Neobank Unit Economics: Active Accounts, Product Use and Cost to Serve

Two digital banks can report the same number of customers and have very different businesses. One may hold primary accounts with regular deposits and lending relationships; another may have many lightly used accounts. Their funding needs, support workload and revenue per active relationship will differ.

This analysis examines the product activity and cost structure behind customer growth. It separates customer-cohort contribution from company profit and preserves the role of partner banks, credit risk and compliance. The evidence window ends in December 2025; the cohort example is illustrative.

Define the active and funded customer cohort

Start with the denominator. A “customer” may be a person who completed identity verification, opened an account, received a card, used a product, or remained in good standing at the reporting date. A company may exclude charged-off, blocked, or voluntarily closed accounts. A business account, joint account, or product account can also be counted differently across issuers.

Nu Holdings gives a useful example of the required disclosure discipline. Its 2024 Form 20-F defines customers as individuals and small and medium-sized enterprises that opened accounts, excluding accounts charged off, blocked, or voluntarily closed. Nu reported 114.2 million customers at December 31, 2024. That is a reach measure with a stated boundary; it is not a claim that each customer generated comparable revenue.

Write the inclusion and exclusion rule beside every customer number. Record whether the measure is an account, legal customer, household, member, or product relationship; the reporting date; geography; and whether inactive or closed relationships remain in the denominator. Without those fields, a customer comparison is a count of labels rather than a comparable market signal.

Separate account count, funded accounts, and active users

Three layers should be separated: relationships that exist, relationships with money or a funded product, and relationships that created qualifying activity during a stated window. Funding can mean a deposit balance, direct deposit, an investment balance, or an outstanding loan. Active use can mean a purchase, transfer, login, bill payment, deposit, revenue event, or any company-defined activity.

Nu defines a monthly active customer as a customer who generated revenue in the last 30 calendar days of the measurement date. It reported 94.9 million monthly active customers and an 83% activity rate at the end of 2024. The definition is company-specific: it is useful because it is disclosed, but it should not be silently substituted for a debit transaction or a funded account.

The CFPB's overdraft core-processors report uses a different operational concept for an active consumer account: at least one consumer-initiated debit transaction or deposit in the preceding month, excluding an account whose only debit activity is fees. The difference shows why active-use rates need a definition before they are compared across banks.

Product use and the revenue relationship

An active customer can use one low-value feature or several products with different economics. Build a product-depth view that separates checking and savings, card spend, payments, lending, investing, insurance, subscriptions, and business services. Add transaction frequency, average balance, direct-deposit status, and share of wallet where the issuer discloses them.

Product counts are leading indicators only when the company explains the path to revenue. SoFi's 2024 Form 10-K says product growth is generally an indicator of future revenue, but is not directly correlated with revenue because members may not fully use or continue using a product and some products, such as SoFi Relay, do not directly generate revenue. That warning applies broadly: a cross-sell opportunity is not current monetization.

Use a cohort table with one row per product and columns for eligible customers, activated customers, funded customers, recurring activity, revenue basis, and avoidable cost. Keep a product that creates engagement but no direct fee visible because it may support retention or cross-sell; classify that effect as an assumption until the company provides evidence.

Map the revenue streams behind the relationship

Neobank revenue can include net interest income, interchange, account or subscription fees, payment fees, lending fees, wealth or insurance commissions, foreign-exchange spread, and partner revenue. Each stream has a different event, counterparty, cost base, and risk. A customer count is upstream of these streams; it is not a common revenue denominator.

Nu's filing says a significant portion of its revenue historically came from interchange fees on card purchases and interest on revolving or financed card balances. For 2024 it reported monthly average revenue per active customer of $11.1, but that metric is total revenue divided by average monthly active customers. It is a monetization diagnostic, not a margin or profit measure.

For each stream, identify the fee-bearing event, price or rate, principal or agent presentation, pass-through items, timing, and associated cost. Do not divide total revenue by total customers when revenue is produced by active customers, borrowers, or deposit balances with a different mix.

Funding, fees and credit exposure

Deposits can supply low-cost funding, but a customer relationship does not automatically create a deposit or a usable balance. Track funded account rate, average and end-period deposits, direct-deposit share, deposit beta, insured versus uninsured balances, liquidity terms, and the assets funded by those balances. A partner bank may hold the deposit while the neobank owns the interface and some economics.

