a lender creates or adopts to make small unsecured loans viable: typically five to thirty members who meet regularly, guarantee each other, and receive loans in a defined order. Many are existing chamas that took on a lending relationship. Some were formed by the loan officer last month.
The difference matters more than it looks. A group that existed before your loan has social capital you did not create and cannot destroy. A group assembled to qualify for a loan has none, and the joint liability is a legal formality rather than a real constraint. If your portfolio-at-risk by group age shows the second kind performing like the first, look harder at your data.
Joint liability, and what it actually means on the day
The promise is that if one member does not pay, the others do. In practice this resolves in one of four ways, and a system needs to be able to record all four:
- The group pays from the group fund. Cleanest. The group holds savings, the shortfall is taken from them, and the defaulting member owes the group rather than you.
- Members contribute individually. Recorded as repayments against the defaulter's loan, with a note of who paid — because the group will settle up internally and will want the record.
- The group refuses. This is the case people design for last and see most. The loan goes to arrears, the group's collective credit standing is affected, and the next cycle's disbursement is at risk.
- The group covers it and then expels the member. Very common, and it needs a membership-history model: the loan was made when they were a member, and it does not stop existing when they leave.
A system that models a group loan as "one loan, one borrower" cannot record any of this. What is needed is a group loan with member-level allocations, so that KES 250,000 lent to a group of ten is ten positions that can each be in a different state.
Meetings and attendance are credit data
In a well-run group programme, attendance is the single best early warning you have. A member who stops attending is a member who is about to stop paying, and they stop attending first — usually by several weeks.
That means attendance has to be recorded as data rather than as a note:
- Meeting date, expected attendees, actual attendees, apologies.
- Collections taken at the meeting, tied to the members who paid.
- Decisions made — a new member admitted, a loan approved by the group, an expulsion.
Then attendance rate becomes a field you can report on next to PAR, and a group whose attendance has dropped from 95% to 60% becomes a visit rather than a surprise. See PAR, OLB and collection rate for how to keep the arrears numbers honest around it.
The guarantor chain
Kenyan group lending often layers individual guarantees inside the group structure: each member's loan is guaranteed by two or three named others, rather than by the group as an undifferentiated whole. This is stronger, because it makes the obligation specific — "you guaranteed Mary" is a different conversation from "the group is jointly liable".
It also creates a data structure that is easy to get wrong. Three rules:
- A guarantee is recorded against a loan, not against a person. Mary guaranteeing John's second loan is not the same fact as Mary guaranteeing his first, and it must be possible for one to be live while the other is closed.
- A guarantor's exposure is a number you can total. If Mary has guaranteed four members for KES 40,000 each and holds KES 25,000 in savings, that is a fact somebody should be able to see before approving the fifth.
- Guarantees survive the guarantor leaving the group. They attach to the loan for its life.
Group savings, and the temptation to net them off
Most group programmes require members to save alongside borrowing — a proportion of the loan held as a security deposit, plus regular contributions. Those savings are a liability of yours: the group's money, held by you.
The temptation is to show a member's position net — "owes 40,000, has saved 12,000, net 28,000" — and it should be resisted in the ledger even where it is shown that way on a statement. The loan is an asset and the savings are a liability; netting them off understates both sides of the balance sheet and makes portfolio metrics wrong. When a default is settled from savings, that is a transaction with two entries, not a smaller number appearing.
Loan cycles and the ladder
The other engine of group lending is the promise of a bigger loan next time. A member who completes a cycle cleanly qualifies for a larger amount; a group that completes cleanly moves up together. This is what makes the joint liability bite, and it is why a group in arrears is a group where nobody can borrow.
Operationally, this needs the system to know a member's cycle number and their history across it — including loans from before they joined this group. A member on their fourth cycle with a perfect record is a different credit from a first-cycle member in the same group at the same amount, and if your product limits are set per group rather than per member you cannot express that.
Running a group programme: what to watch weekly
Related: twelve things that cut PAR 30, several of which apply specifically to group portfolios.