Why payment reconciliation breaks at scale
Manual reconciliation does not get gradually harder as an organization grows. It works fine, then one period it does not, and everyone is surprised.
The mechanism is simple arithmetic, and once you see it the collapse is predictable rather than surprising.
The exception rate stays roughly constant
Some proportion of payments will not identify themselves — a wrong reference, an unregistered number, a third-party payer, a duplicate. Call it five percent. That proportion is a property of your payers and your rails, not of your size, so it does not improve as you grow.
At two hundred payments a month, five percent is ten exceptions: an afternoon. At two thousand, it is a hundred: a full-time job nobody has hired for. At ten thousand, it is five hundred, and there is no version of the afternoon that scales to it.
Exceptions compound rather than clearing
The second mechanism is worse. An unresolved exception does not disappear at the end of the month — it stays in suspense and joins next month's batch.
If the resolution rate falls below the arrival rate even slightly, the backlog grows every period. And a backlog of unidentified payments is not neutral: it makes every downstream figure less reliable, so the collections list starts including people who have paid, and the team's trust in the queue drops precisely when the queue matters most.
Why it looks sudden
The collapse appears abrupt because the person doing the work absorbs the growth silently for a long time. They work later, they batch more, they take shortcuts on the hard ones.
The visible failure — an arrears list nobody trusts, a member chased for paid dues, a treasurer's figure the accountant cannot reproduce — arrives long after the underlying capacity was exceeded.
What actually helps
Two things, in this order. First, reduce the exception rate at source: registered paying numbers, a reference scheme people can actually use, corporate payers modeled properly. Second, make the exceptions that remain into a queue with owners, ages and outcomes, so the backlog is measurable rather than absorbed.
More reminders do not help. Neither does a better dashboard over an unreliable base.