# Customer Acquisition in Financial Services: Cheap Installs Can Cost More

> Define the financial relationship your acquisition budget buys, then compare channels on the same customer event and clock.

- Author: Rishikesh Ranjan · Published: Oct 8, 2026
- Type: Essay
- Tags: Acquisition, Onboarding, Metrics, Frameworks
- Growth levers: Acquisition (primary), also Activation
- ~1899 words

---

A fintech campaign can win the install report and lose the customer budget. If one channel brings twice as many app installs but fewer people fund an account, the acquisition team has two different answers to the same question: which channel deserves the next dollar?

Customer acquisition in financial services needs a definition of the relationship being acquired. An install tells you that an app reached a device. It doesn't establish that a new person became eligible, received a usable account, or completed the financial action they came for. Keep install metrics for diagnosing distribution, then judge customer acquisition against a deduplicated business event within a stated window.

This matters to a growth lead buying mobile reach, but the definition belongs to product and finance too. A savings account, a loan and a payment service earn value differently. Calling all their downloads ‘new customers’ hides those differences precisely when the team needs them to decide how much acquisition it can afford.

![Conceptual illustration of a phone download separated from a funded relationship by identity, eligibility and funding checks.](https://www.productgrowth.blog/media/posts/fintech-installs-new-customers/00-install-to-relationship.webp)
*A download starts a possible relationship; the customer event needs its own definition.*

## Separate access, first value and a continuing relationship

For acquisition reporting, it helps to distinguish a prospect from someone who has received the product's first useful outcome. A person can install, register, submit information, pass a check, and still leave with no usable balance or completed payment. Those intermediate states deserve measurement. They answer different questions from the one finance asks about the cost of winning a customer.

‘Customer’ can also have a [contractual or regulatory meaning](https://www.consumerfinance.gov/rules-policy/regulations/1016/4/) before the first transaction. Keep that official definition intact. The growth metric should add a qualifier such as ‘new funded customer’ or ‘new transacting customer’ rather than pretending that a marketing team can redefine the legal relationship. Record both if the difference matters to the business. For example, the U.S. Regulation P customer-relationship examples include opening a credit-card account or executing a deposit-account contract.

[Mastercard's lifecycle guidance](https://www.mastercard.com/us/en/news-and-trends/Insights/2024/the-guide-to-lifecycle-marketing-for-financial-institutions.html) separates acquisition, onboarding, usage and retention for cardholders. That distinction leaves room for a legitimate account-opening measure while still asking whether card activation and use followed. A signup can be a completed acquisition milestone in one report and an unfinished economic relationship in another.

The useful disagreement is therefore about the label and decision. A distribution team may need cost per install to compare creative. A product team may need verification completion to find an onboarding problem. A finance team needs a customer denominator connected to its payback model. Keep those views together so a cheap early event doesn't inherit the value of a later one.

## The industry data shows why registration and use need separate counts

GSMA's [State of the Industry Report on Mobile Money 2026](https://www.gsma.com/sotir/wp-content/plugins/plugin_gsma_sotir/reports/The-State-of-the-Industry-Report-2026_English.pdf) reports 2.3 billion registered accounts globally in 2025 and 593 million active 30-day accounts. Its reported monthly activity rate was 25.7%. These estimates concern mobile-money services, a category that excludes conventional mobile banking used as another account-access channel.

![Common-zero bars show GSMA's 2025 global registered mobile-money accounts at 2.3 billion and active 30-day accounts at 593 million. Reported monthly activity share: 25.7%.](https://www.productgrowth.blog/media/posts/fintech-installs-new-customers/01-registration-and-use.webp)
*GSMA, 2025 global account estimates. Registration and recent activity measure different states.*

| Measure | 2025 global estimate |
| --- | --- |
| Registered mobile-money accounts | 2.3 billion |
| Active 30-day mobile-money accounts | 593 million |
| Reported monthly activity rate | 25.7% |
*GSMA Global Adoption Survey 2025 and estimates, reported in the 2026 industry report.*

These are account totals, not unique people or a cohort of recent installs. Treating the activity share as an install-to-customer conversion rate would mix populations and timeframes. Nor does monthly inactivity prove churn: a useful account might serve an occasional need. The narrower lesson is that registration and recent use are different quantities.

