Mapping and diagnosing the account-opening funnel
The most expensive screen in retail banking is the one asking for a photo of an ID document. Upstream of it, a weak headline costs you clicks you can buy back tomorrow. Downstream of it sits a stretch you are not free to redesign: identity verification, sanctions screening, the credit or eligibility decision, and the first deposit actually landing. This lesson stays inside that stretch, from application started to account funded, and treats it as a measurable sequence rather than a black box the compliance team owns.
The regulated middle
Two steps sit in the middle of every account-opening flow and neither is negotiable.
KYC (Know Your Customer) is legally mandated identity verification. In the US it flows from the Bank Secrecy Act and the USA PATRIOT Act; in the EU from the anti-money-laundering directives, now consolidating into a single EU AML Regulation with a new supervisor, AMLA, standing up from 2025. Underwriting is the credit decision: for a card or a loan the bank pulls credit data and decides whether to approve, and on what terms. A deposit account skips underwriting but still passes sanctions, politically-exposed-person and adverse-media screening.
You cannot growth-hack these away. What you can change is how many times an applicant is asked for the same data, whether verification happens in-session or by an email link the next day, and how fast a screening hit gets resolved. The great majority of screening alerts are false positives, usually a common name matching a list entry, and each one routed to a manual queue adds hours or days that the applicant experiences as silence.
The stages that actually exist
| Stage | What happens | What you measure |
|---|---|---|
| Application start | First field completed | Start rate from click |
| Identity submitted | Document capture or video ident completed | Submit rate, split by device |
| Screening cleared | Checks pass automatically or refer to a human | Auto-clear rate, manual queue share, hours in queue |
| Decision | Approve, decline or refer | Approval rate on submitted applications |
| Credential live | Card provisioned to a wallet, or plastic posted | Provisioning rate, days to first use |
| Account funded | First deposit or first transaction lands | Funding rate, days to fund |
ImpressionsImpressionsThe total number of times an ad or piece of content is displayed, regardless of clicks. Each display counts as one impression, even to the same person.View full definition → and clicks sit above this table and belong to channel cost work; the denominator argument (funded accounts versus applications) is settled in the acquisition-cost lesson, and this one assumes it.
Worked example: what 10,000 starts do
Illustrative numbers for a card campaign with $50,000 of spend behind it.
- 10,000 application starts
- 55% submit a complete application with identity documents: 5,500
- 88% auto-clear screening; 660 refer to manual review
- Half of those referrals lapse or are withdrawn before a decision: 5,170 reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → a decision
- 38% approved: 1,965
- 70% funded or first-transacted within 30 days: 1,376
Divide spend by those 1,376 and you get roughly $36 against the funded denominator the acquisition-cost lesson insists on. Measured against approvals it looks like $25, which is the number that gets quoted in the channel review and the number that hides the problem.
Look at the 330 applications that died in the manual queue. They are not declines. They rarely appear in a decline report at all, because no decision was ever recorded. They cost the full click price, they passed the form, and in most banks' reporting they simply vanish. Instrument the queue or you will keep optimising the parts of the funnelfunnelThe customer journey from awareness to purchase, typically Awareness, Interest, Consideration, Decision, Action, with prospects narrowing at each stage.View full definition → you can already see.
Where the friction actually lives
Start to submitted is the first crater: document upload failures, glare on a passport photo, liveness checks that fail on the first attempt, and requests to re-key data the bank already holds. Mobile is worse than desktop on every one of these, and mobile is where the traffic is.
The second crater is between approval and money. An approved account that is never funded earns nothing and still costs servicing and reporting. Where funding depends on micro-deposit verification, the applicant is asked to wait one to three business days for two small credits and then come back and type the amounts in. Every day in that gap is drop-off you paid for.
Apple Card shows the other end of that spectrum. Applied for inside Wallet with a decision in minutes and, on approval, provisioned to Apple Pay straight away, so the approval-to-first-transaction gap collapses from the week or so a posted card takes to roughly the length of the walk to a coffee shop. That design choice is a funnel decision, not a product flourish.
Diagnosing: is it friction, fit, or hand-off?
