The hardest reconciliations aren't bank-to-book. They're the ones that cross system boundaries — bank, ERP, billing platform, payment processor, and a spreadsheet someone keeps "just in case." When the same transaction lives in four places with four different keys, matching is a data problem before it's an accounting problem. Here's how I structure multi-source reconciliation with matchdata.ai.
The multi-source problem
Consider SaaS revenue. A single customer's monthly charge touches:
- Stripe / billing system — the charge and refund
- Bank — the net settlement deposit (often batched with other customers)
- GL / ERP — deferred revenue recognition and the cash entry
- Revenue schedule — the amortization across the service period
Each source records amount, date, and reference differently. Batched settlements mean one bank line equals many Stripe charges. Revenue recognition means one Stripe charge equals many GL entries across periods.
Step 1: Normalize before you match
The first thing matchdata.ai does is normalize the inputs: consistent date formats, currency handling, and a canonical amount field. More importantly, it lets me tag each row with its source so I can reason about directionality — settlement flows bank←processor, recognition flows GL←charge.
Step 2: Chain the matches
I don't try to match all four sources in one pass. I chain them:
- Processor → Bank: match Stripe charges/refunds to bank settlements, tolerating batching and settlement lag.
- Processor → GL cash entry: match the same charges to the posted cash side, confirming nothing was missed in posting.
- GL cash → Revenue schedule: tie the cash side to the recognized revenue, surfacing timing differences and unamortized balances.
Each link is a matchdata.ai job with its own rules and tolerances. The output of one feeds the next.
Step 3: Trace, don't just match
The payoff isn't speed alone — it's traceability. When a board member asks "why does recognized revenue lag cash by $42k this month," I can trace a specific bank settlement → its constituent charges → their revenue schedules in seconds. That lineage is what auditors and investors actually want.
When to build this
If your close takes more than a week, or your auditors send the same reconciliation request every quarter, the problem is structural, not effort. Multi-source matching with matchdata.ai is the structural fix.