Competitive Compass Competitive Compass Signal for Financial Leaders
Methodology  ·  v6.0  ·  Q2 2026

Acquisition Compass methodology.

4 measurement layers. 3 treatments that change the answer. Every input published with a confidence tier. Q2 2026 reads against a first quarter rebuilt from primary filings, so every movement on the charts is a movement in acquisition cost.

At a Glance

What this framework measures.

Acquisition Compass turns public disclosure into a comparable cost per account and a comparable value per account for the 10 largest US card issuers. No issuer publishes a customer acquisition cost, so every figure here is a modeled estimate built on disclosed inputs and stated assumptions.

Period
Q2 2026, measured against Q1 2026 rebuilt from primary filings
Panel
10 issuers. 9 publish this quarter, Barclays US publishes a baseline
Primary ratio
5 years of discounted contribution lifetime value over Acquisition Investment
Hurdle
3.0x. Breakeven sits at 1.0x
Secondary view
Steady-state lifetime value over Operating CAC, the shape most published benchmarks take
Held constant
Allocation shares, attrition, margins, revenue per active and expected credit losses
Recalibration
Annual. Quarterly refresh moves acquisition cost only
Section 1

The 4 measurement layers.

Cost per account builds in 4 layers, so a reader can stop at whichever one matches the decision in front of them. Layer 1 answers what the marketing line buys. Layer 4 answers what an account actually costs the company, and that is the layer the published ratio uses.

L1

Operating CAC

Normalized acquisition marketing allocated by product, divided by gross accounts acquired. Use it to compare marketing efficiency across issuers on a like for like basis.

L2

Channel-Loaded CAC

Adds branch personnel on branch-originated opens only, plus identity verification blended at $25 digital and $100 branch-assisted. Use it to compare a branch franchise against a direct bank.

L3

Incentive Yield Offset

Adds the sign up bonus at a 60% qualification rate, carried separately under ASC 606 contra-revenue. Use it to understand how much of the offer a competitor actually pays out.

L4

Acquisition Investment

Layers 1 and 2 plus onboarding and fulfillment plus the expected incentive. This is the economic denominator, and this is what the published return is calculated on.

Gross opens

Where an issuer discloses gross accounts acquired, the disclosure is used. Where it does not, gross opens are modeled as net new plus base times a per-entity quarterly attrition rate. Validation flags any product where modeled closures exceed twice net new, because at that point the attrition assumption rather than the disclosure drives the denominator. That condition holds for consumer checking at every large deposit franchise in the panel, so cost per net add prints beside cost per gross open and the client-facing narrative leads card economics.

Section 2

How a verdict is set.

The verdict reads off 5 years of discounted contribution over Acquisition Investment.

VerdictRangeWhat it means
Excellent5.0x and aboveWell clear of the hurdle.
Above hurdle3.0x to 5.0xHealthy on the published hurdle.
Below hurdle1.0x to 3.0xThe account pays back and falls short of the hurdle.
Below breakevenUnder 1.0xThe account costs more than 5 years of discounted contribution returns.
Confidence grades travel with the number. Every output carries a grade propagated from its inputs. Tier A is a disclosed line item. Tier B is a disclosed figure requiring interpretation, or an annual disclosure applied to a quarter. Tier C is modeled, and should be read as a comparable estimate rather than a measurement.
Section 3

The 3 treatments that change the answer.

Most of the difference between this framework and a straight reading of reported marketing lines sits in 3 places. Each one exists because a reported figure would otherwise mislead.

3.1 Acquired versus onboarded

A gross new account means an account the issuer originated. Accounts arriving through a portfolio purchase, a program transfer or a bank conversion are onboarded, and they carry no acquisition marketing cost. Counting them as acquisitions makes the buyer look like the cheapest acquirer in the market. Q2 2026 carried 3 cases large enough to distort the panel.

IssuerBookSizeTreatment
CitiAdditional American Airlines cobrand portfolio, closed 24 April 20262MM+ accounts
$6.6B loans
Excluded. Organic general purpose acquisitions held at 1,899K, flat against Q1.
SynchronyLowe's commercial cobrand portfolio, April 2026$0.7B receivablesExcluded. Booked as an asset purchase rather than an origination.
PNCFirstBank customer conversion, completed 22 June 2026About 780,000 customers
95 branches
Excluded. Converted relationships are onboarded.

Citi's reported general purpose acquisitions moved from 1,899K in Q1 to 4,004K in Q2, and the press release footnote states that the onboarded accounts are included. Left unstripped, Citi would print the lowest cost per account in the panel by a wide margin, on accounts it purchased.

3.2 Contra-revenue acquisition cost

Some issuers route acquisition cost through revenue rather than expense. Citi is the only panel member where the effect is material. Its advertising and marketing line came in at $283M in Q2, while U.S. Consumer Cards non-interest revenue moved from negative $359M to negative $659M, a $300M swing that Citi attributes to higher partner payment accruals and new account acquisition costs.

