Visa results: a revenue ratio can rise without a rate change

From revenue growth to a question about rates

Read Figure 1 through a question about rates.[S1] An increase in revenue divided by current payment value would not, by itself, identify an increase in contractual rates. The activities behind the revenue must be matched to the activity in the denominator.

This distinction serves more than a cautious reading of strong results. Revenue growth from greater use at unchanged rates has economic value. Calling that growth evidence of stronger pricing power requires further evidence, however. Greater use and a rate change imply different reasons for revenue to persist and different judgments about competition.

The fictional examples below hold two rate coefficients fixed. With current-period payment value also unchanged, one comparison changes only prior-period payment value; another changes only the current transaction count. Separating the changes reveals causes that would be obscured if both inputs moved together. These calculations do not estimate Visa’s actual rates or customer mix.

Readers can use the comparison tables to match revenue in their own documents with the relevant activity period, value and count. The result is a specific question: which further document is needed to support an explanation based on a rate change? Aggregate company figures cannot calculate an individual merchant’s charges.

Reconciling revenue is a different task from finding a unit price

Figure 1 allows reconciliation to be distinguished from a pricing inquiry. A correct total does not establish that dividing it by one measure of activity produces a contractual price.

Consider adding compensation based on payment value to compensation based on the number of transactions processed. Both are revenue in the same currency, so their sum has economic meaning. Dividing that sum by payment value also converts the transaction-based component into a value-based ratio. The result moves with the relationship between count and value. Accurate addition does not remove that relationship.

It helps to put the reconciliation question and the pricing question on separate lines in an analysis memo. A total that reconciles can make the second task look complete. Identifying a price requires an account of the unit, the covered transactions and whether the same contractual terms apply.

The difference between the price of goods and remuneration for a service is explored in the US PPI and fuel-retail contract reference analysis. This article takes the calculation a step further. When revenue contains different activity bases, how much can a ratio with a common denominator change while the rate coefficients stay fixed?

Assign an activity period to each revenue category

To carry Figure 2 into a pricing inquiry, retain separate columns for the recognition period and the activity basis. Compressing them into one makes it harder to establish which inputs are held constant in the later calculation.

This is more than the presence of an older number in a table. It determines which activity change can explain revenue. Revenue reported for one period need not refer to activity from one common period. Giving a table a single quarterly heading does not align its underlying bases.

The order of inquiry follows from this. Match each revenue component to its activity period before asking how compensation per comparable unit changed. Putting current activity under every revenue component can make differences between past and current activity appear to be price changes.

For another treatment of the relationship between transactions and reported earnings, see Exxon’s physical sales and hedges traced through the same transaction. Its settlement and delivery questions should not simply be imported here. The present task is to preserve each component’s activity basis before dividing the components by a common denominator.

Set up two fictional contracts and limit what changes

For the following calculations, assume a fictional service provider receives two kinds of compensation: 1% of prior-period payment value and $0.10 for each transaction processed in the current period. Exclude other revenue, deductions, minimum charges and tiered pricing. This is a small counterexample to an inference from a ratio to a price, not a reconstruction of Visa’s contracts.

Let P denote prior-period value, V current-period value and N the current transaction count. Model revenue R is 0.01 × P + 0.10 × N. All values are in US dollars. Within the model, V and N cover the same transactions. That assumption does not establish that two real published series cover the same population.

In baseline case A, P is 10,000, V is 20,000 and N is 200. Value-based compensation is 100 and count-based compensation is 20, giving revenue of 120. Dividing by current value gives 120 ÷ 20,000 = 0.006, or 0.6%. Average value per transaction is 20,000 ÷ 200 = $100.

Known coefficients let us trace the causes of a ratio change completely in this example. Real analysis may not permit that degree of explanation. A small model first establishes which outcomes require a rate change and which can occur without one; the analysis can then return to actual documents.

A versus B: change only the prior-period value

Case B keeps current value at 20,000 and the current count at 200, as in A, but raises prior-period value to 20,000. Value-based compensation becomes 200. Adding count-based compensation of 20 gives revenue of 220, and 220 ÷ 20,000 = 1.1%. Neither rate coefficient has changed.