That distinction gives a fintech team a better question for its own data: what evidence would show that this newly acquired person received the promised service? A global activity figure can't choose the endpoint or target for a local deposit, credit or insurance campaign.

## Choose a customer event that fits the product

A first-value definition should describe a completed customer outcome that your business can confirm. ‘Funding started’ may mean that a transfer was requested; ‘funds received’ means something different. A loan approval may authorize an offer without putting money in the borrower's hands. Agree on which state the budget decision requires, then name it directly.

| Product job | Proposed first-value event | Separate later quality check |
| --- | --- | --- |
| Deposit account | First qualifying customer-funded balance received | Retained balance and funding cost |
| Credit | Approved loan disbursed | Repayment performance and credit losses |
| Payments or wallet | First qualifying transfer or payment completed | Repeat use, reversals and fraud |
| Investing | First qualifying funded investment completed | Retained assets and servicing cost |
| Insurance | Policy issued with coverage effective and required payment received | Cancellations and contribution over the policy period |
*Proposed acquisition metric definitions. Product terms, eligibility and business economics determine the exact event.*

‘Qualifying’ needs a concrete meaning. Decide whether promotional balances count, how reversals affect the result, and whether a person with an existing relationship belongs in a new-customer denominator. An event that merely moves company-funded reward money into an empty account may not answer the same question as a customer's own deposit.

For example, a deposit product might count a person once when cleared customer funds arrive and exclude internal test accounts. It might separately inspect whether the balance persists. Those are proposed definitions, not a universal minimum deposit or retention rule. Changing the qualifying amount halfway through a campaign changes the denominator, so the apparent improvement needs to be restated on a consistent basis.

Once the endpoint is clear, the cost definition becomes easier to defend. The [customer acquisition cost calculator](https://www.productgrowth.blog/calculators/customer-acquisition-cost-cac) explains the numerator and denominator for a broader acquisition-cost view. A channel report using media spend alone should say ‘media cost per funded customer’ so it can be compared without quietly dropping creative, incentive or operations costs.

## The cheaper install can produce the more expensive customer

Consider an illustrative deposit-account campaign with two channels. Each spends $12,000 on media. Channel A records 4,000 installs and 80 distinct new customers who first fund an account. Channel B records 2,000 installs and 120 first-funded customers. Each customer belongs to that channel's install cohort, has the same 30-day funding window, and is counted once; existing customers and test accounts are excluded.

| Same-window measure | Channel A | Channel B |
| --- | --- | --- |
| Media spend | $12,000 | $12,000 |
| App installs | 4,000 | 2,000 |
| New first-funded customers | 80 | 120 |
| Media cost per install | $3 | $6 |
| Media cost per first-funded customer | $150 | $100 |
*Illustrative scenario: equal media spend, complete 30-day windows and deduplicated new first-funded customers. Costs exclude other acquisition expenses.*

![Illustrative paired panels compare channels A and B: installs 4,000 versus 2,000, new first-funded customers 80 versus 120, media cost per install 3 versus 6 dollars, and media cost per first-funded customer 150 versus 100 dollars.](https://www.productgrowth.blog/media/posts/fintech-installs-new-customers/02-channel-denominators.webp)
*Illustrative equal-spend comparison. Each panel uses a common zero baseline; counts and dollar costs have separate scales.*

Divide the same spend by each relevant count. A's media cost per install is $12,000 / 4,000 = $3; B's is $12,000 / 2,000 = $6. For the funded relationship, A costs $12,000 / 80 = $150 and B costs $12,000 / 120 = $100. The install report favors A. On this first-funded customer measure, B buys the outcome at a lower media cost.

The arithmetic doesn't explain the difference. B might reach people with a stronger reason to fund, show a clearer offer, or simply benefit from a different audience mix. The result supports investigating those possibilities; it doesn't identify which mechanism caused them. Compare eligibility, processing delays and the experience after signup before crediting the ad alone.

Even B isn't automatically ready to scale. Its funded customers could withdraw quickly or require expensive support. A larger spend could reach a different audience. The scenario establishes why the denominator matters, while the real budget decision still needs the next cohort and its economic quality.

