Digital Account Opening: Your Abandonment Rate Counts Two Losses and Hides a Third
The industry figure is 3.36 applications abandoned for every account opened. That single number holds the customer who quit the form and the applicant the bank declined, and it leaves out the account that opened and never received money. Each of the three needs a different kind of work, and your own records can size all three.
Your digital account opening abandonment rate holds two separate failures and leaves out a third, and a better application form fixes only one of the three.
The short version
- The industry figure is 3.36 applications abandoned for every account opened, so roughly 23 in 100 applications become an account.
- Your own abandonment rate mixes two failures: a customer who quit the form, and an applicant your own controls declined.
- A third failure sits outside it entirely: the account that opened and then sat empty. Its application was completed, so the rate counts it as a success.
- The digital team owns the form, risk owns the declines, and retail owns the accounts that opened and stayed empty.
- You can count all three failures from records your bank already keeps: the application log, the decline records, and the funding flag on each new account.
The usual response to the abandonment figure is to shorten the application. A shorter form recovers the customers who quit partway through it, and it does nothing for the applicant you declined or the account that sat empty. Find the biggest of the three before you decide what to fix.
What the published number says
Cornerstone Advisors and Alkami report an average of 3.36 applications abandoned for every one account successfully opened. The figure comes from their 2026 Digital Banking Performance Metrics Report, published in May 2026, on 2025 data.
Out of 100 digital account opening applications started | Count |
|---|---|
Become an account | about 23 |
Do not become an account | about 77 |
The same report puts digital checking openings at 27 percent of all checking openings in 2025, up from 21 percent the year before. It also finds that 87 percent of checking accounts are tied to an active digital banking user.
A funded account is also worth more than it was two years ago. Mercer Capital's June 2026 bank watch reports that median net interest margin at the community banks in its data set, those with $100 million to $5 billion in assets, rose from 3.23 percent in the first quarter of 2024 to 3.78 percent in the fourth quarter of 2025, before easing to 3.74 percent in the first quarter of 2026. Mercer reads that rebound as having run its course, because deposit competition stays intense. So an application that never funds gives up a wider margin than it would have in early 2024.
One thing to remember. Only 23 of every 100 digital account opening applications become an account, on the Cornerstone and Alkami ratio. Each account that does open earns a wider margin than it would have in early 2024.
Three losses in one digital account opening number
The published ratio does not say how declines were counted, so treat it as a reference point. Your own rate counts applications that did not become accounts, which puts the first two losses in one number.
The loss | What happened | Who owns it | In your abandonment rate? |
|---|---|---|---|
The application that died in the form | She started, got partway, and stopped: a long form, a session lost between phone and laptop, a step that sent her to a branch | Digital channel | Yes |
The application your bank stopped | Fraud, identity or deposit screening declined it; some were fraud, some were a real customer whose address on file trailed her driver's license | Risk | Usually, depending on how your bank counts declines |
The account that opened and never funded | Application completed, screening passed, account exists, money never arrived | Retail, or whoever carries the deposit number | No |
Your account opening platform records a reason for every decline. The single reported rate gives you a count of lost applications without those reasons.
The third loss sits between opened and funded. A bank that requires funding inside the application moves this loss into the form. There it counts as abandonment. The retail banking page sets out those steps.
Banks miss this loss because the form funnel and the deposit number sit with different teams, and neither report has a line for it. The customer finished the form, passed screening, and received an account number, so the digital team's funnel records a win and closes her file. Her account then sits in the core at a zero balance until deposit operations closes it past the cutoff. Retail reads the deposit number in aggregate. Unless somebody owns that handoff, the lost conversion stays unreported.
Empty new accounts fall into two groups. In the first, she tried to fund and failed. The record shows a micro-deposit verification that failed, a transfer that came back with a return code, or a hold the customer did not expect. Those are logged events. Pull them out first. They tried to pay you and your funding flow stopped them. Repairing that flow is what recovers them.
In the second, she skipped funding or gave up after one attempt. For them the record shows an open date, a balance of zero and nothing in between, and only the customer can tell you why she stopped.
The answers usually fall into a short list:
- a document she could not find or could not produce
- a transfer screen she did not trust with her money
- wrong or conflicting information from a branch or the call center
- a better rate somewhere else
- a funding attempt that failed and was never retried
- a plan to fund later that never came back around
Keep the list in the customer's own words. Most customers stay silent when asked. The ones who answer supply the labels. For the silent ones, the bank matches their behavior against the behavior of the customers who answered, and estimates which reason on the list is most likely. That estimate is projected onto each silent customer. Where it is weak, the bank sends nothing. The companion piece on growing deposits from the customers you already have covers how to ask a few customers and act for all of them.
Suppose your bank starts 10,000 digital applications in a quarter at the published ratio. The figures that follow are illustrative.
Step | Count |
|---|---|
Applications started | 10,000 |
Became accounts (at the published ratio) | about 2,300 |
Did not become accounts | about 7,700 |
Of the 2,300, funded by day 30 | 1,700 |
Of the 2,300, open with no money at day 30 | 600 |
The 7,700 split between customers who quit and applicants your controls declined, with applications still in manual review set aside because they have neither failed nor finished. The 600 empty accounts split again, between the customers whose funding attempt failed and the customers who skipped funding. Only the second group is a candidate for a message.