The move from A’s 0.6% to B’s 1.1% is a difference of 0.5 percentage points. It is not evidence of a 0.5% rate increase. Compensation tied to prior activity rises by 100; dividing that increment by the unchanged 20,000 gives 0.005. Figure 3 assigns the increase to the value-based component.

Even confirmed equality of current activity is insufficient to identify a price here. Prior activity must also be held constant to fix that component. Similar current conditions do not establish similar past activity. Assuming they do can manufacture an apparent pricing difference.

If an analysis sheet contains only current activity, add the field for the relevant prior period. The purpose is to connect revenue growth to the right period, rather than to relabel B’s growth as weak performance. Whether elevated past activity persists, and how revenue composition changes as current activity feeds through, are distinct hypotheses for subsequent results.

A versus C: how many transactions carry the same value?

Case C keeps prior value at 10,000 and current value at 20,000, as in A, but raises the count to 400. Value-based compensation remains 100; count-based compensation becomes 40. Revenue is 140 and the ratio to current value is 0.7%. The rate per transaction remains $0.10.

Processing 20,000 across 200 transactions rather than 400 gives average transaction values of $100 rather than $50. The same total is split into smaller transactions. With unchanged compensation per transaction, the total of that compensation rises with count. A higher ratio per dollar does not imply a higher price per transaction.

Figure 4 isolates that change. A versus B changes the past value; A versus C changes the present count. Selecting a comparison pair each time makes clear which difference is being explained. Comparing B directly with C changes both inputs, so it cannot isolate either effect.

More transactions do not necessarily produce more real-world profit. Processing costs and other contractual terms lie outside this model. Its conclusion is narrower: count-based compensation can move a revenue ratio even when total payment value is unchanged. Profit and processing-capacity judgments require their own inputs.

Two inputs sit inside the single ratio

Dividing R by V gives R ÷ V = 0.01 × (P ÷ V) + 0.10 × (N ÷ V). The left side is one ratio, but the right side has two inputs. P ÷ V compares prior and current values. N ÷ V measures transaction density: the number of current transactions per current dollar. The expression contains more than rates.

P ÷ V is dimensionless because both amounts use the same currency. N ÷ V, in contrast, has units of transactions per dollar. Multiplying it by a coefficient in dollars per transaction converts it into a value-to-value ratio. The two terms become comparable after conversion; their original rate units are different.

For A, P ÷ V is 0.5 and N ÷ V is 0.01. The first term is 0.005 and the second 0.001, totaling 0.006. Only the first input changes to 1 in B; only the second changes to 0.02 in C. Written this way, the expression helps select the next input to investigate after observing a ratio difference.

This decomposition does not claim that two categories explain real revenue. It is the smallest arithmetic example of converting several bases into one denominator. Adding real revenue components would require adding each component’s basis and unit. An unexplained remainder cannot simply be assigned to a convenient rate-change term.

Transaction density is an inverse average only with matched coverage

Within the model, V and N cover the same transactions, so N ÷ V is the inverse of average transaction value V ÷ N. The inverse of A’s $100 average is 0.01 transactions per dollar; the inverse of C’s $50 average is 0.02. Halving the average doubles this density, providing another view of a ratio rising without a rate change.

Two real aggregates need not cover the same transactions. If coverage differs, the division can still be performed, but its interpretation as an average lacks support. Adding transactions from another route to the model’s count, for example, raises density without changing value. That rise would not establish that the same customers split their purchases.

The worksheet therefore includes fields for included and excluded transactions as well as period and currency. A common company or network name does not prove a common population. Definitions of both payment value and processed count must be checked. If coverage cannot be matched, the average-value field can remain blank.

This is why the article does not calculate an average ticket from the actual published series. The fictional example explicitly aligns the populations and therefore permits that calculation. Applying it to real documents requires evidence for the assumption. Having all the numbers is different from having matching coverage.

