Measuring true acquisition cost across banking channels

A regional US bank celebrates 10,000 new checking account "wins" from a Q1 campaign. Marketing spent $1.2 million, so the deck reports a Customer Acquisition CostCustomer Acquisition CostCustomer Acquisition Cost (CAC) is the total sales and marketing spend divided by the number of new customers gained in a period. It measures how efficiently you grow.View full definition → (CAC) of $120. Clean, defensible, wrong. Six months later only 4,100 of those accounts hold money and show any activity. The real cost per genuine customer was closer to $293, a number 2.4 times higher off identical spend.

This lesson computes CAC the way banking should report it: fully loaded, measured against funded accounts, and honest about what a consideration cycle measured in months does to attribution.

Why "applications" is the wrong denominator

An account opening is not a customer. Applications leak at identity verification, at approval, and at first deposit, in the pattern the account-opening funnelfunnelThe customer journey from awareness to purchase, typically Awareness, Interest, Consideration, Decision, Action, with prospects narrowing at each stage.View full definition → lesson maps stage by stage. What matters here is which survivor you put under the division line.

CAC is total acquisition spend divided by the number of customers acquired. The whole game is defining "customer" honestly.

The banking standard is the funded and active account: a qualifying deposit received, plus activity (direct deposit, card swipe, bill payment) inside a defined window, typically 60 or 90 days.

Rule of thumb: if you would not count it in your active deposit base, do not count it in your CAC denominator.

Cards need a different denominator. An approved card that is never activated consumes underwriting and card production and returns nothing. For card portfolios the honest test is first purchase within 90 days, which is why Capital One's long history of direct mail testing was judged on booked-and-active accounts rather than response rate. Mail response for card offers runs well under one percent, so a numerator holding every piece posted over a denominator of mere approvals produces a figure nobody can act on.

Blended vs. fully-loaded CAC

Two adjustments separate a napkin number from a boardroom number.

Blended CAC combines all channels into one figure: what did an average funded account cost across everything we did?

Fully-loaded CAC adds every cost that made those accounts happen, not just media buys:

• Paid mediaPaid mediaVisitors arriving via paid ads or sponsored placements, where you pay a platform to display your message rather than earning visits organically.View full definition → (search, social, display).
• Branch acquisition cost (staff time attributable to new-account opening, not the whole branch).
• Referral and signup bonuses paid to customers.
• Agency fees, martechmartechThe connected set of software tools a marketing team uses to plan, run, measure and automate campaigns across channels.View full definition → tooling, creative production.
• Sometimes an allocation of marketing salaries.

The most common banking mistake is leaving out the referral bonus. A $200 "open a checking account and get $200" promotion is a direct acquisition cost. If you paid it, it belongs in the numerator.

A worked example: three channels, one campaign

Here is our checking account campaign. All figures are illustrative for teaching, not real bank data.

Step 1: The naive number

$1,200,000 / 10,000 applications = $120 per application. This is what the launch deck showed. Ignore it.

Step 2: Blended CAC on funded accounts

$1,200,000 / 4,100 funded accounts = $293 per funded customer.

Step 3: channel-level fully-loaded CAC

• Branch: $500,000 / 2,600 = $192
• Digital: $450,000 / 1,000 = $450
• Referral: $250,000 / 500 = $500

Digital produced the most applications and the worst funding rate (1,000 / 5,200 = 19 percent). Branch produced the cheapest real customers, because someone who walks in and opens an account usually funds it that day (2,600 / 4,000 = 65 percent).

That branch number survives only because of an allocation choice. We charged branch with new-account staff time and nothing else. Allocate a share of occupancy, security and cash handling and branch CAC often triples, which is the calculation that has closed thousands of branches. Both versions are defensible; publishing one and comparing it to a channel costed the other way is not.

Referral looks expensive per account and may still be your best channel, since referred customers tend to retain longer. That comparison runs against the per-customer value the LTV lesson models, never against CAC alone.

One more number the table hides: the marginal cost of the next account. If digital spend went from $300,000 to $450,000 and funded accounts from 800 to 1,000, the extra 200 accounts cost $750 each, not $450. Blended channel CAC is an average over a saturating curve. Budget decisions need the slope.

AttributionAttributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition →: the hard part

A customer sees a digital ad in January, asks a friend, compares rates for six weeks, then opens the account in a branch in March. Which channel gets the credit?

