Adding a Payment Method: Incremental Revenue or Just Cannibalization?

Six months after adding a new checkout option, the report says 30% of transactions now use it. Everyone treats this as success. It is not evidence of anything. If 28 of those 30 points came from customers who would have paid by card, the method generated almost no new revenue — and if it costs more to accept, it destroyed margin while appearing to succeed. Payment methods create value in exactly three ways: converting customers who'd otherwise abandon, raising order values, or costing less than what they displace. Usage measures none of them. This guide covers the arithmetic of when a method pays for itself, and the test that actually answers the question.

The three sources of value

A payment method can only help you in three ways:

1. Incremental conversion. A customer completes a purchase they'd otherwise have abandoned.
2. Higher order value. The same customer buys more than they would have.
3. Lower cost. The method is cheaper than the one it displaces.

Everything else — brand association, customer preference, competitive parity — is either a proxy for one of these or isn't a financial benefit.

Now note which of the three is measurable from your existing reporting. Only the third. You can compute cost per transaction by method directly. You cannot see incremental conversion or order value lift in usage data, because both require knowing what would have happened otherwise — and that customer, by construction, didn't leave a record of the purchase they didn't make.

Which produces the structural problem: the one benefit you can measure is the one that's usually negative, since new methods frequently cost more than cards, and the two that justify the decision are invisible in standard reporting. So the evaluation defaults to usage, which measures neither.

Why usage misleads

Decompose that 30% adoption figure:

Segment	What they'd have done without the method	Value created
Switchers	Bought using another method	Negative if the new method costs more
Converted	Abandoned	Full margin on the order
Expanded	Bought less	Margin on the increment

Usage counts all three identically. And the switcher segment is almost always the largest, for a reason worth stating: a method convenient enough to attract new customers is convenient enough to attract existing ones. The same properties that drive incremental conversion drive cannibalization, so you cannot have the first without the second.

Which means high adoption is weak evidence and can be evidence of the problem. A method reaching 40% of transactions in a business whose total conversion rate didn't move has redistributed 40% of your volume onto a different cost structure and created nothing.

The diagnostic question that cuts through it: did total conversion rise? Not usage of the method — total orders per visitor. If that number is flat, everything the method did was redistribution.

The break-even arithmetic

Solve for the incremental share a method needs to justify itself.

Setup: $2.4 million annual volume, 30% moves to the new method ($720,000), current blended acceptance cost 2.6%, new method costs 3.4%, gross margin 40%, integration cost $18,000 amortized over three years plus $4,000 annual overhead.

The cannibalization cost. Let x be the incremental share of the $720,000. The switched portion is $720,000 × (1 − x), and each switched dollar costs an extra 0.8%:

• Cannibalization cost = $720,000 × (1 − x) × 0.008

The incremental gain. The incremental portion is new revenue, earning gross margin less the acceptance cost:

• Incremental gain = $720,000 × x × (0.40 − 0.034) = $720,000 × x × 0.366

Fixed costs = $6,000 amortization + $4,000 overhead = $10,000.

Set gain equal to cost plus fixed:

$720,000 × x × 0.366 = $720,000 × (1 − x) × 0.008 + $10,000

Solving: 263,520x = 5,760 − 5,760x + 10,000 → 269,280x = 15,760 → x ≈ 5.9%

So roughly 6% of the method's volume must be genuinely incremental for it to break even.

Two readings of that number, and both matter.

It's a low bar, which is the case for adding methods. The margin on incremental sales is large enough that even modest genuine conversion pays for a lot of cannibalization. A merchant convinced a method brings some new business should probably add it.

But it's not zero, and it rises fast with cost gaps and low margins. Re-run with 20% gross margin and a 1.5-point cost gap: the required incremental share climbs to roughly 15%. For low-margin businesses accepting expensive methods, the bar is genuinely high — which is exactly the situation where merchants add methods on the assumption that adoption equals success.

The costs beyond the rate

The processing rate is the visible cost and rarely the largest one.

• Integration and engineering, including the ongoing maintenance every additional integration creates.
• Reconciliation. Each method settles on its own schedule with its own reporting format, and each adds work to the process our reconciliation guide describes. This is a persistent labor cost that nobody attributes to the decision that caused it.
• Dispute handling, since methods have different dispute processes, timelines, and evidence requirements.
• Support burden from customer questions about an unfamiliar option.
• Settlement timing differences, which affect working capital — the arithmetic in our settlement timing analysis.
• Checkout complexity. The one people miss entirely: too many options reduce conversion. Adding a sixth method to a checkout that already has five can lower total conversion through choice friction, which means the method can be net negative even at zero cost.
• Compliance and contractual obligations that come with certain methods.