Use a two-axis table to separate combined changes

Figure 5 combines prior value of 10,000 or 20,000 with a current count of 200 or 400. Current value remains 20,000 and the rates are common throughout. Case D adds prior value of 20,000 and a count of 400 to cases A, B and C. Its revenue is 200 + 40 = 240, a ratio of 1.2%.

Moving down the table changes only prior value; moving across changes only count. The vertical revenue difference is 100 in either column, while the horizontal difference is 20 in either row. The additive model allows these differences to be traced separately. D’s revenue exceeds A’s by 120: the vertical 100 plus the horizontal 20.

The larger 100 does not establish that prior value matters more than count in the real business. Those differences follow from the input changes and coefficients chosen for this example. Different changes in value or count produce different revenue increments. The table traces causes rather than ranking their real-world importance.

The same discipline helps when comparing published ratios across companies or regions. Once a difference is found, ask whether the underlying activity combinations match before attributing it to rates. If those combinations are undisclosed, several explanations remain possible. Figure 5 makes that missing information specific.

The same 0.6% is compatible with different rates

Now hold A’s activity and revenue fixed and work backward toward rates. One row containing prior value of 10,000, a current count of 200 and revenue of 120 cannot determine both a value coefficient a and a per-transaction coefficient b. The equation is 10,000a + 200b = 120, with two unknown coefficients.

The original a = 0.01 and b = 0.10 give 100 + 20 = 120. Another pair, a = 0.008 and b = 0.20, gives 80 + 40 = 120. Both produce the same 0.6% ratio to current value. An identical ratio is therefore not evidence of identical contractual terms.

The limitation applies both to inferring a rate increase from a rising ratio and to inferring unchanged rates from an unchanged ratio. Several changes can offset one another. Stability in the total is a useful observation, but another relationship is needed to determine what remained stable inside it.

The purpose of further evidence is not simply to accumulate numbers. It is to obtain different information about the two coefficients. Restating the same row in another unit, or adding its percentage in another column, supplies no independent condition. Ask which unknown a document can reduce before adding it to the comparison.

When one further row identifies the coefficients

Suppose A and B are known to share both rate coefficients. With count unchanged, the revenue difference of 100 divided by the prior-value difference of 10,000 identifies a = 0.01. Returning to A leaves 120 − 0.01 × 10,000 = 20; dividing by 200 transactions gives b = 0.10.

A and C also work under the common-rate assumption. Prior value is unchanged, so divide the revenue difference of 20 by the count difference of 200 to find b = 0.10 first. The remaining 100 divided by prior value of 10,000 gives a = 0.01. Changing inputs at fixed rates shows both a counterexample and what helps identify the rates.

Picking two real periods does not automatically justify this calculation. Common coefficients, no other revenue or deductions, and matched transaction coverage are conditions of the model. If those conditions change, other factors enter the difference and it cannot uniquely be allocated to these two coefficients.

A reader with contractual documents can check whether rates, effective dates or covered transactions changed. Aggregate company revenue tables alone do not establish those conditions. The example must not be repurposed as a way to recover Visa’s actual rates from two quarters of results.

Two rows can repeat one relationship

From A to D, both prior value and count double, and revenue doubles from 120 to 240. The rate equations are 10,000a + 200b = 120 for A and 20,000a + 400b = 240 for D. The second is just twice the first; it supplies no new constraint. The number of rows does not determine whether the system can be solved.

The alternative coefficients that fit A, a = 0.008 and b = 0.20, also fit D: 160 + 80 = 240. Two different rate structures explain both observations. More observed activity and more revenue do not necessarily provide more information separating the two forms of compensation.

Figure 6 compares the choice of row to pair with A. A with B, or A with C, changes the relationship between inputs if the other model conditions hold. A with D preserves their proportions. Recognizing that difference avoids adding rows that do not add information to a revenue explanation.

Changing only one input is sufficient here, but it is not necessary in every solvable system. Several inputs can change together and still supply different information, allowing simultaneous equations to identify coefficients. A/B and A/C make the reason for the difference easy to follow; they do not define the only possible identification method.

