Lending journeys are leaking applicants who actually want to borrow — and what positive friction fixes

Of all the moments a customer can have with a financial services provider, a lending application is the one with the highest declared intent. The applicant has identified a need, chosen a provider, and opened the front door.

And yet lenders routinely lose most of them before the decision is made. Take the credit union sector: application-to-funded conversion rates commonly sit as low as 10–15%. Four in every five applicants — all of whom actively wanted to borrow — walk away. The same dynamic plays out across community banks, consumer finance, and SME lending, just in slightly different proportions.

This is not a demand problem. It is a journey design problem.

Friction isn't the enemy — sacrificed data quality is

The instinct to shorten lending journeys isn't wrong. Fewer fields, faster submissions, a cleaner applicant experience — all worth having where they can be had. If the same decisioning confidence can be achieved in fewer questions, that is unambiguously a better journey.

The instinct goes wrong when "fewer questions" becomes the measure of success in its own right, rather than "the right questions, yielding the data required to decision cleanly." When question count is the KPI, data quality is the variable that gives way. And when data quality gives way, the friction does not disappear — it moves. It reappears as "we need one more document" emails, manual underwriting queues, and decisions that sit for days. It reappears as applicants being declined on incomplete information when they could have been approved with the right questions asked up front. It reappears in the quiet, expensive form of every data pull run on an applicant who was never going to qualify in the first place.

The real target is not a shorter journey. It is a journey with the minimum friction required to produce clean, decisionable data — and sometimes that means asking more, not fewer, questions at the point where an extra validation step widens the funnel and eliminates downstream chasing. Positive friction, in the right place, is how data quality is preserved without the applicant feeling punished for it.

The Positive Friction Model

A better design starts from the opposite premise: some friction, deliberately introduced and sequenced, is the single biggest lever a lender has on conversion, cost, and data quality. The table below captures the shift.

Layer	Traditional journey	Positive-friction journey
1. Journey design	Linear, one-size-fits-all	Optimised around applicant intent and profile
2. Data pulls	All checks run in parallel	Sequenced — cheapest checks gate the expensive ones
3. In-journey validation	Minimised or deferred post-submit	Purposeful friction that captures everything needed first time
4. Data schema	Messy, unstructured outputs	Clean, structured — ready for auto-decisioning and AI agents

Each layer compounds on the last. Journey design determines what gets asked. Sequencing determines what gets paid for. Validation determines what gets captured. Schema determines what gets used downstream. A lender improving any one layer in isolation will see modest gains. A lender redesigning all four will change the economics of their book.

Layer 1 — Journey design

A traditional lending journey treats every applicant the same. The same questions, in the same order, regardless of product, profile, or intent. This is operationally simple and conversionally expensive.

A journey optimised around applicant intent and profile adapts. A thin-file applicant for a small consumer loan should not navigate the same journey as an established SME seeking asset finance. The order of questions, the validation steps required, and the checks run behind the scenes should be shaped by what the applicant has already told the system. Applicants feel the journey is built for them. Lenders recover the cost of every question they ask.

Layer 2 — Sequenced data pulls

Most lenders run their data checks in parallel. An application hits submit, and identity, bureau, affordability, fraud, and — increasingly — open banking pulls all fire at once. This is fast, and it is wasteful.

A sequenced approach runs the cheapest and most decisive checks first. Identity and eligibility gates cost pennies and rule out the applicants who were never going to qualify. Only when those gates are passed do the expensive pulls — full bureau reports, affordability assessments, open banking retrievals — get triggered.

The applicant experiences no additional friction; the lender pays for data it actually uses. The economics compound quickly. A full data stack on a UK unsecured application — identity, fraud, bureau, affordability, and open banking — commonly runs £8–12 when fired in parallel. A sequenced journey that filters out unqualified applicants for pennies before triggering the expensive checks typically cuts average data spend per application by 50–60%. For a lender processing 10,000 applications a month, that is in the order of £500,000 a year in data costs alone, with no change to credit policy and no downside to conversion.

