Why reconciliation is still the hardest hour of the week

Payments in Africa went digital years ago. Reconciliation didn't. A customer sends mobile money in three seconds; figuring out which invoice that payment belongs to can still take a finance officer the better part of a morning.

That gap — instant money, slow understanding — is the quiet tax on almost every small business we talk to. It rarely shows up as a line item, so it rarely gets fixed. But it is real, and it is expensive.

Where the hour actually goes

Reconciliation feels like one task. It is really four, and each one leaks time:

• Collecting records — MoMo SMS, bank statements, cash notes, card settlements — from four or five different places into one view.
• Matching each payment to the invoice or customer it belongs to, usually by squinting at names and amounts.
• Chasing the ones that don't match — a payment with no reference, a part-payment, a transfer from a relative's number.
• Updating the books so the next person who looks sees the truth.

None of these is hard on its own. Together, repeated across a few hundred transactions a month, they become the hour nobody wants — the one that gets pushed to Friday afternoon and then to Monday.

Why spreadsheets don't close the gap

A spreadsheet is a brilliant place to record what you already know. It is a poor place to figure out what you don't. It can't tell you that "Ama M." on the MoMo log and "Mensah, Ama" on the invoice are the same person. It can't flag that a GHS 200 payment against a GHS 500 invoice is a part-payment, not a mistake. Every judgement call still lands on a human.

The money moved instantly. Understanding it still takes days. Closing that gap is the entire job.

What good automation looks like

Good reconciliation automation is not a black box that declares everything matched. It is a system that does the obvious work instantly and shows its reasoning on everything else.

• Exact matches — same amount, clear reference — clear themselves the moment a payment lands. No human touch needed.
• Likely matches are surfaced with a confidence score and the evidence behind it: the name similarity, the amount, the timing.
• Genuinely ambiguous payments go into a short review queue, not a 300-row spreadsheet — so the human spends their hour only on the calls that actually need a human.
• Nothing is hidden and nothing is silently edited. The history stays intact.

The goal isn't to remove the finance officer. It is to make sure their judgement is spent on the ten payments that need it, not buried under the two hundred that don't.

Reconciliation will probably always need a person in the loop. It should not need a person doing data entry. That is the line Rekava is built to draw.