That checkout-complexity point deserves the emphasis. The evaluation is not "does this method help versus nothing" — it's "does this method help versus the checkout I currently have." A method that would be valuable in isolation can be harmful as an addition to a crowded page.

Where genuine incrementality comes from

Not all methods are equally likely to bring new business. The ones that genuinely convert share a property: they serve customers who couldn't or wouldn't transact with your existing options.

High incrementality:

• Methods serving customers without your current options. Bank payment for customers without cards, or cash-adjacent options for the population in our unbanked analysis. These are genuinely additive because the alternative was no sale.
• Financing for high-ticket items, where the customer's constraint is the amount rather than the mechanism — the incremental case in our deferred payment analysis.
• Local methods in new geographies, where your existing options genuinely don't work — per our cross-border guide.
• Business payment terms for commercial buyers who require them procedurally, per our terms guide.

Low incrementality:

• A wallet layered over cards your customers already have — convenience for existing customers, which may still be worth it for conversion friction but is mostly redistribution.
• A second method serving the same population as one you offer.
• Methods added for competitive parity without any hypothesis about who they serve.

The screening question before any integration: who specifically cannot buy from you today, and does this method let them? If you can't name the customer, the method is probably redistribution — and that's a five-minute conversation that prevents a five-figure integration.

Testing properly

The only method that answers the question is a randomized rollout, and it's more feasible than most merchants assume.

1. Expose the method to a random share of traffic — half, or less if volume permits — and withhold it from the rest.
2. Measure total conversion per visitor in each group, not usage of the method.
3. Measure revenue per visitor, which captures order value effects.
4. Measure blended acceptance cost per group, capturing the cannibalization directly.
5. Run long enough to cover normal variation and to let customers who need multiple visits complete.
6. Compare contribution per visitor — revenue times margin minus acceptance cost — which is the number that decides it.

What the outcomes mean:

• Conversion up, cost up: compute net. This is the interesting case and the arithmetic above resolves it.
• Conversion flat, method used heavily: pure cannibalization. If the method costs more, remove it.
• Conversion down: checkout complexity. Remove it or simplify the page.
• Conversion up, cost flat or down: unambiguous win, roll out fully.

Where a randomized test isn't possible, the weaker substitutes: staged geographic rollout, before-and-after with careful seasonal adjustment, or comparing customers who used the method against matched customers who didn't. All are meaningfully worse — the last one especially, since it compares self-selected groups — but any of them beats reading an adoption chart.

Why incrementality decays

A point that changes how often you should re-evaluate.

A method that's genuinely incremental at launch tends to become less so over time. The mechanism: the customers it converted become customers, and the method becomes their habit. Once a customer who previously abandoned now buys regularly using the new method, their transactions are no longer incremental — they're your baseline. Meanwhile existing customers keep discovering the method and switching.

So the incremental share falls and the cannibalized share rises, mechanically, over the life of the method. A method that cleared a 6% bar in year one may not clear it in year three.

Which argues for two practices almost nobody follows:

• Re-test periodically, at least for expensive methods, by withholding from a small holdout and measuring conversion.
• Track blended acceptance cost as a trend, since a drifting mix raises cost without any single decision causing it. If your effective rate has risen 40 basis points over two years with no repricing, mix shift is the explanation and the methods driving it should be examined.

The case for removing methods

Rarely considered and frequently correct.

Candidates for removal:

• Methods with low usage and real fixed cost. Each integration carries maintenance and reconciliation overhead. A method at 2% of volume may not cover its own administrative burden.
• Expensive methods with demonstrated non-incrementality. If a holdout test shows conversion is flat without it, its entire cost premium is loss.
• Redundant methods serving the same population as another option.
• Methods adding checkout clutter without a distinct customer segment.

Removal carries risk — some customers genuinely rely on a method, and the ones who leave don't tell you — which is why removal should be tested the same way as addition: withhold it from a random share and measure conversion. Same design, opposite direction, same answer.

The general principle worth ending on: a checkout is a portfolio, and portfolios should be pruned as well as expanded. Most merchants only ever add.

Frequently asked questions

---

This guide is educational and does not constitute financial advice. Worked figures are stylized illustrations chosen to demonstrate the arithmetic; acceptance costs, margins, and conversion effects vary substantially by business, and results should be established through testing rather than assumed.