SoFi reported $26.0 billion of deposits at December 31, 2024 and described deposits, warehouse funding, equity, debt, securitizations, and other financings as funding sources. The same filing says deposit funding generally has a lower borrowing cost than warehouse financing. The business implication is conditional: the value of a customer rises through funding only when balances are durable, usable, and matched to the asset and liquidity profile.

The OCC's 2024 joint statement highlights operational, compliance, strategic, liquidity, concentration, consumer-protection, and deposit-insurance risks in third-party deposit arrangements. Include the bank and the platform in the funding map, and state which party controls servicing, ledger reconciliation, disclosures, and contingency obligations.

Read interchange and fee economics carefully

Card spend can be a valuable activity measure without being a profit measure. Interchange depends on payment network rules, merchant category, geography, card type, regulated status, partner splits, rewards, fraud, disputes, and processing costs. A debit purchase can produce interchange while also consuming support, fraud, and network capacity.

Compare purchase volume with interchange or payment revenue using the same period and scope. Then subtract rewards, network and processing fees, fraud losses, disputes, and partner shares where disclosed. A higher purchase volume per active customer may improve economics, remain neutral, or reduce contribution if rewards and risk rise faster than revenue.

The Federal Reserve's 2024 household report is useful context for payment and account behavior, but its population survey is not a neobank revenue model. Market context can describe the addressable behavior; issuer filings must establish the issuer's monetization.

Account for lending and credit-loss exposure

Lending can deepen a relationship and increase interest income, but it introduces underwriting, funding, servicing, delinquency, charge-off, and expected-credit-loss exposure. Separate customers eligible for credit from customers with an outstanding balance, and separate loan growth from risk-adjusted return. Report vintage performance, loss allowance, net charge-offs, delinquency, recoveries, and collections cost when available.

Nu's 2024 consolidated statements show credit-loss allowance expense of approximately $3.17 billion, alongside customer support and operations expense of about $604.6 million. Those figures are not a per-customer benchmark; they demonstrate why a larger customer base must be reconciled with credit and operating costs before profitability is inferred.

SoFi's 2024 annual report presents financial-services segment revenue, provision for credit losses, directly attributable expenses, and contribution profit as separate lines. Preserve those layers. A customer-growth story that uses revenue while omitting provision expense can mistake credit expansion for durable earnings.

Separate interest income from fee revenue

Interest income is driven by earning assets, rates, duration, funding cost, and credit performance. Fee revenue is driven by activity, price, customer mix, and contractual sharing. Net interest income is not the same as gross interest income, and neither is the same as operating profit.

Build a bridge with interest income, interest expense, fee and commission income, transaction costs, expected credit losses, operating expenses, taxes, and any management adjustments. State whether the bridge is consolidated, segment-level, or a contribution view. Keep the accounting presentation and the management presentation side by side when they differ.

The FDIC Quarterly Banking Profile and its Graph Book provide the banking industry's standard lenses for net income, net interest margin, return on assets, funding cost, noninterest expense, efficiency, and credit performance. These are comparison frameworks, not proof that a particular neobank has bank-like economics.

Bank and platform responsibilities

“Neobank” describes a customer-facing model, not one legal structure. A company may hold a charter, operate through a sponsor bank, use multiple banks, or combine banking with a technology platform. The customer may see one brand while deposits, lending, card issuing, compliance, and settlement sit across several entities.

The OCC request for information on bank-fintech arrangements notes that rapid deposit growth tied to a bank-fintech arrangement can create funds-management, liquidity, interest-rate, credit, operational, and nested-relationship risks. Map contractual roles before attributing deposit economics, regulatory costs, or customer ownership to the branded app.

For a buyer comparison, record charter owner, deposit holder, processor, lender of record, servicing party, compliance owner, and termination or migration rights. A customer count that spans partner products may be useful for reach while being unsuitable for the neobank's standalone profit denominator.

Cost to serve and cohort durability

Customer economics need both acquisition and ongoing service cost. Track paid and organic acquisition, verification, onboarding, card issuance, incentives, fraud review, customer support, disputes, collections, cloud and data processing, compliance, and partner fees. Identify which costs are variable with active customers, funded balances, transactions, or loan accounts.