Watch for this. Your abandonment rate is the first two losses added together, and the third is not in it. Get the three counts before you decide what to fix.
Why one fix does not cover all three
Form abandonment. Cut the number of fields. Pull data from a document scan or a credit file so she types less. Keep her session alive when she moves from phone to laptop. Remove any step that sends her to a branch. If the split shows that few customers quit the form and many applications were declined, the form is the wrong place to start.
Rejections. Customer Identification Program rules require your bank to collect and verify name, date of birth, address and an ID number. Sanctions screening is mandatory as well. Your bank cannot loosen either control. Deposit screening declines an applicant for a prior history of account abuse. It runs on a separate system with its own reason codes. A decline that relies on a consumer reporting agency triggers an adverse action notice naming the agency and the applicant's dispute rights.
Above the mandatory checks sit a risk-based identity score and a fraud score, and a review path gives a caught customer a second look. Risk can retune both scores. Loosening a tunable threshold lets more real customers through and more fraud with them. Risk owns that change because it comes with change control, model validation and a fair lending review of who is newly declined and who is newly approved.
Unfunded accounts. If her funding attempt failed, the fix is operational and sits with whoever runs the funding flow, since no message clears a returned transfer. If she skipped funding, a message about her reason can move her, and retail owns that work. Before the first send, run a fraud suppression pass on every account that opened and sat empty, because some are fraud that cleared screening and went dormant.
The split runs on records you already keep. Your application platform logs the last step each incomplete application reached, and whether it was referred to manual review or pended for a missing document. Your fraud, identity and deposit screening systems record each decline with a reason code, and your core flags whether an opened account holds a balance.
Joining those records takes real work. The application platform, each screening vendor and the core keep their own identifiers and reason codes, and the application ID does not always carry through to the account record. Most banks leave the split undone until someone asks for it.
The log has the same limit for form abandonment. It shows the step she left, and a customer stopped by a failed upload and a customer stopped by a missing utility bill look identical there. Ask a few. Act for everyone. covers why the reason has to come from the customer.
The number to pull. Ask your data team for last quarter's incomplete applications, split into quit and declined, with applications still in review set aside. Add opened accounts with no balance at day 30. Those three counts show which of the three losses is the largest at your bank.
How you know the fix worked
To know whether a fix worked, you need customers who did not get it. Pick that group at random. Leave them on the old flow, or send them no message, for the length of the test. That group is your control group, and it is the baseline the fix is measured against.
When the window closes, compare the two groups. The difference is the lift. The lift is what your fix added. Without a control group you are comparing this quarter with last quarter, and every change in between, from a rate move to tax refund season, takes the credit. The companion piece on incrementality testing shows the arithmetic and how big the control group has to be.
Split the number first. It shows which loss is biggest. Ask customers why they stopped, because their answers decide what you say to the ones who skipped funding. Run the control group from the day the change goes live, so the test can tell you whether you aimed at the right loss.
Before you start. Agree in writing how the result will be measured, and keep a random control group for the length of the test. Write down the smallest lift the test can detect before anything goes live.
What to pull from your own systems
- Last quarter's incomplete digital applications, split between no decline record and a decline record, with applications still in manual review or pended at quarter end set aside from both.
- The last step reached for each incomplete application with no decline record.
- Each decline by reason code across fraud, identity and deposit screening, and how many of those applicants came back through a review path.
- Accounts opened last quarter with no balance at day 30, as a count and as a share of accounts opened.
- The measurement plan for any change you make, naming the control group, its size and the test window.
Frequently asked questions
What is a normal digital account opening abandonment rate?
Cornerstone Advisors and Alkami's 2026 Digital Banking Performance Metrics Report, published in May 2026, reports an average of 3.36 applications abandoned for each account successfully opened. That works out to about 77 percent of applications never becoming an account. Treat it as a reference point, since the report does not say how declines were handled. A bank that counts its declines is not comparable with one that leaves them out.
Why do customers abandon account opening applications?
The log shows the last screen they touched. You get the reason by asking the customer soon after she stops and sorting the answers into a short list kept in her own words. Most customers stay silent, so for them the bank estimates the most likely reason from how they behaved, matched against the customers who answered. A weak estimate means the bank holds the message back.
Does a faster application fix unfunded accounts?
No. A customer with an open, unfunded account has already finished the application, so the work that would make the form faster happens at a step she passed weeks ago. She needs a message about why she has not funded, and that message is a retail program.
How do we tell abandonment from rejection in our own data?
Join the application platform's last-step log with the decline records from fraud, identity and deposit screening and with the core's funding flag. An incomplete application with no decline record and no review flag is abandonment, and one with a decline record is a rejection. An application still in manual review or pended for a missing document at quarter end has not failed yet, so leave it out of both counts. An opened account with no balance at day 30 is the third loss.
How do we know a new application flow actually worked?
Keep a random share of applicants on the old flow for the length of the test. Compare the completion rate of the two groups when the window closes. Last quarter is a poor control group, because rates and the season both changed since then.
How Numr runs this
Numr runs PXI on the customer who opened an account and skipped funding, or tried once and stopped. The bank sends Numr a weekly file. Numr asks a sample of them why they stopped, projects the reasons onto the rest, and sends every message with a random holdout from the first send. Numr reports the lift as treated versus held out, in dollars with its confidence interval, and hands the client the arithmetic behind it.
