Guides

5 questions to ask any reconciliation vendor.

Ask what the match rate is a share of, what happens to payments with no statement, where your customer data goes, whether you can audit the code that touches it, and how to trace one payment back to its source rows. Good answers are specific and testable. Vague answers usually hide the hard cases.

1. What is your match rate a share of?

A match rate question tests whether a vendor knows its own denominator. A match rate can be a share of payments by count or of turnover. It can include every payment or only those that had something to compare against. It can mean "the ID was found" or "the amount and outcome agreed". Why it matters: each choice moves the figure, sometimes from modest to near perfect, on the same data. A good answer names the denominator, says what was excluded and why, and shows the excluded groups as figures of their own. A red flag is a single percentage with no definition, or a claim of perfect accuracy. How Reconora answers: outcomes are shares of turnover in the reconciliation scope, with counts alongside. On one anonymised merchant day, 75.9% of turnover was matched and confirmed by statements, and 10.1% had nothing to reconcile against. Both figures are shown. The match rate guide explains the arithmetic.

2. What happens to payments with no statement?

A missing-statement question tests how a vendor treats what it cannot verify. Some providers send no statement, some send one for only part of the day or only some currencies. Why it matters: a payment without a statement line cannot be confirmed. If the vendor quietly removes such payments from scope, or books them as the provider's debt, the report looks cleaner than the day was. A good answer says these payments are shown as an outcome of their own, with the reason: no statement, none for the currency, or an incomplete statement. A red flag is "we only reconcile providers that send statements" without a figure for the rest. How Reconora answers: such payments end in "nothing to reconcile against" with a reason. On one anonymised merchant day, 26 providers sent no statement at all. Reconora cannot create a statement that does not exist; it names the gap so the payments team can request one.

3. Where does my customer data go?

A data question tests what leaves your company and who can read it. Payment logs carry names, e-mails, card fragments and IP addresses. Why it matters: every copy of that data outside your perimeter is a risk and a contract to manage. A good answer lists exactly which fields leave, where they are processed, and which sub-processors touch them. Ideally, personal data never leaves at all. A red flag is "we are secure" with no field list, or a requirement to send full exports before you see any result. How Reconora answers: an anonymisation module runs inside your perimeter, offline, and replaces personal data and provider names with codes. Reconora receives only the anonymised artifacts. The encoding table and the secret stay with you, so only you can decode the result. Amounts, timestamps and statuses do leave, because reconciliation needs them. The security page lists what leaves and what never does.

4. Can I audit the code that touches my data?

A code-audit question tests whether a vendor's claims can be checked rather than trusted. Why it matters: "we anonymise" is a promise. Code your security team can read is evidence. A good answer is yes, with the full source of every component that runs on your systems, plus a description of how it is tested. A red flag is a binary agent with network access and no source, or a refusal to describe test results. How Reconora answers: the anonymiser is delivered with its full source code, on request, before it touches your data. It runs offline and uses only the Python standard library. Its test results are published, including the weak spot. Document numbers in free text with no marker word were caught at 26.4%, below the release threshold of 95%. Under known field names they are caught completely. Reconora does not hand out the source of its reconciliation engine; that engine runs on its side and only ever sees anonymised data. Details are on the anonymiser page.

5. How do I check a single payment back to its source rows?

A tracing question tests whether a result is evidence or only a summary. Why it matters: a provider will not accept a claim built on a dashboard total. An auditor will want to see one payment end to end. A good answer shows, for any payment, the log line, the callbacks and the statement line it was matched against, with file and row addresses. A red flag is a report that only offers totals and a match rate. How Reconora answers: every line in the discrepancy queue carries a reason and the row addresses of its sources. Every package and source file carries a SHA-256 fingerprint, so evidence can be traced to the exact file. Claim candidates per provider come with row-level evidence. Every decision is recorded in a tamper-evident audit trail. Reconora does not submit claims to providers; the payments team does that with the evidence.

Using the five questions

The five questions work for any vendor, including a module inside your orchestrator and an in-house script. Ask them in writing and keep the answers. A vendor that answers all five specifically, with figures and their definitions, is easy to compare with another. The quickest test is a pilot on one day of your own data, so the answers are measured rather than described. The Reconora pilot is free, needs no integration and no personal data, and the report stays with you. See also independent reconciliation vs your orchestrator.

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