Nu reports customer support and operations, general and administrative, and marketing expenses separately. Its 20-F also describes a lifetime-value-to-acquisition-cost framework. Those disclosures are useful starting points, but an issuer's LTV/CAC claim remains a management estimate; test the cohort definition, payback window, contribution margin, churn treatment, and allocation of shared costs.

Do not spread total operating expense over total customers and call the result cost to serve unless the denominator and allocation are explained. A dormant account may cost little in one month but still carry compliance, data-retention, and support obligations.

Add churn, retention, and cohort age

Customer additions can mask attrition. Add gross additions, closures, reactivations, churn, cohort age, active-rate decay, funded-balance retention, and product retention. A monthly active customer may be active for one period and disappear in the next; a customer with a balance may remain profitable or become costly as credit conditions change.

Prefer cohort curves to a single lifetime average. For each acquisition month or quarter, show active customers, funded customers, revenue, direct cost, credit loss, service contacts, and cumulative contribution. Separate organic referrals from paid acquisition when possible. Retention can increase lifetime value, but only if the retained relationship continues to produce contribution after risk and service costs.

Use a survival definition that can be reproduced from the issuer's records. If a company changes its activity definition, preserve the old series and mark the break; do not splice the numbers into a false trend.

Contribution, compliance and company profit

Contribution profit can answer a product or segment question, while GAAP net income answers the consolidated accounting question. Neither should be substituted for the other. Reconcile excluded corporate costs, stock compensation, funding transfers, credit provisions, taxes, depreciation, acquisition costs, and shared technology expense.

SoFi's filing reports contribution profit for its Financial Services segment and total net revenue for the company, while its consolidated results include the full cost and accounting structure. Nu's statements likewise distinguish gross profit, operating expenses, and net income. Use the issuer's definition and period; do not compare one company's contribution profit with another company's net income.

The buyer takeaway is a two-line conclusion: what does the relationship contribute before shared costs, and what remains after the complete consolidated cost and risk stack? A large customer base can support future operating leverage while the current business remains unprofitable; both statements can be true.

Test regulation, fraud, and compliance burden

Regulatory and control costs scale with products, jurisdictions, transaction volume, fraud exposure, complaints, and partner complexity. A fast-growing account base can require more identity verification, transaction monitoring, customer support, card controls, data security, audit, and remediation. Treat these as operating requirements rather than generic overhead.

The federal banking agencies' third-party guidance emphasizes due diligence, ongoing monitoring, operational resilience, consumer protection, and contingency planning in bank-fintech relationships. A profitability model should therefore include control ownership and failure scenarios. If the filing does not disclose a cost, label the gap instead of assigning a convenient percentage.

Fraud and compliance can also change revenue quality. Reversals, blocked accounts, disputes, and refunds may reduce recognized revenue while increasing service cost. Track net revenue after those adjustments and disclose whether the activity metric counts the affected accounts.

Build a buyer-specific profitability filter

Use a decision filter that asks whether the customer count is connected to a durable economic engine. The evidence should include a defined customer denominator, active and funded cohorts, product and transaction depth, revenue by stream, funding cost, credit performance, service and acquisition cost, retention, and a reconciliation to consolidated profit.

Scroll horizontally to compare all columns.

This filter is more useful than a league table of customer counts because it makes missing evidence visible. It also allows a buyer to compare a deposit-led neobank with a lending-led or partner-platform model without pretending that the same customer label has the same economics.

How an expanding cohort can earn less

Consider a hypothetical mature cohort with 100,000 active accounts. Suppose annual revenue averages $60 per active account and direct product and service costs average $40. Contribution before shared costs is $2 million: 100,000 × ($60 − $40).

A following-period cohort of 120,000 active accounts looks stronger by count. But assume a lower-use mix reduces average revenue to $48 while direct cost remains $40. Contribution is then $960,000. The cohort is 20% larger and its simplified contribution is 52% lower. These values are invented to expose mix sensitivity, not to describe a real bank.

Acquisition spending, corporate overhead, tax, credit provisioning and regulatory capital are outside this simplified calculation. In a lending business, loss costs must be measured consistently for the relevant credit cohort. For a real comparison, decompose the averages by account age and product use; a newly acquired cohort can have temporary onboarding costs and an unproven retention curve.

The commercial test is whether the intended customer relationship produces durable contribution after its relevant costs. Compare like-age product cohorts, reconcile the bank and platform responsibilities, and connect the result to consolidated accounts. Customer growth is most informative when those economics improve alongside it.