Choosing local payment methods. 6 chapters and a final quiz.

Chapter 1. Reading a market share without getting the figure wrong.

A market share is the result of a measurement taken on a given population, in a given unit, on a given channel, and at a given date, by an author whose business shapes what it counts. The same country shows very different results depending on whether you count transactions or add up the amounts collected. Neither count is wrong. But a decision built on the wrong figure costs a whole integration, its testing, support training, and then its removal two years later. Four questions establish these attributes before you reuse any percentage.

Four questions to ask before copying a percentage

• Which population? The whole country, internet users, online shoppers, or the merchant’s own customers. A method used by half of adults may be absent from a given site’s customer base.
• Which unit? Number of transactions or value collected. A micropayment rail dominates by volume but weighs little by value. The reverse is true of business credit transfers.
• Which channel? In store, online, person-to-person, recurring billing. The ranking changes from one channel to another in almost every country.
• Which year, and published by whom? A 2023 figure describes a market that may have shifted since. With no year and no named publisher, the figure does not get copied.

The classification trap, illustrated

A classification trap arises when a single statistical label covers products with different economics. The “wallets” category is a case in point, since it groups three distinct products. The first hold a balance and collect funds on their own account, which makes them institutions you contract with directly. The second only present a tokenized card, and the rail that settles is still the card, with its interchange, network rules, and chargeback rights. The third are merely an interface to a public account-to-account rail, whose pricing is set by a central bank. Cost, risk, and integration work differ in all three cases. Standard practice is to read the source’s definition before reusing its percentage.

Chapter 2. The full cost of a method.

The full cost of a payment method adds up six items. The rate the provider quotes is only the first. The other five appear in no sales proposal, because none of them goes through the provider’s invoice. They are paid for in tied-up cash, accounting hours, and support tickets. A mix decision based on the fee alone routinely ranks methods in the wrong order, and the mistake only comes to light after a full fiscal year. The six items are listed below in the order in which they arise.

Gross collected ............................  100.00

  - percentage fee .........................   -x.xx   % of amount
  - fixed fee spread over avg order ........   -x.xx   fixed fee / order value
  - FX, sale -> settlement .................   -x.xx   0 if settled in sale currency
  - FX, settlement -> treasury .............   -x.xx   0 if account held in that currency
= net collected ............................   xx.xx

Costs outside collection, spread over the same volume:
  - reconciliation .........................   -x.xx   hours / month * hourly cost
  - support and disputes ...................   -x.xx   contacts * cost per contact
  - amortized integration ..................   -x.xx   project cost / volume over N months
= full cost per 100 collected ..............   xx.xx

Lost opportunity, outside the calculation above:
  failures = (attempts - successes) * avg order value * unit margin

Order value decides between a fixed fee and a percentage fee

The break-even point is the order value at which two pricing models built on different bases cost the same. It is used to compare a method charged as a percentage of the amount with a method charged a fixed sum per transaction, the same whatever the order value. It lies at an order value equal to the fixed fee divided by the percentage rate. Below that order value, the fixed fee costs more. Above it, the percentage fee becomes the heavier one, and the gap then widens linearly with the amount sold. A site that sells both top-ups worth a few units and furniture worth several hundred therefore sits on both sides of the point, which rules out a single answer for the whole site.

Chapter 3. The effect on conversion: captured, displaced, lost.

A payment method’s contribution is measured by the number of orders that would not have happened without it, not by the volume it processes. That volume almost always rises after launch, because some shoppers who were already being served switch to the new method. The matching counter is the first figure shown to the committee that funded the integration, yet it lumps together orders that were created and orders that merely moved. Three effects overlap in the observed figure. They have to be separated before drawing any conclusion about the method’s contribution.

• Captured demand: shoppers who used to leave because they had no way to pay. This is the only effect that creates revenue.
• Displaced demand: shoppers who would have paid another way. Volume migrates from one method to the other, and only the cost difference matters.
• The method’s own success rate: the share of attempts that end with funds confirmed. A method that is often chosen but rarely completed destroys conversion.