## A new app event can belong to an existing customer

[Google Analytics documents first_open](https://support.google.com/analytics/answer/9234069?hl=en) as the first app launch after installation or reinstallation, rather than the download itself. That event definition shows why an analytics label can't settle whether the business acquired a new customer. A returning customer can generate a fresh app lifecycle event while keeping the same financial relationship.

Use the business customer record to answer newness where you have a permitted, reliable match. An existing web customer who installs the app represents adoption of another channel. A cardholder who opens a savings account is new to that product but may already be a customer of the institution. A returning wallet user may be reactivated. Each can be valuable without being reported as a net-new relationship.

Keep new-to-firm, new-to-product and reactivated populations separate in acquisition reporting. Mixing them can give a channel credit for people who already knew the provider and needed little persuasion. You can then compare the cost of each job honestly instead of making the new-customer cost look lower through a change of labels.

An exact person-level join won't always be available. Use only the identity and attribution data your permissions and systems support. If an aggregate report includes unresolved matches, retain that uncertainty rather than declaring every install a distinct new person. An unmatched record is a measurement status; it doesn't by itself prove a duplicate, a fraudulent user or an acquisition failure.

## The observation window changes the answer

A customer who installed yesterday has had less time to fund than one who installed last month. Comparing both as if they had an equal opportunity to reach the endpoint can make a recent channel look weak just because its outcomes are pending. Choose a window suited to the product's actual decision and settlement lag, then compare acquisition cohorts at the same age.

In the illustrative campaign, 30 days means 30 days from each person's install, not funding completed before the calendar month's last day. The entire cohort needs that observation time before its final comparison. A weekly review can show provisional figures, but it should carry the cohort age and pending count alongside them.

Separate approval declines, unfinished applications and pending reviews where those states are available. Each calls for a different explanation. An eligibility mismatch may require a clearer offer or a change in audience. An upload failure may require a product repair. A pending review may need time or operational capacity. Combining all three into abandonment can send the team toward a fix that doesn't address the cause.

The same care applies to acquisition cost. A calendar-period report can be useful for financial planning, but it isn't automatically a channel cohort comparison. State which costs and customers belong together, and keep the convention stable. If you revise the window or attribution rule, show the previous cohort under the revised definition before claiming progress.

## Keep quality beside the customer count

A completed first financial action is a useful endpoint, but it isn't profit. Ask finance which part of a funded balance or financial transaction contributes revenue and margin before putting it into a payback calculation. Work with finance on the contribution expected from the relationship, including the costs and risks relevant to the product. A lender needs credit performance; a payment service needs to understand reversals and fraud.

Use a later quality measure appropriate to the customer job. Repeated payment can show continuing use of a wallet, while an insurance policy may deliver value without a customer opening the app every month. A deposit relationship can also remain useful with few screen visits. Forcing every product toward daily app activity can turn the measurement target away from the financial service.

This leaves a sensible role for cheap install tests. They can tell you whether a message reaches an audience and persuades people to try the app before enough deeper outcomes arrive. Treat an early event as a provisional signal, and inspect how it connects to the customer endpoint once the cohort matures. Cost per install earns a place in the diagnosis without carrying the whole acquisition claim.

Finally, campaign attribution and incremental customer acquisition answer different questions. An attributed funding event records credit under a measurement rule; it doesn't show that the person would have remained unfunded without the campaign. A suitable holdout or another credible comparison is needed when the decision requires that causal answer. Keep platform-attributed outcomes distinct from proven additional customers.

## Write the denominator before approving the budget

Before the next campaign review, finish this sentence: ‘We count a new customer once when a person with no prior qualifying relationship completes this confirmed outcome within this window.’ Add the product-specific exclusions and cost scope. If the team can't finish the sentence, an install forecast is still describing app distribution rather than the customer outcome the budget is meant to buy.

Then put the install report next to the mature customer cohort. A channel can be strong at distribution and weak at funding, or slower to produce a relationship that is worth keeping. The next dollar should follow the outcome and economics the team agreed to measure, with the early events explaining where the path needs work.

**Next job: Make the acquisition promise match the first action.** Use the fintech acquisition playbook to carry fees, eligibility and verification expectations from the campaign into a usable account. [Continue](https://www.productgrowth.blog/p/fintech-mobile-app-user-acquisition)

---

All posts: https://www.productgrowth.blog/archive · Site: https://www.productgrowth.blog