Guides

How to reconcile payments across dozens of payment providers

To reconcile payments across many providers, collect three sources for every payment — your own log, the provider’s notifications and the provider’s statement. Translate them into one format, match on several identifiers, compare amount and outcome, and put each payment into one named outcome. Name what cannot be checked instead of hiding it.

What does multi-provider payment reconciliation mean?

Multi-provider payment reconciliation is the daily check that every payment taken through every provider ended the way your records say. The check covers amount, currency and outcome, not only whether an ID exists somewhere. With one provider, a lookup in a spreadsheet often looks good enough. With dozens of providers, each one speaks its own dialect. Fields have different names. Amounts arrive in different units. Times arrive in different zones. Statuses use different words for the same result. Some providers send no statement at all. The method below is vendor-neutral. It works whether you build the process yourself, run it in a spreadsheet for a while, or buy software. The goal is simple to state and hard to reach: every payment ends in one named outcome, with a reason, and nothing disappears silently. The terms used here are defined in the glossary.

What are the three sources of truth, and why do they disagree?

The three sources of truth for a payment are the merchant log, the provider’s notifications and the provider’s statement. The merchant log is your own record of what you asked for and what you believe happened. Notifications are the provider’s real-time messages: callbacks and status responses. The provider statement is the provider’s later, settled account of the day. The three are produced by different systems, at different times, for different readers. A callback can arrive before the provider’s own processing finishes. A statement can be cut at the provider’s midnight, not yours. A notification can repeat the same fact several times. A provider can correct an outcome in its statement and send no new callback. None of this means anyone is lying. It means each source answers a slightly different question. Reconciliation is the work of asking all three the same question and recording where the answers differ.

How do you reconcile payments across many providers, step by step?

The method has eight steps, run once per business day.

  1. Collect. Export the merchant log, all notifications and every statement for the same period. Record which provider sent which file, and which sent nothing.
  2. Translate to one format. Convert amounts to one unit, times to one zone and statuses to one vocabulary. See the next section.
  3. Match on identifiers. Use your ID, the order ID and the provider’s ID. Behind an orchestrator, match on the end provider, not the orchestrator.
  4. Compare amount and outcome. A found ID is not a reconciled payment. Check the amount, the currency and the final result on all three sides.
  5. Classify. Put each payment into one of four outcomes, plus “awaiting” for payments with no final status yet.
  6. Name blind spots. List providers, currencies and hours that no document covers.
  7. Group claims. Collect provable discrepancies by provider, with row-level evidence.
  8. Keep an audit trail. Record which files were used, what was decided and by whom.

How do you translate dozens of formats into one?

Translation into one format is the step most teams underestimate. Three things need normalising before any comparison makes sense. First, amount units: one provider sends 150.00 in major units, another sends 15000 in minor units. The unit cannot be guessed from the number, and a wrong guess is a hundredfold error. Keep a rule per provider and per message type. Second, time zones: some timestamps carry a zone, some are epoch seconds, some have no zone at all. Convert everything to one zone and keep the original. Third, status vocabulary: “approved”, “OK”, status: 1 and “captured” may all mean success, while one word can mean different things in different messages. Map every raw value to a small set of canonical statuses. Keep the mapping as a reviewed table, not as formulas scattered across sheets. When a value is new, stop and classify it instead of guessing.

Why does a payment not match? A taxonomy of reasons

Reasons a payment does not match fall into seven families. Naming them helps a team decide what to do with each line.

ReasonWhat happensWhat to do
Timing or day edgePayment near midnight lands in the next statementWait for the neighbouring day before flagging
UnitsMajor and minor units mixedFix the unit rule; not a money discrepancy
Status vocabularyRaw status mapped to the wrong meaningCorrect the mapping; re-run
Missing IDStatement line carries only the provider’s IDMatch on provider ID or order ID
Incomplete or missing statementStatement covers part of the day, or never arrivesRequest the full statement; do not book a debt
Duplicate notificationsSame fact reported several timesFold repeats into one event
Reversed outcomeDeclined on one side, successful on the otherInvestigate; often a real money issue

The first four are timing or translation issues, not money discrepancies. The last three need a decision.

What do the four outcomes look like on synthetic data?

A worked example makes the four outcomes concrete. The rows below are synthetic and illustrative. Providers, IDs and amounts are invented; all amounts are in CUR-1.

PaymentProviderMerchant logNotificationStatementOutcome
PAY-1A2BProvider Asuccess, 150.00success, 150.00settled, 150.00Matched
PAY-3C4DProvider Asuccess, 150.00success, 15000 minorsettled, 140.00Differs in money
PAY-5E6FProvider B via orchestratordeclineddeclinedsettled, 75.00Differs in facts (reversed outcome)
PAY-7G8HProvider B via orchestratorsuccess, 40.00success, 40.00absent from a complete statementDiffers in facts
PAY-9J0KProvider Csuccess, 60.00success, 60.00no statement sentNothing to reconcile against
PAY-2L3MProvider Apending at 23:58none yetnot yetAwaiting

PAY-3C4D shows why units matter: 15000 minor units equals 150.00, so the notification agrees with the log, and only the statement differs. PAY-5E6F is a reversed outcome: the customer was likely charged but not credited. PAY-9J0K is not a debt. Nobody sent a document to check it against.

