The scale problem no one plans for
Payment reconciliation works at low volume. It stops working, quietly, somewhere between $2M and $10M in monthly payment volume — not because the process is wrong, but because the process was never designed to scale.
Finance teams at Series B fintechs consistently run into the same three failures. They're not exotic edge cases. They're structural.
Failure one: settlement lag mismatches
Payment processors settle on T+1 or T+2. Internal systems record transactions at the time of authorization. When volume is low, the delta is manageable. When 400 transactions clear in a single batch, the lag creates ghost discrepancies that look like missing funds but aren't.
The fix is to reconcile against settlement date, not authorization date — and to automate the daily settlement file pull from the processor so it happens without manual intervention.
Failure two: fee rounding accumulation
Stripe, Braintree, and Adyen each round interchange fees differently. At low volume, rounding errors are immaterial. At $5M monthly volume, a $0.002 rounding difference per transaction compounds into a meaningful variance that shows up in the GL and triggers audit questions.
Track processor fees at the transaction level, not the batch level. The difference is a few columns in your reconciliation model but it eliminates an entire category of unexplained variances.
Failure three: multi-currency ledger drift
If any portion of payment volume runs through non-USD rails, FX rates at booking versus settlement create systematic differences. Most reconciliation processes apply a single monthly FX rate. Processors apply the rate at settlement time. The gap widens as volume grows.
The practical fix: pull the processor's applied FX rate per transaction and apply it consistently in the GL. This requires a data feed, not a manual process.
What this means for the close
These three failures — settlement lag, fee rounding, FX drift — account for the majority of unexplained reconciling items finance teams spend time chasing at month end. They're not mysteries. They're engineering problems with engineering solutions.
Finance teams that fix the data feeds and automate the matching logic get their close time back. The ones that don't keep hiring more people to do more manual matching.
openmemo's reconciliation tools are built around these exact patterns — [openmemo.co](https://openmemo.co) if you're looking for a starting point.