A longer reporting window does not automatically align activity

Would moving from a quarter to a full year resolve the timing mismatch? A longer window can reduce the relative influence of its boundaries. It does not remove the shifted activity basis from the sum. Extending the window and aligning the relevant activity are different operations.

Extend the fictional example to two quarters. Period 1 is A: prior value 10,000, current value 20,000, count 200 and revenue 120. Period 2 has prior and current values of 20,000, a count of 200 and revenue of 220. Combined revenue of 340 divided by combined current value of 40,000 is 0.85%, different from A’s 0.6% and B’s 1.1%.

The value-based compensation refers to 10,000 + 20,000 = 30,000 across the two periods. The denominator instead totals 20,000 + 20,000 = 40,000. Adding count-based compensation of 40 gives 0.01 × 30,000 + 40 = 340. Calling the two-period window a half-year does not make the two value bases equal.

The boundaries remain

For consecutive periods, the sum of prior-period values equals the sum of current-period values plus the value immediately before the window, less the value in its final period. Interior periods overlap; the boundaries remain. Figure 7 shows this relationship. Equal boundary values cancel the value-basis mismatch, but an annual heading is no evidence that they are equal.

Even when the value mismatch cancels, the count-based component still depends on total count relative to total value. A longer aggregation period does not turn two forms of compensation into one rate. Each issue must be assessed separately: which becomes less influential with a longer window, and which remains?

A unit conversion changes the coefficient’s number

The number written as a coefficient changes when counts are expressed individually or in thousands. A charge of $0.10 per transaction is $100 per thousand transactions. Writing 200 transactions as 0.2 thousand gives 100 × 0.2 = 20, the same compensation. Keeping the coefficient unchanged after changing the count unit would shrink revenue by a factor of 1,000.

Amounts require the same care. If $10,000 of prior value is written as 10 thousand dollars, an output in thousand dollars is 0.01 × 10 = 0.1 thousand dollars. Keeping the output in dollars requires an additional conversion. A familiar-looking percentage is no reason to overlook different table units.

These conversions also affect N ÷ V. A ratio calculated from thousands of transactions and millions of dollars cannot be compared numerically with transactions per dollar without conversion. An intermediate ratio with dimensions must carry the units of both numerator and denominator. A percentage for the final R ÷ V does not make every intermediate column dimensionless.

Writing units next to the formula helps catch the error early. When people bring values from different tables, a field name alone may not reveal the required conversion. Preserve the original unit first, then convert to the common comparison unit so another reader can retrace the calculation.

The ratio remains useful as a descriptive statistic

Dividing revenue by activity is not inherently wrong. With clear coverage and definitions, the result can describe that particular combination. Tracking a consistent definition provides a starting point for locating change. The problem arises when the value is called a contractual price and several possible causes are fixed into one explanation.

It is correct to say that the ratios for B and C exceed A’s. The reasons differ, however. A ratio compresses that distinction into one line, so preserve a route back to its two inputs. Adding explanation fields often retains more comparative value than discarding the summary measure.

Refusing all comparisons until every basis is matched can also be impractical. The key is to specify the question. Describing revenue relative to present scale and identifying a contractual rate change require different information. Missing conditions for the latter do not make the former meaningless.

A memo recording the ratio can give separate lines to its name, numerator, denominator, period, population and intended use. Labeling the use “contractual rate” requires supporting evidence. Labeling it “company revenue relative to current scale” allows the mixed activity periods to be stated explicitly. Precise naming preserves the number while reducing misinterpretation.

A deduction adds another moving component

Adding a variable deduction I to the fictional model gives R = aP + bN − I. Dividing by the common denominator adds a third term, −I ÷ V. Aligning the two activity bases does not also hold deductions constant. Figure 1’s reconciliation helps locate the net amount.

For a separate illustration, deduct 20 from A’s 120 to obtain a net amount of 100, or 0.5% of current value. Reducing the deduction to 10 gives 110, or 0.55%, at unchanged rates and activity. This extension isolates deductions. The main A/B/C comparisons contain no deduction; their revenue remains 120, 220 and 140.