What did one measured day look like?

One anonymised merchant day shows the scale of the problem. These figures describe that single day only; they are not an industry benchmark. On that day the merchant exchanged messages with 96 payment providers. The messages came in 1,198 distinct shapes. Payment status arrived as 318 raw values, which reduced to 9 canonical meanings. 54.2% of statement lines did not carry the merchant’s own ID, so they could only be matched through the provider’s ID. Those counts describe the full day. In the reconciliation scope of the same day, 26 providers sent no statement at all. 10.1% of turnover in scope had nothing to reconcile against. The engine also found 320 payments declined in the merchant’s records but successful in the provider’s statement. Your own day will differ. The measured day and the methodology explain how each figure was produced.

How should you handle orchestrators, duplicates and missing statements?

Orchestrators, duplicate notifications and missing statements are the three structural traps of multi-provider reconciliation. A payment orchestrator routes a payment to an end provider, but the statement that settles money comes from that end provider. Reconcile the end provider’s statement against the end provider’s identifiers, and read which end provider handled the payment from the orchestrator’s message. Duplicate notifications are normal. Providers resend callbacks, and merchants poll for status more than once. Fold every repeat of the same fact into one event, keyed on payment and status, before counting anything. On the measured day, 401,255 repeated observations were folded this way. Missing statements need an explicit list. Record which providers sent nothing, which sent a statement for some currencies only, and which sent a file covering part of the day. Then request the missing documents instead of inventing a debt.

How do you turn discrepancies into claims a provider will accept?

A claim a provider will accept rests on evidence the provider can check in its own systems. A list of “missing payments” from a spreadsheet rarely meets that bar. Start by grouping provable discrepancies by provider and by reason, so each claim describes one kind of problem. For every payment in the claim, include the provider’s own ID, the amount and currency, the time in the provider’s zone, and the outcome each source reported. Add the address of the row in each file, so anyone can find the original line. Leave out payments that are only awaiting a statement, and payments covered by an incomplete statement. Those are requests for documents, not claims. A false claim costs time and goodwill; a precise one can be checked line by line. Reconora groups provable discrepancies into claim candidates with row-level evidence. On one anonymised merchant day, that produced 11 claim candidates covering 128 payments.

How do you report statement coverage to finance?

Statement coverage reporting tells finance how much of the day’s turnover could be checked at all. Coverage is a separate figure from the share that matched. A process can report a high match rate while large volumes were never compared with any statement. Report the day in shares of turnover, one share per outcome: matched, differs in money, differs in facts, nothing to reconcile against, and awaiting. Then list the providers and currencies behind the unverifiable share, with the reason for each. Typical reasons are no statement, no statement for that currency, and an incomplete file. Track the list from day to day. A provider that stops sending statements should be visible the next morning, not at month end. Keep the shares in the transaction currency where exact figures matter, and label any converted totals as estimates with the rate date. The exposure calculator shows how unverified shares translate into money for your own turnover.

What belongs on a daily reconciliation checklist?

A daily reconciliation checklist turns the method into a routine that does not depend on one person’s memory.

  1. Confirm that the merchant log for the day is complete and closed.
  2. List which providers sent statements, and check each one covers the full day and every currency.
  3. Run the unit, time zone and status translation; review any new raw values.
  4. Fold duplicate notifications before matching.
  5. Match on every available identifier, including the end provider’s ID.
  6. Compare amount and outcome, not only the presence of an ID.
  7. Move day-edge payments to “awaiting” and review reversed outcomes first.
  8. Group provable discrepancies into claims by provider, with row references.
  9. Publish the share of turnover with nothing to reconcile against, and record every manual decision.

When does a spreadsheet stop working?

A spreadsheet stops working when the question changes from “was the ID found” to “did the money and the outcome match”. Lookups answer the first question well. They struggle with the second. A lookup from the log never sees payments that exist only in a statement. A found ID counts as reconciled even when the amount or outcome differs. An incomplete statement looks like a debt. A payment near midnight looks missing. Each new provider adds another set of formulas and another person who understands them. On one measured day, a spreadsheet-style lookup marked 690 lines as reconciled although amount or outcome differed. The signs are practical: the morning check takes most of a day, and nobody can say what share of turnover was never checked. The same-day comparison shows what changes, and the three-way reconciliation guide explains why a third source matters.

Try it on one day of your data

One day of your data, reconciled, and the report stays with you whether or not you continue.

By Reconora Research · Updated