Engineering acquisition economics for regulated deposit and lending products

A $2 billion easy-access savings book, a target of $200 million in new balances, and two ways to buy it: lift your rate by 25 basis points, or pay a $300 cash bonus to every new checking customer who sets up direct deposit. The rate move never appears in a marketing budget, which is why it looks free. Do the arithmetic and it is usually the more expensive of the two.

This lesson assumes the fully loaded per-channel acquisition costacquisition 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 → the measurement lesson builds, and the multi-year value model its sibling constructs. What is specific to a regulated balance-sheet business is that you can buy growth with price as well as with promotion, and the two land on different P&L lines, under different owners, with different regulatory treatment of the dollar you bought.

The two ways to buy a deposit

Rate is a price change. It applies to every dollar already on the book, it recurs every month until you reprice, and the cost sits in treasury's cost of funds where no campaign report will find it.

A bonus is a one-off cost, applied only to new accounts, sitting in the marketing line, switchable off next quarter without an existing customer noticing.

Referral sits between the two: cash per customer, but the screening work is done by someone who already banks with you.

The same split runs on the lending side. You can buy loan volume with the APR you advertise, which compresses margin on every loan you write for its whole term, or with acquisition spend, which is one-off and recoverable. The advertised rate cannot be targeted at the segment you want; a broker fee or a paid-search bid can.

What the 25 basis points actually cost

Take the numbers above. Moving from 4.00% to 4.25% costs $5 million a year on the $2 billion you were keeping anyway. The new $200 million costs $8.5 million at the new rate. Incremental interest expense: $13.5 million to raise $200 million, so a marginal cost of funds of 6.75%, not 4.25%.

If brokered CDs or an FHLB advance would have funded the same balance sheet at 5%, the rate rise destroyed value and marketing never saw the invoice. That comparison, retail acquisition against wholesale funding, is the hurdle every deposit campaign should clear before it is signed off.

The obvious dodge is to price new money separately: a new-customer-only tier, or a whole new brand with no back book. ING Direct did the second properly in the late 1990s, one savings product, no branches, a rate above the high-street average, and a cost base low enough that the above-market rate still produced a competitive all-in cost of funds. ING sold the US arm to Capital One in 2011 and the Canadian arm to Scotiabank, now Tangerine. Two catches. A brand with no back book has one by year three, and by then it owns the same repricing problem with a base of rate-shoppers. And dual pricing of new and existing savers is exactly what the FCA pressed banks on in its 2023 cash savings review, which is the fair-treatment question its own lesson covers.

What a $300 bonus actually costs

The bonus is the floor, not the number. Add the media to drive a funded account, the wasted media on applicants who never qualify, a loss reserve for bonus churners who open, collect and leave, and the KYC and AML screening every single account must pass whether it funds or not. A "$300 offer" commonly lands near $500 per funded, qualified account.

Now convert it into the unit that matters for a deposit product. $500 to acquire an account carrying $5,000 of average balance is 10% of the balance, paid up front, against a spread of roughly two to three points a year. You have pre-spent three or four years of net interest income. If the account closes in month 14, you did not fund anything: you rented $5,000 at an annualised cost of about 8.5%.

Where deposit value actually comes from

Net interest income on balances

The spread between what the bank earns on the money and what it pays the depositor is NIM (net interest margin). A $5,000 checking balance funded at a two to three point spread throws off roughly $100 to $150 a year. Modest, recurring, and sticky. The FDIC quarterly banking profile publishes industry NIM trends for calibrating your spread assumptions.

Not every dollar carries that spread, though. Under the Basel III liquidity coverage ratio, stable retail deposits are assumed to run off at 3% in stress while less stable retail deposits are assumed to run off at 10%, and rate-shopped, uninsured or non-relationship balances fall in the worse bucket. A dollar in the worse bucket forces the bank to hold more low-yielding liquid assets against it, so the same headline spread earns less. Rate-bought money is discounted twice: shorter tenure and thinner effective spread.

Interchange

Interchange is the fee a merchant's bank pays when your customer swipes. An active debit user spending $1,500 a month generates tens of dollars a year. Under the Durbin Amendment, banks above $10 billion in assets face capped debit interchange, so the same account is worth measurably less on the large side of that line. Your economics change on the day you cross it.

