Online orders fall 4% in a quarter while site traffic holds steady, and the finance team cannot say where the missing orders went. Shoppers reach the checkout, type in their details and press the button, and somewhere between that click and a confirmed sale a share of them drop out. The analyst assigned to the problem starts by splitting the payment flow into stages that can each be timed and counted.
A card payment passes through 7 stages that matter to an analyst. The customer completes the checkout form, the merchant sends an authorization request, the issuer approves or declines it, the merchant captures the approved amount, the network settles the batch, the processor funds the merchant's bank account, and finance reconciles the deposit against orders. Each stage leaves a timestamp and a status in some system, such as the payment gateway or the accounting ledger.
The first deliverable is usually a funnel built from those records. It shows how many attempts enter each stage and how many leave it in the expected state. A drop of a few points at one stage gives the team a specific target and tells it which system owner to bring into the room.
Many losses happen before a card number reaches the processor. Baymard Institute puts the average documented cart abandonment rate at 70.22%, based on 50 studies. In its survey of shoppers who abandoned during checkout, 17% cited a checkout that was too long or complicated, another 17% pointed to website errors or crashes, 10% said their card was declined and 9% left because there were not enough payment methods.
A 2020 study by Deloitte for Google, covering 37 brands in Europe and the US, found that a 0.1-second improvement in mobile site speed raised retail conversion rates by 8.4%. Analysts check how long the payment form takes to load and how often it returns an error after submission.
Some bottlenecks come from the processing agreement. An analyst who sees deposits arriving 5 days after capture, or volume flattening at the same figure every month, should read the contract before blaming the bank. Delayed funding, monthly volume caps and holds triggered by sudden growth are common terms for merchants in categories such as travel or nutritional supplements. Businesses in those categories often move to high risk payment processors, which underwrite the category at onboarding and set any limits in the original agreement.
Once a request reaches the network, the main measure is the approval rate, approved authorizations divided by attempts. A single blended figure hides most of the useful detail. Analysts split it by issuing bank, card type, country, amount band and new or returning customer, and a sharp dip in one segment usually points to a specific cause.
Decline codes add the reason. Code 05, "do not honor," is the issuer's general refusal and the most common response merchants see. Code 51 means insufficient funds, and code 14 means the card number does not exist, often because of a typing error. A rise in code 14 after a checkout redesign points to a form usability problem. A rise in code 05 from one issuer suggests the issuer's fraud model is reacting to something in the request, such as a missing billing address.
For response time, analysts log the time from request to issuer response at the median and at the 95th percentile, because an average hides the slow tail where timeouts happen, and a timeout at checkout looks to the customer like a failed payment.
Visa has charged fees for excessive reattempts of declined transactions since April 2021, and Mastercard returns merchant advice codes that tell the merchant when to stop. A subscription business that retries every failed renewal on a fixed schedule can pay those fees while recovering very little.
Fraud screening is a frequent source of false declines, because a rejected order rarely produces a complaint. The analyst pulls every order declined by the merchant's own rules, separate from issuer declines, and compares that group with the orders that later turned into fraud chargebacks. If a rule blocks 400 orders a month to stop 12 fraudulent ones, the revenue it costs can be far larger than the fraud it prevents.
Orders held for manual review wait hours or days, and some customers cancel or buy elsewhere in the meantime. The analyst measures how long each order waits and what share is eventually approved. Revenue lost to cancellations during the wait belongs in the same report.
A batch cutoff time decides if a sale made at 9 p.m. settles that night or a day later, and weekends and bank holidays add more days to the funding cycle. Analysts track the gap from capture to cash in the bank by day of the week, because a cutoff set too early can cost a full day of float on a large share of sales.
Account reconciliation is slowed by how processors pay out. Many deposit a net amount, with fees, refunds and chargebacks already deducted, so a single deposit has to be matched against hundreds of orders and several fee lines. The analyst counts unmatched items each month and the staff time spent resolving them, then checks if the processor's reporting can supply transaction-level detail that removes the manual work.
On $20 million in annual card sales, a 1 point rise in the approval rate is about $200,000 in additional revenue. Cutting 2 days from the funding cycle on the same volume frees roughly $110,000 in cash at any given time. Ranking problems this way lets the team fix the stage that pays most first, even when another stage looks more urgent.
Fraud remains a top problem for banks, and issuers keep adjusting the models that approve or decline transactions. New card types enter the market, and a checkout change months later can reintroduce an old error. A weekly report that tracks approval rate by segment next to days to funding catches most of these early. Who owns that report once the project that produced it has closed?