Layer 3 — Positive friction

This is the counterintuitive part, and it is the one that matters most for conversion. The right validation steps, asked at the right moment, widen the funnel rather than narrow it.

Two effects are at work.

First, positive friction captures applicants who would otherwise be auto-declined on thin data. An applicant prompted to verify income, connect a bank account, or provide clarifying context moves from "insufficient information" to "assessable." Many of those applicants are good credits that a stripped-down journey would have lost.

Second, positive friction eliminates the post-submit back-and-forth that quietly kills conversion. When everything required for a decision is captured in the journey itself, there is no "we need one more document" email. There is no hand-off to an underwriter who has to chase the applicant for a payslip. If the application cannot be auto-decisioned, the file that reaches manual review is complete. The applicant hears one answer, quickly, rather than a sequence of requests that wears them down.

Applicants do not resent friction that is obviously purposeful. They resent friction that feels arbitrary, and they resent being asked for things twice.

The conversion arithmetic is striking. Credit Canary's journey implementations regularly deliver application-to-funded conversion rates of around 80% — against a traditional-journey baseline of 10–15%. This is not a marginal improvement. It is a step-change, and it transforms the economics of a lending book. Even a more modest uplift — from 15% to 25%, say — represents a 67% lift in productivity per pound of acquisition spend. A move to 80% multiplies the funded output of the same top-of-funnel many times over. The leak, once sealed, is worth more than most new-customer acquisition channels a lender could build.

Layer 4 — Clean data for AI agents

The fourth layer is where today's conversion problem meets tomorrow's infrastructure question.

Every lender is being asked what their AI agent strategy is. The honest answer, for most, is that it doesn't matter yet, because the data underneath won't support it. Agents — whether for auto-decisioning, portfolio monitoring, proactive member engagement, or collections — are only as good as the schema they sit on top of. Messy, unstructured, inconsistently-captured application data will not support useful agentic capability no matter how capable the underlying model is.

A well-designed journey produces clean, structured, validated data as a by-product of solving the conversion problem. That data is the precondition for every agentic capability lenders will want to deploy over the next five years. The lenders building on Credit Canary-grade data schemas will be ready when the agents arrive. The lenders building on whatever their current journey happens to collect will not.

This is not a future argument for the work. It is today's conversion problem and tomorrow's AI strategy, solved by the same design decision.

What the model unlocks

To make the economics concrete, take an illustrative mid-sized lender: 10,000 monthly applications, 15% conversion, £120m of annual origination, and the standard parallel-fire data stack.

Applying the Positive Friction Model to that lender typically yields:

• ~£500,000 a year in data cost savings, from sequencing cheap gates ahead of expensive pulls.
• A multi-fold increase in funded volume from the same top of funnel. Every ten-point conversion lift is worth roughly £80m in additional annual origination; moving from a 15% baseline towards the 80% ceiling Credit Canary journeys deliver takes that into the hundreds of millions — with no incremental marketing spend.
• A 30–50% reduction in manual review volume, because applications that reach underwriting arrive complete rather than requiring document-chasing.
• A clean, structured data schema that is genuinely deployable against AI agents for auto-decisioning, monitoring, and member engagement.

Three of those benefits hit the P&L this year. The fourth compounds every year after.

Model the impact on your own book

Plug in your own monthly application volume, average loan size, current and target conversion rates, and data pull economics to see the annual uplift a positive-friction journey would unlock.

The reframe

The question lending leaders have been asking is the wrong one. "How do we remove friction from our journey?" assumes that friction is the enemy. It is not. Unstructured friction, in the wrong place, is the enemy. Structured, sequenced, purposeful friction is how a lender earns an applicant's money without losing them — and how it builds the data foundation for the decade ahead.

Four in every five applicants at the sharp end of the market want to borrow from you. The journey is the reason they don't.

← Back to News & Opinion