- Low completion, healthy approvals among completers: a UX and verification problem. Fix the capture flow.
- Healthy completion, low approvals: a targeting problem. Your audience does not match the credit box, and no form change will help.
- Both healthy, weak funding: a hand-off problem. The gap between yes and money is too long or too manual.
🎬 [VIDEO: "How Banks Reduce Onboarding Drop-Off" - youtube.com - a short walkthrough of digital account-opening friction points and KYC abandonment]
Instrumenting a flow you are not allowed to log
Client-side tags stop firing the moment the applicant is redirected to a verification vendor's hosted page, which is exactly where the drop-off is. Fire server-side events on state changes instead, each carrying an application ID that survives the redirect, so the return leg stitches back to the original session.
Keep the payloads clean. Send `id_document_rejected` with a reason code, never the document, the ID number or the selfie. Segment, a customer data platformcustomer data platformSoftware that unifies customer data from every source into one persistent profile that marketing, sales and service teams can act on.View full definition → and therefore an interested party here, is a common collection layer for this pattern; the discipline holds whatever tool you use, and a data-protection review of the event schemaschemaA schema is the formal blueprint that defines how data is structured, named, typed, and related within a database, file, or message.View full definition → is cheaper before launch than after.
Count unique applicants, not sessions. Someone who retries on a phone, then a laptop, then the phone again reads as three abandonments and one success, which quietly inflates both your drop-off rate and your start volume.
One more instrumentation break worth knowing: Apple's App Tracking Transparency, introduced with iOS 14.5 in April 2021, cut the join between ad click and application for a lot of mobile campaigns. When your funnel starts in an ad network and finishes on a bank server, expect the two halves to disagree, and decide in advance which one is the system of record.
When smoothing the funnel costs you the growth
N26 built its proposition on speed: sign up in minutes with a video identification call rather than a branch visit. The regulator then set the ceiling. BaFin fined the bank €4.25 million in 2021 over late suspicious activity reports, appointed a special monitor, and in November 2021 capped new customer onboarding at 50,000 a month. The cap was eased to 60,000 in 2023 and lifted in June 2024, and a further fine of roughly €9 million followed in 2024 over delayed filings.
The second-order effect matters more than the fines. For over two years, the binding constraint on that funnel was not conversion rate, it was a quota. Paid acquisitionPaid acquisitionVisitors arriving via paid ads or sponsored placements, where you pay a platform to display your message rather than earning visits organically.View full definition → had to be throttled, waiting lists appeared, and marketing's optimisation target flipped from volume to the quality mix inside a fixed monthly allowance. Any onboarding improvement that shaves friction by weakening a control can be reversed by a supervisor at a speed no growth team can match.
Mortgage: same wall, longer clock
Same shape, stretched over weeks, with a much heavier verification load: income documents, appraisal, title. Rate shoppers pre-qualify and never apply, and applicants walk between rate lock and close if a competitor undercuts. Track the funded loan rather than the lead, and set 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 → windows long enough (60 to 90 days is common) to catch the whole thing.
Your own numbers first
Published digital account-opening abandonment figures range from around 40% to 70% of started applications depending on who counted and what they counted, and the benchmarks lesson covers when a published figure is comparable to yours at all. For ongoing coverage, see the Financial Brand. Until then, your best reference point is your own best channel over trailing quarters.
A practical diagnostic checklist
- Log every state change server-side, including referrals into and out of manual review.
- Split identity-step abandonment by device and by document type.
- Report the manual queue as a funnel stage with its own volume and its own clock.
- Segment approval rate by acquisition channel to catch fit problems early.
- Measure days from approval to first funding, and treat any median above one day as a defect.
Key takeaways
- The middle of a banking funnel is regulated. You smooth it, you never remove it, and most drop-off clusters there.
- Applications that lapse in a manual review queue are invisible in decline reporting and are pure paid-media waste.
- Friction, fit and hand-off failures produce different-looking funnels and demand opposite fixes.
- Instrument server-side with reason codes, not payloads, and count applicants rather than sessions.
- N26 spent over two years with its onboarding volume capped by BaFin: the ceiling on an account-opening funnel can be regulatory rather than behavioural.