A smaller deduction and a higher charged rate may move company net revenue in the same direction, yet represent different changes in a contract and its persistence. A single aggregate ratio cannot replace rate schedules, volume conditions and the terms governing discounts or benefits.

Classification must follow the actual definitions. Not every payment to a customer belongs in the same deduction line, and not all such payments are operating expenses. The task is to match the reader’s classification to the company’s. For a different, concrete KPI-scope problem, see Robinhood net deposits classified by the origin of each item.

Match the payee before moving to a merchant statement

Visa’s US-facing explanation describes interchange as a transfer between the acquiring and issuing financial institutions. A merchant negotiates the merchant discount with its financial institution; that charge may include processing services.[S3] Aggregate company revenue ratios cannot be carried directly into a particular merchant contract: the recipient of the payment matters as well as the calculation basis.

To examine a shop’s own charges, first identify whose invoice or statement is being read. A revenue line in company accounts and a line on the merchant’s bill need not aggregate the same service on the same terms. A published ratio is insufficient to establish overcharging or a rate increase before the actual agreement and statement are matched.

The real contract determines whether the merchant has per-transaction, value-based or fixed components. The two coefficients in this article are not a tariff for automatically classifying those charges. If the actual structure differs, retain its additional inputs instead of forcing it into two coefficients.

The terminology in a US-facing explanation should not be extended uniformly to every country’s contracts. Location, counterparty, currency and covered transactions can affect what must be checked. The reader’s task is to select documents explaining their charges and specify the differences from an aggregate ratio, rather than predict a personal rate from this article.

Equal payment totals do not establish equal consumption

V is the total value of the model’s transactions. It contains no information about the goods, quantities or services purchased. A and C have different average values, but possible explanations include purchase frequency, product mix and whether payments are combined or split. The average alone cannot select among them.

That also limits a direct inference from a company revenue ratio to household living conditions. A higher ratio does not immediately establish a greater burden, nor a lower one an improvement in living standards. What consumers purchase, whom they pay for payment services and what the company recognizes as revenue require different evidence.

International comparisons also require care with amounts translated into a common currency. This is not a numerical claim about foreign exchange’s contribution to these results. It is the arithmetic point that changing a conversion basis can change a common-currency amount even when local-currency transactions are unchanged. Record the translation method when comparing count-to-value ratios.

Real company analysis often asks simultaneously whether demand, price or mix changed. The counterexamples show why one ratio cannot select among them. Rather than discard demand information, give demand measures and explanations of revenue mechanics their respective roles.

Choose separately what persists in a revenue projection

For company valuation, the remaining question is whether the source of revenue growth can recur. Extending A-to-B as a rate increase requires different assumptions from extending it as earlier activity feeding into revenue. Applying the final ratio increase mechanically to the next period drops that distinction from the calculation.

A-to-C likewise raises the question of whether transaction density persists. A continuation of smaller transactions and a return to more consolidated payments imply different counts per dollar. Fixed rates and a fixed pattern of use are separate hypotheses. Holding one constant does not hold the other constant.

Adding costs and taxes to analyze profit introduces further questions. More transactions may fit within existing facilities and contractual capacity, or require additional spending. The revenue equation does not answer this. Turning the counterexample into a profit forecast requires new inputs and evidence.

Comparison with market expectations also requires information about what investors had expected. Revenue growth alone does not determine a share-price direction. The calculation can improve an earnings hypothesis by making its dependence on rates, activity and the combination of activities explicit.

Keep six matches in the evidence worksheet

Figure 8 is a worksheet for manually transferring evidence into a personal memo. First record the revenue label and whether it is a total or a component. Next match it to the activity period. Third distinguish the activity unit: value, count, fixed amount or another basis. These fields establish the question of why a particular denominator is appropriate.

Fourth record coverage for revenue and activity: the whole company, a region, a contract or a processing route, for example. Fifth distinguish deductions from items accounted for as separate expenses. Sixth record the evidence for unchanged or changed rate coefficients.

