Ajo runs on trust and a card, and both hold right up until they do not. One agent walks a line of shops holding contributors' money, marks the card, and squares up later from memory. When a balance is disputed, there is no third record — only two people who each believe themselves.
Digitising that round is not complicated. Doing it without slowing the agent down, and without accidentally misdescribing what your business is, takes a bit more care. Here is the order that works.
1. Decide what you are before you decide what to build
This comes first because it is the one software cannot answer for you.
A rotating savings and contribution arrangement — ajo, esusu, adashe — is not automatically a loan business. A cooperative society's member service is not the same as lending to the public. These are different activities with different registration, capital and reporting conditions, and the Central Bank of Nigeria's microfinance bank framework covers licensing, governance, individual and group lending, documentation, portfolio at risk and provisioning for the institutions on the lending side of that line.
Which side you sit on decides what your records must show. Settle it with your regulator and your lawyer first. Then the software's job is to make the answer evidenceable: contributions recorded as contributions, loans recorded as loans, each with its own balance and history.
2. Move the entry to the stall
The highest-leverage change is moving the record from the-notebook-at-night to the doorstep. Not because evening entry is careless, but because it is reconstruction — and reconstruction is where amounts drift and contributors get skipped.
An entry made while the contributor is standing there carries the actual amount, the actual time, the agent's identity and the channel the money came through. The contributor keeps a receipt. The office sees it in seconds. Two matching records instead of one contested one.
The blocker is always speed. An agent with a cash bag and a queue behind them will not type into a small form in the sun — they will defer it, and deferred is where entries die. That is why voice entry earns its place on a market round: the agent says the amount, the app matches the contributor and shows it for confirmation, and the round keeps moving. On Nigerian routes that works in English, Nigerian Pidgin, Hausa and Yoruba, including mixed phrases — a sentence in one language carrying a number in another.
3. Keep contributions and repayments apart
Most rounds involve both. The same agent takes a contribution from one shop and a loan repayment from the next, and they land in the same bag.
They are opposite obligations. A contribution is money you owe the contributor. A repayment reduces money the contributor owes you. If the system records a single daily total, nobody can decompose the day afterwards — not you, not your auditor, not the contributor asking for their balance.
Record the meaning at the point of collection and the day's cash separates back into its two halves on its own.
4. Close per agent, the same evening
A shortfall found at month end points at nobody. A shortfall found the same evening points at one person, one round, one day.
So the close needs three properties:
- Per agent, not pooled across the office
- Daily, not weekly
- Against an expected figure derived from the contributor schedules, not just what the agent declared
The point of a daily close is not catching thieves. It is that an honest agent who is short ₦3,000 can explain why on Tuesday evening, and cannot on the 31st.
Pair that with a planned round assigned to a named agent and tracked by GPS with a location history, and "I went, the shop was closed" becomes checkable. The behaviour usually changes before anyone checks.
5. Gate the adjustments, not the collections
Here is where most rollouts go wrong in one direction or the other. Require approval for everything and your agents quietly return to the card. Require approval for nothing and your book is whatever the last person typed.
The split that survives contact with a real round:
| Action | Volume | Risk | Treatment |
|---|---|---|---|
| Contribution or repayment | Dozens a day | Mishearing | Posts on confirmation |
| Balance adjusted by hand | Rare | Judgement or intent | Held for approval |
| Contributor signed up in the field | Rare | No office oversight | Held for approval |
| Loan rescheduled after a bad market day | Rare | Judgement | Held for approval |
Fast where volume lives, slow where consequence lives. That is what a maker-checker approval workflow is actually for — and it is set per role and per resource, so you are not choosing between slowing every agent down and controlling nothing.
6. Assume no network
Markets are where signal dies. An app that stops recording on a dead spot will have a paper workaround within a week, and the workaround becomes the real system.
Offline is the assumed condition of a market round, not a fallback. Entries save on the device and sync when the signal returns, and the agent should not have to think about it.
7. Record why a payment was missed
Where your operation also lends, recovery becomes part of the record — and this is worth being explicit about. Harassment, public shaming, contact-list abuse and threats are not collection techniques. They are the fastest route to a regulatory problem and they do not work twice.
The practical alternative is small: the agent logs why a contributor could not pay, that reason sits next to the missed instalment, and the follow-up decision gets made by someone looking at the full history rather than by whoever shouted loudest at the stall.
None of this is a technology problem. The card is not deficient because it is paper — it is deficient because it is the only copy, written by one party, checkable by nobody. Produce a second matching record at the moment of collection and most ajo disputes stop existing.
Nigeria is available at company setup today, in naira on Lagos business dates. Bring one agent's contributor list and yesterday's card and we will build that round and close it end to end on your own numbers.