Payment Processing KPIs: Success, Cost, and Exceptions

Payment processing KPIs measure completed payments, processing cost, refunds, and settlement timing against clearly defined populations. Review them together by channel: a channel can complete more payments while costing more, or show strong capture results while cash arrives late.

Define success before calculating a rate

Choose the event that counts as success. An authorization outcome, a completed capture, and an eventual paid order are different measures. This example defines a successful attempt as one that completed capture; an authorization without a completed capture does not qualify.

Keep the attempt ID and order ID separate. A retry is another attempt at the same order, not another customer purchase. If you report eventual order success, group the attempts by order and measure whether each eligible order was eventually paid. Do not label that result an attempt success rate.

A KPI also needs an owner and an action. The payments team owns the completion review, finance owns the fee reconciliation, customer operations owns refund investigation, and treasury owns the payout and settlement review. The distinction between a KPI, metric, and measure helps keep routine counts from crowding out those decisions.

Use the right denominator for each question

Payment counts and payment value should remain separate. The Federal Reserve Payments Study reports trends in both the number and value of noncash payments; a count trend and a dollar trend describe different aspects of activity.

Capture success rate = Successful captured attempts / Eligible attempts

Effective fee rate = Included processing fees / Captured payment value

Refund value rate = Attributed refund value / Original captured payment value

An eligible attempt needs a recorded outcome at the reporting cutoff. Keep unresolved attempts visible as pending rather than silently excluding them. For settlement, use the expected date in the company's agreement and the relevant business-day calendar. Count both captures that settled late and captures still unpaid after that deadline at the review cutoff. A capture pending before its deadline is not late.

The Federal Reserve's ACH processing schedule distinguishes transmission deadlines, target distribution times, and settlement timing. Do not treat a submitted payment as settled, and do not apply an ACH schedule to card payments. Use the timing rules for the actual payment channel.

Gather the files and preserve their relationships

Bring an attempt export with attempt ID, order ID, channel, initiation date, outcome, capture ID, captured amount, and currency. Add the fee report, refund register, and settlement report. Refunds need the original capture ID; settlements need the capture or batch identifier used in the payout report.

Keep each source at its own level. An order may have several attempts, a capture may have several refund events, and a payout may contain many captures. Repeating an entire capture amount beside every refund row will overstate captured value when the rows are summed.

Set the reporting cutoff so the selected cohort has had time to reach its expected settlement dates. Compare refunds after a consistent observation window as well. A recently captured cohort has had less time to accumulate refunds than an older cohort, even if customer behavior is identical.

A worked comparison by channel

An illustrative US company compares online and in-person card attempts from one closed period. The attempt outcomes are final at the review cutoff. Refunds are attributed to those original captures, and every capture has settled by the cutoff, although some settled late.

The table uses USD throughout. Fees represent the included processing charges assigned to these captures; refunds are customer funds returned, not fees. The settlement denominator includes all captured attempts because this example has a fully mature cohort.

Percentages are rounded to two decimal places after calculating the underlying ratios. Company rates divide the appropriate totals; they do not average channel percentages.

For a spreadsheet with attempts in column B and captured attempts in column C, the total success rate is:

=SUM(C2:C3)/SUM(B2:B3)

Reconcile the figures to the payout

First check the attempt identity. Online has 600 minus 540 = 60 failed attempts; in person has 400 minus 380 = 20. Successful attempts of 920 plus failed attempts of 80 equal all 1,000 eligible attempts. There are no unresolved outcomes in this illustrative cohort.

Next reconcile the money. Assume all listed fees and refunds were deducted from the associated payouts, with no reserves, disputes, currency conversion, or other adjustments. This assumption must be checked against the actual payout report.

The proof is $184,000 minus $4,760 minus $2,920 = $176,320. Match that amount to the payout records and their bank deposits. If actual reports include withheld reserves or adjustments, add explicit bridge lines rather than forcing the difference into processing fees. Connect the resulting costs to the broader financial KPIs, while preserving the distinction between captured funds, recognized revenue, and cash received.

Read the differences as investigation questions

Online has lower capture success and higher fee and refund rates in this example. That identifies a review population; it does not establish a cause. Compare transaction types, customer populations, and outcome reason codes before attributing the gap to the channel itself.

Settlement needs its own exception list with expected date, actual date, amount, and owner. A late-payment count does not reveal whether the delayed value is material. Inspect the affected captures before deciding which exceptions deserve escalation. When choosing a reporting view, require the checks described in dashboard software, rather than judging the payment review by its appearance.

Where it goes wrong

• Counting orders as attempts. Retries distort the rate when the numerator counts attempts and the denominator counts orders. Use matching identifiers and populations.
• Subtracting refunds from the fee denominator. That produces a different ratio from fees divided by original captured value. Name and apply one definition consistently.
• Comparing refunds at different ages. Allow equivalent observation windows, or label immature cohorts clearly before comparing channels.
• Calling pending funds late. A payment is not late until its agreed settlement deadline has passed. Missing expected dates need investigation, not an assumed zero delay.
• Counting payout deductions twice. A refund listed in the transaction file and payout report is one economic event. Match identifiers before building the cash bridge.

Check your payment exports in Covirage

Bring a sample of your own payment, fee, refund, and settlement exports or a Google Sheet to Covirage. Covirage maps the columns, its tools compute and check the channel comparisons, and the external AI model explains the results with every figure traced to its rows. Create a free Covirage account, then set up your business and bring your file.

See Covirage for financial services for the supporting industry context.