All case studiesProduction system
4.6 million payments per period, reconciled at a 99.25% match rate.
A betting and gaming operator reconciled about 4.6 million transactions per period across several payment providers, largely by hand. A five-pass reconciliation engine now matches 99.25% automatically, and the finance team no longer touches routine reconciliation.
- Client
- A betting and gaming operator
- Sector
- Gaming, finance operations
- Engagement
- Production system
- Measured result
- 99.25%match rate, 4.6M transactions
99.25%match rate, 4.6M transactions
Growth goal
Add payment providers and volume without reconciliation falling behind the business.
Operating constraint
Each provider delivered its own format, and manual matching was both a bottleneck and a control risk.
Measured result
99.25% of 4.6 million transactions per period matched automatically.
Summary
- Challenge
- Settlement files and internal records were matched in spreadsheets. Knowing how to read each provider's file depended on a few people.
- What SIEL built
- AI maps every provider file into one standard transaction record. Five deterministic passes then match the records, each handling a defined class of match.
- Controls
- Every match links to its source lines and the rule that made it. Each unmatched item is assigned to a named person, and rules change only through change control.
- Impact
- Routine reconciliation runs without the finance team. Exceptions arrive with the passes already tried and the fields that disagree.
The full account
01More providers, more formats
The operator was adding payment methods to reach more players in more markets. Each provider meant another settlement file with its own layout, reference fields and timing. Reconciling around 4.6 million transactions per period by hand lagged the business. Because manual matching left no consistent record of how each item had been treated, it was also a control risk.
02AI for intake, rules for matching
We kept AI away from the matching decision. Its job is intake: it reads each provider file, including irregular layouts and noisy reference fields, and maps it into one standard record. Matching is done by five deterministic passes. Each pass handles a defined class of match, and anything it cannot match falls through to the next. Every rule can be read, tested and explained to an auditor. What remains after the fifth pass is an exception with a named owner.
03What finance has now
The engine matches 99.25% of about 4.6 million transactions per period. The finance team stopped touching routine reconciliation and works the exceptions and the rules instead. Any period can be reviewed from result back to source lines. Because intake and matching are separate, a new payment provider needs a new intake mapping, not new matching rules.
Discuss Your Operation With Our Engineers.
Describe one workflow and the systems it relies on. A senior engineer responds within two business days with an initial assessment: what we would build, what we would not, and why.