A market’s payment mix shifts from year to year, and a measurement that is a year old describes an earlier state of that market. Bizum recorded 105.6 million online purchases in Spain in 2025, up 82.1% year over year. Its annual report, published in January 2026, counts 1,237 million transactions across all uses. Growth of that magnitude can move a method from third to first place in a payment list within two fiscal years. Standard practice is to review the list’s make-up at least once a year, starting from the merchant’s own checkout logs before any external study.

Chapter 4. The three-method rule.

The three-method rule is a convention for building a payment list. It organizes the list around three distinct roles and requires any additional method to prove its value with figures. It responds to the running cost of every method you turn on: an integration, a test cycle, a reconciliation line, a refund flow, and support training. Those costs recur with every change to the rail. A large provider’s catalog runs to more than 100 lines. Few merchants properly run more than four methods per country. Building by role keeps that burden down, since each method covers a need the other two leave unmet.

Three roles, three methods

Chapter 5. Testing before rolling out.

A bounded activation trial is a production release limited to a share of traffic. Its scope, duration, indicators, and decision thresholds are written down before the first line of code is written. It is how you make the case for turning a method on in a market, in place of a slide deck. With no threshold set in advance, a team that has spent six weeks integrating a rail will read ambiguous results in favor of the effort already made. This bias is the costliest in the whole process. The trial sheet sets the thresholds and the default decision before launch, which leaves no room for interpretation when the results come in.

market: <ISO 3166-1 alpha-2 country code>
method: <official name of the system, as used in its own country>
target_role: mainstream | fallback | segment
scope:
  traffic_share: <e.g., 50% of the country's sessions, stable assignment per visitor>
  control: <the other sessions, payment list unchanged>
  display_position: <fixed and identical for the whole trial>
duration:
  minimum: <covers a full cycle: sale, payout, refund, close>
  end: <a written date, not "when we have enough data">
indicators:
  - selection rate
  - confirmation rate
  - median time to confirmation
  - full cost per 100 collected
  - incremental orders vs. control
decision_thresholds:
  roll_out_if: <quantified condition on incremental orders AND cost>
  extend_if: <volume too low to decide, once only>
  remove_if: <quantified condition, including support cost>
default_decision: remove

Four pitfalls that invalidate a trial

• The novelty effect. An unfamiliar logo draws clicks in the first few weeks, then the selection rate falls back. A trial that is too short measures curiosity.
• Seasonality: sales, local holidays, and payday. In markets where wages are paid monthly, the payment mix at the end of the month looks nothing like the mix mid-month.
• Display position. Moving a method to the top is enough to double its selection rate. If the position changes during the trial, the result no longer measures the method.
• No control group. Without sessions that stay on the original list, nothing separates captured demand from displaced demand. The control group is not a refinement; it is the measurement itself.

Chapter 6. Removing a method without breaking the customer journey.

Removing a payment method takes it out of the purchase flow without extinguishing the obligations created by sales already collected. On the surface, the job comes down to unchecking a box in the provider’s dashboard. But a method that has been live for two years has left dependencies in recurring mandates, unused store credit, open disputes, and the reconciliation trail. A removal carried out without an inventory of those dependencies interrupts refunds for six months. The inventory therefore comes before setting a date. Removals that have actually happened in the market provide the typical timeline.

The removal sequence, step by step

• Freeze new entries: remove the method from new recurring payments and new accounts before any announcement.
• Inventory the dependencies: active mandates and subscriptions, refunds in progress, open disputes, unused store credit, and orders awaiting an asynchronous payment.
• Migrate recurring payments, the longest step, because it requires action from the customer. It takes several reminders and a fallback method that is already live.
• Publish the date to support, to affected customers, and in the integration documentation. Short notice gets paid for in inbound contacts.
• Remove it from display. The method disappears from checkout, but the refund route stays open as long as sales can be disputed.
• Keep reconciliation running. Statements, identifiers, and supporting documents are still needed after closure, for accounting and for disputes.
• Clean up the documentation: help pages, terms and conditions, support scripts, and automated tests. This is the step teams forget, and it generates tickets for months.