• Last-touch: the branch takes 100 percent. Simple, overcredits the closing channel.
• First-touch: the digital ad takes 100 percent. Overcredits awareness.
• Multi-touch / fractional: split credit across touches. More accurate, harder to build.

Long consideration cycles break all three. Deposit and mortgage decisions run weeks to months, while click and view windows are configured in days, so a 30-day lookback quietly reclassifies paid conversions as organic or direct. Rate moves make it worse: when central bank rates jump, savings inflows rise whether or not you advertised, and whichever campaign is live absorbs the credit. The defence is a geogeoThe practice of making your brand and content visible and citable inside AI-generated answers from tools like ChatGPT, Gemini and Perplexity.View full definition → or audience holdout, run long enough to cover the real decision cycle, so you measure incremental accounts rather than correlated ones.

For a first pass, many banks use last-touch and accept the bias. The discipline that matters is consistency: same model across channels and quarters, or comparisons mean nothing.

For a primer on attribution logic that transfers cleanly to banking, see Google's attribution documentation. Google sells both the media and the measurement, so read its default model choices as commercially interested.

The funding-rate multiplier

The strongest lever on banking CAC is not spend, it is funding rate: funded accounts divided by approved applications.

If digital's funding rate rises from 19 percent to 30 percent on identical spend:

• Funded accounts from digital: 5,200 x 0.30 = 1,560
• New digital CAC: $450,000 / 1,560 = $288

A 36 percent cut in CAC without an extra dollar of media, from onboarding fixes the funnel lesson details. This is why the CAC number belongs to marketing and product jointly, and why a product roadmap freeze is a marketing cost.

Sector benchmarks (treat as estimates)

Published banking CAC figures move for reasons that have nothing to do with efficiency: some banks count bonuses, some do not; some divide by approvals, some by funded accounts; some leave branch labour out entirely. Which sources are legitimate and how to normalise them is the benchmarks lesson's work. Three numerator problems belong here.

• Incumbents disclose marketing expense, not CAC. Capital One reports marketing expense running into the billions of dollars a year across cards, retail banking and brand, with no split by product or funnel stage. Dividing it by net new accounts is arithmetically clean and analytically empty.
• Organic growth flatters the blended number and hides the marginal one. Monzo reached millions of UK customers largely through word of mouth and referral, including the early "golden ticket" invite scheme, so its blended CAC always looked low. The question a CMO answers is what the next cohort costs, not what the cheapest one averaged.
• Bonus economics are constrained before you model them. UK firms have operated under the FCA's Consumer Duty since July 2023, and promotional terms have to stand up to a fair-value test, so some CAC-efficient offers are simply unavailable.

Your fully-loaded, funded-account CAC will almost always sit far above the per-application figure your first dashboard shows.

Common ways banks fool themselves

Counting the signup bonus as a "reward" not a cost. It is a cost. Put it in the numerator.

Using approved instead of funded accounts. Inflates volume, hides the leak.

Ignoring branch labor. Staff are salaried, so branch acquisition feels free. Their new-account hours are allocable.

Mixing time periods. Spend lands in Q1, funding matures over 90 days. Dividing Q1 spend by Q1 funded accounts undercounts. Match spend to the cohort it acquired.

Counting win-backs as new. A dormant customer who reopens an account is a reactivation with a different cost base and a different value curve.

Letting a rate move take the credit. If your savings CAC halved the quarter the central bank raised rates, you learned about the macro, not the campaign.

A quick cohort discipline

Tag every account with its acquisition campaign and channel, then measure funding at a fixed maturity.

CAC (cohort) = total_channel_spend / funded_active_accounts_at_day_90

funding_rate = funded_active_accounts / approved_applications

Only include accounts where:
  - qualifying_deposit = TRUE
  - activity_in_last_90_days = TRUE
  - prior_closed_account = FALSE

This keeps the denominator honest and lets you compare Q1 against Q2 on equal terms.

Key takeaways

• Divide by funded and active accounts, never raw applications. In the worked example that alone moved CAC from $120 to $293.
• Fully-loaded means everything: media, branch labor, agency fees, and especially referral and signup bonuses. State your branch allocation rule, because it decides whether branch looks cheap or ruinous.
• Compute CAC per channel, and then compute the marginal cost. The next 200 digital accounts cost $750 each while the average said $450.
• Funding rate is your cheapest lever. Better onboarding cut digital CAC 36 percent with zero extra spend.
• Long cycles defeat click windows. Hold out a geo and measure incremental accounts before you believe any channel report.