ComponentQ1 2026Q2 2026
Reported advertising and marketing$233M$283M
USCC non-interest revenue-$359M-$659M
Acquisition-attributable share applied20%20%
Contra-revenue component added$72M$132M
Normalized acquisition marketing$305M$415M

The 20% share holds constant across quarters, so neither driver absorbs the whole sequential change, and a 10% to 30% band is carried on the entity tab. This is a Tier C estimate and it is labeled as one. The alternative, using the reported line alone, produces a figure the segment footnotes contradict. Stating an assumption openly beats inheriting a presentation choice silently.

3.3 Derived numerators

Where an issuer publishes no marketing line at any level of disclosure, the numerator is built from segment operating expense times a documented marketing share, converted at the disclosed quarterly average exchange rate, then cross-checked against a peer benchmark scaled by receivables. Barclays US Consumer Bank is the only panel member on this basis.

StepValueNote
Segment total operating expenseGBP 380mQ1 2026 results announcement.
GBP/USD 3 month average1.35Barclays-disclosed rate.
Marketing share of operating expense12%Tier C. Branchless direct issuer with no branch or teller cost in the line.
Derived numerator$61.6MEnters the Normalization tab.
Peer cross-check$47MSynchrony marketing per $B of receivables, scaled to Barclays end net receivables of $35.2B.

The derived figure sits above the private label benchmark, which fits a general purpose travel cobrand book carrying higher acquisition intensity. Two events reshape the Q2 base and are flagged on the tab: the American Airlines exit on 24 April, and the Best Egg acquisition around 1 May, which adds roughly $45M of monthly cost belonging to an origination platform rather than to acquisition marketing.

Section 4

Lifetime value and payback.

Lifetime value = Annual revenue per active × Contribution margin × NPV factor NPV factor = (1 − annual attrition) ÷ (discount rate + annual attrition), at a 10% discount rate.

One churn mechanism runs across all 5 products, which removes the double counting that a separate tenure assumption introduces.

Annual contribution = Revenue per active × Contribution margin Contribution margin = 1 − rewards and benefits − funding cost − expected credit losses − servicing and fraud − retention and partner share.

Three views print for every entity: 12 month contribution, 5 year discounted contribution lifetime value, and steady-state lifetime value.

Payback

Discounted payback solves for the point where cumulative discounted contribution equals Acquisition Investment. It prints beside every ratio, because a finance audience reads payback before it reads a multiple.

Section 5

Measuring movement.

A movement marker publishes only where both quarters read from verified filings. Section 11 on every entity tab recomputes Q1 allocation, Q1 gross opens, Q1 Operating CAC and Q1 Acquisition Investment from the Q1 input column, then compares them against Q2.

Lifetime value holds constant

Allocation shares, attrition, margins, revenue per active and expected credit losses hold stable within the year. That makes every quarterly movement an acquisition-cost movement, which is what a reader can act on. Card loss rates improved at 7 of the 10 issuers in Q2, so holding expected credit losses stable understates the improvement in lifetime value. That is the deliberate trade: a clean attribution of what moved is worth more than a marginally more current ratio.

New panel members publish a baseline

U.S. Bancorp, PNC and Barclays US carry a Q1 column that establishes a baseline and no published verdict movement. They enter the movement set at the Q3 refresh, once a second verified quarter exists on the same basis.

Where an issuer publishes 2 measures, pick one and say which

Bank of America reports card new accounts twice on 2 different scopes. The earnings presentation states over 1 million on a Consumer Banking plus Small Business plus GWIM basis. The Supplemental Information carries a precise quarterly series on a Consumer Banking plus GWIM basis: 999,000 in Q2 2026 and 884,000 in Q1 2026, with a 6 month column of 1,883,000 that confirms the line is a quarterly flow rather than a stock. The framework uses the supplement series because it is exact and consistent across both quarters. Pairing a numerator from one document with a denominator from the other introduces a scope error that stays invisible in the output.

Where an unquantified item could move a result, publish the threshold

Bank of America discloses marketing firmwide with no segment or category split, and the Q2 line carries a brand campaign the bank does not size. The base case excludes nothing, because no quantification exists to work from. The entity tab then solves for the share of that line which, treated as brand rather than direct acquisition, would leave the standalone no fee card ratio at breakeven. That threshold is about 15%, and publishing it is more useful than picking an exclusion and presenting the result as settled.

The general rule. Where a single undisclosed input could change a published conclusion, the model states the threshold at which the conclusion turns rather than choosing a value quietly.
Section 6

Basis breaks a reader must know.

4 issuers changed how they report between 2025 and 2026. Each break is tested on the Validation tab, and none of them is visible to a reader working only from a current quarter document.

