Consent Architecture: Designing Bank-Connection Flows People Actually Complete
Half of bank-connection journeys went uncompleted, and 77% of people didn't feel informed by consent terms. A playbook for intake's highest-stakes screen.
Alfred BEditorial Reviews
The bank-connection step is the highest-stakes screen in a lending application. It asks for more than any other field, it arrives when the applicant is already invested, and it is where a meaningful share of otherwise willing people stop.
The evidence says most of that loss is design rather than refusal. Open Banking Limited found UK authentication completion running near 50% in 2019, with complex multi-step journeys approaching zero. The Financial Services Consumer Panel found 45.2% of consumers do not read consent terms at all and 77% did not feel informed by them. People are not refusing to share. They are failing to get through, and agreeing to things they have not understood.
This is a playbook for the screen itself.
How bad is the drop-off?
Open Banking Limited's September 2019 analysis of the UK's 90-day re-authentication requirement remains the most detailed public measurement of consent friction. It found approximately 50% average completion across all customer authentications, and journeys with many steps approaching 0% conversion.
The re-consent requirement itself carried a recurring cost: roughly 13% customer attrition every 90 days, and around 30% of customer connections broken. Among customers who did complete re-authentication, 30.4% found the journey too long and too obstructive, and 12.9% said they were less likely to keep using the service.
Two caveats matter. This is UK data, and it is from 2019, before the Financial Conduct Authority removed the 90-day bank re-authentication requirement in November 2021 and shifted reconfirmation to the third party. So treat these as evidence of what consent friction costs, not as current benchmarks. The mechanism has not changed even where the rule has.
Do people understand what they are agreeing to?
Largely not, and this is the finding that should shape the copy.
The Financial Services Consumer Panel published research in March 2018, conducted by Dr Edgar Whitley and Dr Roser Pujadas of the London School of Economics, surveying 191 participants on how they consent to share data. It found 45.2% do not read terms and conditions at all, a further 40.9% only skim-read them, and 77% did not feel informed when reading them.
The same research asked what would improve consent terms. Participants ranked shorter text length first, ahead of highlighting risks at the start.
That ordering is the practical instruction. The instinct when a screen underperforms is to add explanation. The people using it are asking for less.
What does the standard actually prescribe?
Open Banking Limited's Customer Experience Guidelines, currently at version 4.0.1, set out a three-step consent model and name the specific failure patterns directly: superfluous information, poor or confusing choice of words, repetition, large amounts of text, too many steps.
The guidelines also contain the single best line written about this screen: the journey should feel like an experience and not a contract.
On mechanics, the Open Banking Implementation Entity reported in August 2020 a direct correlation between making app-based authentication available and higher consent success rates. Where an applicant can authenticate in their banking app rather than by typing credentials into a browser, more of them finish.
The willingness ceiling
There is a limit that no design work removes, and it is worth knowing before setting targets.
The Bank for International Settlements published research in March 2026 finding that 27% of people indicated they would share financial data to obtain improved loan terms. That is a low ceiling, and it means a share of your applicants will decline a connection regardless of how good the screen is.
Which makes the fallback path a design requirement rather than an afterthought. A flow with no document route loses those applicants entirely.
It also reframes what a good completion rate looks like. If roughly three in four people are unenthusiastic about sharing bank data for a better rate, a connection step converting well above that is doing unusually well, and one converting far below it has a fixable design problem rather than a market problem. Knowing which of the two you have is the whole point of measuring the step on its own.
The playbook
1. Cut the text before adding any. Consumers ranked shorter text as the top improvement. Every sentence on a consent screen should survive a deletion test: remove it, and if the screen is still clear and still accurate, leave it out.
- State three things plainly: what data is accessed, what it is used for, and for how long. Purpose and duration are what turn a credential handover into an informed decision.
- Reduce step count above everything else. Open Banking Limited's finding that complex multi-step journeys approached zero conversion is the most dramatic number in the whole dataset.
- Support app-based authentication wherever the institution offers it, given the correlation with higher success rates.
- Never make the connection step the first thing an applicant sees. Ask it after they are invested, not before they understand what they are applying for.
- Build the document fallback properly, and treat its usage rate as a metric rather than a failure. With willingness capped around 27% in the BIS research, some share of applicants will always take it.
- Instrument the step separately. Aggregate completion tells you nothing about where inside the connection people leave.
What the research says to do, and what it says to stop doing:
| Do | Stop |
|---|---|
| Shorten the text | Adding explanation when completion drops |
| Name purpose and duration | Generic "we take privacy seriously" copy |
| Minimize steps | Multi-screen consent journeys |
| Offer app-based authentication | Credential entry as the only route |
| Build a real fallback | Treating non-connectors as lost |
| Measure the step in isolation | Reading it off funnel-level numbers |
What we don't know
No Canadian data exists on bank-connection consent completion or drop-off. Every figure above is UK.
There is also no published controlled experiment on consent screen copy, layout or permission granularity from a regulator or academic body. The vendor A/B results circulating on this topic are not independently verifiable, and we have not cited them. The design principles above rest on measured friction and stated preference, which is weaker evidence than a controlled trial and stronger than instinct.
Current completion benchmarks are also missing. The roughly 50% figure is from 2019 and predates a rule change that removed part of the friction it measured.
Common questions
Why do applicants abandon bank-connection steps?
Open Banking Limited found completion near 50% in 2019, with multi-step journeys approaching zero, pointing to journey complexity rather than unwillingness.
Do people read consent terms?
Mostly not. The Financial Services Consumer Panel found 45.2% do not read them at all, 40.9% skim, and 77% did not feel informed.
What improves consent completion most?
Shorter text, fewer steps, and app-based authentication, which Open Banking Limited correlated directly with higher consent success rates.
How many people will share bank data at all?
Bank for International Settlements research found 27% would share data to obtain improved loan terms, which is why a document fallback is required.
Should the bank connection come early in a flow?
Placing it before an applicant understands what they are applying for asks for the most at the point of least investment.
Carousel's bank-connection step is built consent-first, with a document fallback for applicants who decline. See how verification fits your flow
