Confidence Beyond Go Live

A data solution can pass testing and still lose business confidence after go-live. When figures disagree, teams need a repeatable way to identify the difference, explain its cause and demonstrate that it has been addressed. That is the role of reconciliation.

Data reconciliation compares information across systems or stages of processing. It identifies differences and provides evidence that those differences are understood and addressed. As one client put it, “it’s not right until it’s reconciled”.

System integration testing and user acceptance testing provide evidence that a solution works under the conditions tested. They do not establish that it will continue to work as transactions arrive, interfaces fail, business rules change, and users amend records.

Reconciliation needs to be part of the operating solution.

Agree what should match

Before choosing a tool or writing comparison rules, establish what the reconciliation needs to prove.

Are all completed jobs represented in billing? Do purchase orders retain their value as they move between systems? Are vendor records consistent across enterprise resource planning systems?

The comparison needs an agreed basis: which records belong in scope, how they are matched, which reporting period applies and what differences are acceptable.

Timing matters. Two systems may show different totals because one has processed a transaction and the other has not. Different currencies, hierarchies or accounting periods can also produce legitimate differences.

A useful reconciliation distinguishes expected differences from exceptions requiring investigation.

It also has limits. Two systems can agree while both contain the same incorrect information. Reconciliation contributes to confidence, but it does not replace other data quality and business controls.

Build comparisons that help explain differences

A total can show that something is wrong. Transaction-level detail helps explain where and why.

Where practical, reconciliation should support investigation from a summary difference down to the affected records. Master data matters too: inconsistent vendor, customer or organisational records can undermine comparisons even when the transactions have transferred correctly.

The comparison should also avoid reproducing the same failure as the integration it checks. If both rely on the same faulty extraction or transformation, agreement may provide false confidence.

Independence therefore needs to be considered in the design: what evidence is being compared, where it comes from and which failures the reconciliation can detect.

Make exceptions manageable

A reconciliation that produces thousands of unexplained exceptions can quickly become another operational burden.

Exceptions need to be prioritised by their consequences. Financial exposure, operational impact, age and recurrence can help teams decide where to act first.

Tolerances should be explicit and agreed. They can reduce noise, but they can also conceal a developing problem if nobody reviews them. A small difference repeated across many transactions may deserve more attention than its individual value suggests.

Each exception also needs a clear owner, an expected response and an escalation path. Some will require a business decision; others will require investigation by an integration or application team.

The reconciliation process should make the next action clear.

Correct the cause and preserve the evidence

The reconciliation service should detect differences and support their investigation. Corrections should occur in the appropriate authoritative system or processing step, through controlled processes.

This keeps the distinction clear between identifying a problem and changing the information that caused it.

The evidence should show what changed. If a purchase order initially arrives with a value of $1,200 and later arrives with a corrected value of $1,207, the reconciliation history should retain both observations and show how the exception was resolved.

Rule history matters as well. If a tolerance changes, teams need to know which version applied when an earlier result was produced. Otherwise, historical outcomes become difficult to explain or reproduce.

Access to reconciliation rules and exception closure should reflect their importance. Where reconciliation supports financial or fraud controls, changes to its behaviour deserve appropriate scrutiny.

Design for operational reality

Several issues deserve attention early.

Source records may be deleted. Transactions may arrive late or out of sequence. Systems may use different definitions and hierarchies. Business rules may change without corresponding changes to interfaces or comparisons.

Interface contracts help establish expectations: what will be supplied, when it will arrive, how amendments and deletions are represented, and who handles failures.

Reconciliation also needs a schedule that matches the business need. A month-end comparison may suit one process; another may require daily or more frequent checks.

Reporting should expose recurring causes and trends. Repeated exceptions are evidence that an underlying process, interface or business rule may need attention.

Deliver reconciliation with the solution

Reconciliation is most useful when it is available during delivery.

It can help teams investigate discrepancies during testing, provide evidence for acceptance and establish the process that will operate after go-live. Leaving it until later creates a gap precisely when users are learning whether they can trust the new solution.

Business sponsorship matters because reconciliation requires decisions about acceptable differences, ownership and resolution. Technical teams can build the comparisons, but they cannot make all those decisions on the business’s behalf.

Measure whether the service is helping: how long exceptions remain open, how much effort investigation requires, which problems recur and whether unexplained differences are reducing.

Reconciliation provides continuing evidence that information is moving and being processed as intended. Agree what must reconcile, assign responsibility for exceptions and preserve the evidence of resolution.

Build that capability alongside the solution, and maintain it as the business and its systems change.

Scroll to Top