Wells Fargo restated card metrics to exclude cobrand, effective Q1 2026

Q2 2025 new accounts moved from 643,000 as originally published to 452,000 restated, a 30% reduction. Purchase volume moved from $46.4B to $39.9B. Card loan balances were not restated. The 2 quarters of 2026 are comparable to each other and comparable to nothing published before April 2026. The Q2 2026 supplement drops the restatement footnote, so the caveat is invisible in that document. A residual issue remains: the card denominator now excludes cobrand while the advertising and promotion numerator stays firmwide, which biases Wells card cost per account upward. Both quarters share the basis, so the movement marker holds. The absolute level carries a known bias and resolves at the annual recalibration by carving cobrand marketing out of the numerator.

Citi resegmented, effective Q1 2026

U.S. Personal Banking is replaced by U.S. Consumer Cards, with general purpose and private label reported separately, and U.S. Retail Banking moved into Wealth. Prior history was recast in the 3 April 2026 8-K. Both 2026 quarters read on the new basis, and neither ties to pre-2026 Branded Cards series.

Capital One added Brex and the legacy corporate card to Domestic Card in Q2 2026

Q1 carries no recast. Q2 purchase volume rose 15.1% sequentially against 10.5% at Chase and 9.6% at Bank of America, so roughly 5 percentage points of the step is inorganic. The Q2 denominator is derived from volume adjusted to peer-consistent organic growth, stripping about $11B. Capital One declined to quantify the transfer, so the adjustment is Tier C.

American Express combined loans and receivables into Card balances, effective Q1 2026

A balance presentation change with no effect on the account denominators used here.

Section 7

Panel and denominators.

Each cobrand card belongs to exactly one entity panel, so a program is never counted twice. Where a program changes issuer mid-quarter, the accounts follow the issuer of record and the transfer is disclosed on both tabs.

IssuerClassStatus this quarter
JPMorgan ChaseUniversal bankPublished
Bank of AmericaUniversal bankPublished
CitigroupUniversal bankPublished
Wells FargoUniversal bankPublished
U.S. BancorpUniversal bankBaseline. New to panel
PNC Financial ServicesUniversal bankBaseline. New to panel
American ExpressPremium card issuerPublished
Capital OneCard-led bankPublished
Synchrony FinancialSpecialty and private label issuerPublished
Barclays US Consumer BankCobrand card specialistPending. H1 2026 results publish 28 July 2026

Card denominator basis

Card denominatorBasisTier
Wells FargoDisclosed quarterly, restated basis excluding cobrandA
American ExpressDisclosed proprietary new cards acquiredA
Citi private labelDisclosed quarterlyA
Bank of AmericaDisclosed quarterly in the Supplemental InformationA
Citi general purposeDisclosed less the onboarded portfolioB
JPMorgan ChaseAnnual 10-K disclosure applied to the quarterB
SynchronyDisclosed on the earnings call only, absent from filingsB C
Capital OneModeled from purchase volume. No account count disclosed at any frequencyC
U.S. BancorpModeled from card payment volumeC
PNCModeled from card loan balancesC
Barclays USModeled from end net receivablesC

Aggregation

Median and interquartile range lead every peer view, with n printed beside each result. A volume-weighted aggregate prints as the secondary view, and it is the right figure for any single headline cost per account, because the middle issuer changes between quarters. Universal banks and card-focused issuers aggregate separately, because private label and premium franchises carry structurally different cost per account. Pending entities stay out of every aggregate.

Private label cards cost far less to originate than general purpose cards, and that difference is real rather than an artifact. 42% of Citi's Q2 acquisitions were private label, which pulls its blended figure below the universal bank group. The fee card row is the clean read on Citi general purpose economics, and Synchrony is the right benchmark for the private label share.

Section 8

Limits.

Stating limits plainly is what makes the rest of the document usable.

Section 9

Refresh discipline.

The framework holds its value only when each refresh runs the same sequence. Three steps matter most.

Structural assumptions hold stable within the year. The annual recalibration is where revenue per active, contribution margin components, attrition bands and allocation shares get rebuilt. The highest value input to revisit at that point is fee card revenue per active, given the repricing cycle running across Chase, American Express and Citi.

Version history. v6.0 rebuilt Q1 2026 from primary filings, added U.S. Bancorp, PNC and Barclays US to the panel, built up the Citi numerator, stripped purchased portfolios from every denominator, and introduced the movement markers. Version 5 seeded Q1 with placeholders that sat far from reported figures for 5 of the 7 issuers then in the panel, so any movement measured against them would have described a data correction rather than a change in behavior. Figures published under v3.1 sit on the Q1 2026 first edition page and do not tie to this edition.
Section 10

Sources.

Prepared by Anuj Shahani, Comperemedia. Questions or feedback: anuj@competitive-compass.com