Onboarding Friction

Friction is any step in a digital flow that costs the user effort, attention, or time. In identity verification it is the tax paid for confidence: every check that raises assurance also gives some proportion of legitimate users a reason to stop. Managing it well means removing the friction that buys nothing, not removing all of it.

How it works

Friction is usually discussed as a single dial running from easy to secure. That framing is wrong and it produces bad decisions, because the useful distinction is not how much friction a flow has but whether each piece of it buys anything.

Necessary friction changes a decision. Capturing a document, comparing a face, asking for a second factor before a high-value transfer — each costs the user something and returns information the institution acts on.

Unnecessary friction costs the same and returns nothing. Asking for data already supplied, form fields nobody reads, an error message that does not say what to fix, a capture step that fails three times because the guidance never mentioned glare. The user experiences all of it identically; only one kind is defensible.

Deliberate friction is the case people forget. Sometimes slowing a flow is the control — a delay on a first payment to a new payee, a confirmation step before an irreversible action. Here the effort is the point, and removing it removes the protection.

The measurement problem is that friction is usually recorded in the wrong place. A user who abandons at document capture is counted as a funnel loss, not as a verification failure, so the cost lands in a report nobody in the risk team reads. Drop-off measured per step, with retry counts alongside, is what makes the true distribution visible.

Why it matters for identity verification

Verification sits at the point of maximum friction in almost every onboarding flow, which makes it the step most often blamed and the one where improvement pays best.

The asymmetry is what matters. Applicants who abandon under friction skew legitimate, because a fraudster has a payoff justifying the effort and an ordinary customer opening a current account does not. Tightening a flow therefore filters out good customers more efficiently than bad ones — the same asymmetry that makes manual review queues so expensive. A program that measures only fraud caught, and never conversion lost, is reading one side of its own ledger.

Much of what gets attributed to verification is really capture. Users do not abandon because a document check is thorough; they abandon after three failed photographs with no useful instruction. Better capture guidance and auto-capture removes far more friction than loosening a threshold, and removes none of the assurance. The same logic applies to liveness: passive methods ask nothing of the user at all, where an active challenge costs several seconds and excludes people who cannot perform it.

The goal is not a frictionless flow. It is a flow where every remaining step earns its place. Automated document and biometric verification that resolves in seconds is how the necessary checks stop feeling like an obstacle.

Necessary friction vs unnecessary friction

Most programs argue about the first column while the second quietly does the damage. Auditing a flow for effort that changes no decision is cheaper than any threshold debate and usually recovers more completions.

What reducing friction can’t do

It cannot substitute for a decision about risk appetite. Where thresholds sit is a business judgment. Removing steps without deciding how much fraud is acceptable moves the number without anyone owning the change.

It cannot fix an inaccessible flow. A step that excludes users with motor or visual impairments is not high friction, it is a barrier. Speeding it up leaves those users exactly where they were, which is why accessibility is a separate question from conversion.

It cannot help customers the method never fits. Data-based checks fail thin-file applicants regardless of how smooth the interface is. That is a method problem, and no amount of interface work reaches it.

It cannot be measured only in aggregate. A healthy overall completion rate hides a step where a specific cohort fails badly. Friction is a distribution, and programs that track a single average never see the group they are losing.

Frequently asked questions

What is friction in customer onboarding?

Any step that costs the user effort, attention, or time — forms, document capture, verification checks, waiting. It is not inherently bad; the question is whether a given step changes a decision the business acts on.

Does reducing friction increase fraud?

Only if the steps removed were doing something. Removing effort that changes no decision costs nothing in assurance. Loosening a threshold is a different action and does trade fraud against conversion.

What is positive friction?

Deliberate slowing where the delay is the control — a cooling-off period before a first payment to a new payee, or a confirmation step ahead of an irreversible action. The effort is the point rather than a cost to minimize.

Where does most onboarding friction actually come from?

Usually capture rather than verification. Users abandon after repeated failed photographs with unhelpful guidance, not because the check itself was rigorous. Better capture assistance recovers completions without weakening any control.

Related reading

• Document capture — where most avoidable friction originates
• Manual review — the queue that filters out good customers fastest
• Digital onboarding — the flow friction is measured across
• Passive liveness detection — a check that costs the user nothing