The second product

Direct-deposit customers are the pool the cross-sell sequencing lesson works on. It is usually the largest value component and the easiest to overstate. Count it only where the second product actually lands, and only at the conversion rateconversion rateThe percentage of visitors or prospects who complete a desired action (purchase, sign-up, contact form), calculated as conversions divided by total opportunities.View full definition → you have observed, not the one in the business case.

The teaser-rate trap

A promotional APY pulls in deposits from people who are, by construction, rate-sensitive. When the teaser expires or someone bids higher, the money moves. That is hot money.

Marcus by Goldman Sachs shows both halves. Launched in 2016 with no legacy retail book and a consistently top-of-table savings rate, it gathered tens of billions in deposits and gave Goldman a cheaper funding source than wholesale markets. The deposits worked. The consumer lending business bolted onto them did not: Goldman disclosed billions in cumulative losses and retreated from most of it, keeping the deposits. Rate buys balances reliably. It does not buy the relationship that makes balances profitable.

The failure mode is measuring these campaigns on deposits raised rather than deposits retained. Model a decay curve. A 90% annual retention assumption versus 60% swings lifetime valuelifetime valueLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → by multiples, and the campaign that looked best on day one usually has the steeper curve.

Putting it together: which channels pay back

Compute per channel, never blended. Blended numbers hide the losers.

| Channel | Cost shape | Deposit behavior | Verdict |

|---|---|---|---|

| Referral from existing customer | Low, one-off | High balances, sticky, cross-sells | Best payback |

| Branch / relationship opening | High, one-off | Sticky, strong cross-sell | Slow but durable |

| Paid search on "best checking bonus" | High, one-off | Bonus-seekers, high churn | Often loses money |

| Rate-comparison sites | Permanent, on the whole book | Hot money, LCR-penalised | Payback only if retained |

Nubank is the referral extreme: past 100 million customers, acquisition costacquisition 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 → of a few dollars per customer, with most arriving organically or through someone they know, and spend going into product and service rather than rate-table placement. The catch is timing. A referral engine cannot be dialled up to hit a funding target by quarter end, and when treasury needs the money by a date, rate wins by default. That is how banks end up with expensive books they did not intend to build.

The payback period test

A 4 to 1 lifetime ratio is useless if recovering $500 takes seven years. You bleed cash throughout and you are betting on retention assumptions that far out are guesses.

Walk the checking example: $500 acquired, roughly $160 a year from net interest income and interchange before cross-sell, so payback on the core account alone runs past three years. Above the Durbin threshold, slower still.

Lending inverts the arithmetic. Cost per funded loan equals cost per application divided by your approval rate, so a 12% approval rate turns a $60 application into a $500 loan. Then adverse selection bites: comparison-table shoppers skew toward thinner credit. On an unsecured loan priced at 12% with a 5% cost of funds, moving annual credit losses from 4% to 5.5% erases the margin entirely. No creative fixes a decline curve.

Marketing's real job here

• Reprice offers by predicted value. A prospect who looks like a direct-deposit, high-balance customer can justify a bigger bonus than a serial churner. Same headline offer, different real cost.
• Bring treasury the wholesale-funding comparison before proposing a rate move, and bring the LCR bucket the new money will land in.
• Design against selection effects. Requiring direct deposit to earn a bonus screens out churners and shifts channel economics more than any creative change.
• Kill vanity metrics. Accounts opened and deposits raised are inputs; retained, funded, cross-sold accounts are the outcome.

The Consumer Financial Protection Bureau's guidance on deposit account terms and disclosures matters here, since bonus conditions and rate changes are regulated and misdisclosure carries real legal exposure.

Key Takeaways

• Rate and cash are both acquisition spend. Rate reprices your whole back book, so raising 25bp on $2 billion to win $200 million puts the marginal cost of funds near 6.75%, not 4.25%.
• Every deposit campaign should clear the wholesale funding hurdle. If brokered CDs are cheaper, you bought expensive money with a marketing story attached.
• Convert acquisition costacquisition 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 → into cost per retained deposit dollar. $500 against a $5,000 balance pre-spends three to four years of spread.
• Rate-bought dollars are discounted twice: shorter tenure and a worse LCR run-off assumption, which forces more liquid assets and thins the effective spread.
• On lending, cost per funded loan is cost per application divided by approval rate, and 150 basis points of extra credit loss can wipe out the whole margin.
• Referral economics win on payback (Nubank's few dollars per customer) but cannot be scheduled. Plan them a year ahead of the funding target, not a quarter.