Leave unknown values blank and name the document needed. “Missing the contractual terms applying to this revenue category’s activity period” is more useful than “need more pricing information.” The worksheet’s output is an account of whether a calculation is blocked by period, unit or scope.

Record both publication date and the period described. Updated material about an earlier period need not establish new contractual terms. Separating a value retrieved from a newer document from a change in coverage or definition prevents unnecessary overwriting and mistaken historical comparisons.

One completed field can change the question

Suppose a fictional investigation records current value of 20,000, current count of 200 and model revenue of 220, without recording whether rates are unchanged. At that stage the ratio is known to be 1.1%. Comparing it with A’s 0.6% does not establish a higher rate.

Now add evidence that prior value was 20,000. The inputs match B, and unchanged coefficients can explain revenue of 220. This refutes the claim that a rate increase is necessary to produce the difference. It does not prove rates were actually unchanged: another coefficient pair could also explain the total.

Finally, documents showing that both model contracts remained unchanged narrow the explanation further. In this example prior value explains the revenue difference. Evidence does more than strengthen a conclusion; it removes unnecessary hypotheses. Specifying what each document can eliminate makes the purpose of research concrete.

The same sequence can guide comparison of a personal statement with a company announcement. A personal statement cannot fill undisclosed company-wide terms, however. Separate fields answerable for the reader’s own transactions from those still open for the company. Combining different scopes does not create a complete contract table that never existed.

What would support an explanation based on a rate change?

The article does not rule out rate changes. Evidence linking a contractual amendment, effective date, covered transactions and actual invoicing would support that interpretation. The counterexamples show that an aggregate ratio change cannot substitute for that evidence. Direct documentation can change the assessment.

Comparisons with aligned quantities and activity periods can also add information, provided the matched conditions are explicit. The same quarter, company and currency alone do not fix both activity bases. An estimate of common coefficients must explain how other changes were treated before its meaning can be judged.

Conversely, unchanged contractual terms combined with evidence of a changed activity combination strengthen an explanation that does not depend on higher rates. Changed deductions or coverage call for another comparison. Evidence should narrow the explanations that remain, rather than defend whichever hypothesis was chosen first.

If the transaction population is common, all compensation applies to the same current value at one coefficient, and no deductions or fixed components exist, revenue divided by that value may be interpretable as a price. The two-basis counterexample does not deny such a case. Stating the conditions under which a ratio can identify a price is more useful than treating all ratios as suspect.

SG Group View: preserve the total without overdetermining its cause

SG Group emphasizes the conditions required to move from comparing totals to comparing causes. Separate comparisons like Figures 3 and 4 avoid compressing the explanation into one price. An observed total can stand while the questions about its causes become more specific.

The benefit is not perpetual skepticism toward a strong number. It is a clearer choice of the next document: prior-period activity for a missing prior-value field, series definitions for uncertain count coverage, or applicable terms for a contractual amendment. These are different unanswered questions, not one generic request for more detail.

For a comparison extending to cash receipts, Celestica’s bank receipts and customer collections offers a separate evidence route. The two-basis calculation does not solve that cash-collection problem. It establishes the activity match first, after which cash evidence can be examined if needed.

Putting fixed-rate counterexamples alongside conditions that identify rates lets the reader move beyond “this is not a price.” The tables show which completed field permits which interpretation.

In the next document, check the matches before updating the numbers

When reading subsequent results or supplementary material, first check whether categories, activity periods, coverage and units remain comparable. A document retaining the definitions can supply another row. Changed scope or definitions require a bridge. Updating a growth number while preserving outdated comparison assumptions would mislead.

Additional rate information must also be read for its scope: the company, one customer or a particular region. Specific local terms have value, but extending them to all transactions requires further evidence. The worksheet’s coverage field records which gaps a new document fills and which remain.

A single ratio fits neatly in one column. A reader trying to judge a rate change still needs the prior/current value relationship, the count/value relationship and evidence supporting common coefficients. Start by checking A/B and A/C separately, then ask whether the reader’s own table preserves their difference.