The friction audit: finding and removing conversion drag

Most conversion problems are not persuasion problems. The visitor arrived. They were interested enough to scroll, to consider, maybe to hover over the CTA. Then something stopped them: not a missing benefit or a weak headline, but a sense that the next step was harder than it was worth.

That is friction. And unlike traffic or trust, friction is almost entirely within your control. A friction audit is the structured process of walking your own funnel as a suspicious first-time stranger, cataloguing every point of resistance, and making deliberate choices about what to remove, reduce, or keep. It is one of the highest-leverage CRO activities available, because it surfaces real drag without requiring a single A/B test to diagnose.

What friction actually is #

Friction is any force that increases the perceived cost of completing an action. It is not just slow load times or long forms, though both qualify. The cost comes in three distinct flavours, and naming them is what turns a vague “this feels clunky” into a fixable list.

Effort friction is physical or cognitive work: fields to fill, steps to complete, decisions to make, copy to decipher. Each click and keystroke is tiny on its own. Cumulatively, they compound into abandonment.

Anxiety friction is doubt and perceived risk: “Is this secure? Will they spam me? What am I signing up for? Can I cancel?” The visitor rarely voices these, but they hesitate because of them. Unresolved anxiety kills conversions even when the offer is strong.

Confusion friction is ambiguity: unclear labels, contradictory messages, unexpected flow, jargon the reader never agreed to. When someone cannot parse what you are asking or what happens next, the safe response is to stop.

A useful mental model: friction is the gap between what a visitor wants to do and how hard it feels to do it. Narrow that gap, and conversion rate follows.

The walkthrough method #

The audit begins with a disciplined walkthrough: not from the perspective of someone who built the funnel, but from someone who has never seen it.

Start at the top: the first touchpoint, the landing page, the ad-click destination. Pretend you know nothing about the product. Then work forward through every step until the conversion is complete, or until you give up, which is itself the most useful data point you will collect.

At each step, ask three questions and write down every answer.

Quantity matters here. You want an exhaustive raw inventory before you start judging: every field, every click, every question the page raises but does not answer. Editing comes later; capture comes first.

Once you have a walkthrough list, layer in your behavioural data. Session recordings and heatmaps show where real users pause, rage-click, scroll back, or leave, letting you pattern-match your inventory against observed behaviour. Session replay is especially good at catching confusion friction a solo walkthrough misses, because you watch someone encounter the page genuinely fresh. A heatmap confirms whether the friction you flagged is where attention actually stalls.

Categorising what you find #

With a raw list in hand, sort each item by friction type, then add a second dimension: is it on a critical path or off it?

Critical-path friction sits directly between the visitor and the conversion: a required form field, a forced account-creation step, an ambiguous CTA. Off-path friction (a cluttered sidebar, a tempting link to the blog, a secondary modal) bleeds attention and adds noise but does not technically block the action.

That gives you a simple matrix. Critical-path friction is where you focus first; off-path friction is worth cleaning up but rarely produces a dramatic lift on its own.

Every field is a question you are asking your visitor. Every question has to earn its place.

Prioritising removals #

Not all friction is equally costly, and not all of it is equally easy to remove. After categorising, score each item on two axes: how much drag does it likely cause, and how hard is it to fix?

High-drag, easy-to-fix items go first. Imagine a SaaS sign-up form that requires a phone number nobody ever calls. Deleting that field is trivial and almost certainly reduces drop-off. Or a checkout that forces account creation before purchase: research from the Baymard Institute consistently ranks forced account creation among the top causes of checkout abandonment. These are your quick wins.

High-drag, hard-to-fix items go on a roadmap. Restructuring a multi-step wizard, redesigning a pricing page, or rebuilding onboarding is worth doing, but in a sequenced programme, not a Friday-afternoon sprint.

Low-drag items, regardless of effort, are low priority. The temptation to polish is real; the compounding returns live in the high-impact work. If you already run a scoring framework, friction items slot straight into ICE alongside your hypothesis-driven tests.

Good friction: what not to remove #

Here is where most friction audits go wrong. The goal is not zero friction. It is the right friction.

Some friction is genuinely useful. It qualifies intent, filters out poor-fit visitors, and protects your downstream economics. Strip it blindly and you raise your conversion rate while lowering the quality of everything that converts.

A B2B SaaS tool asking for company size during sign-up is not purely a tax. It is qualifying data that routes the user to the right onboarding and helps prioritise follow-up. Removing it saves the visitor thirty seconds but may cost the business real pipeline clarity.

A booking flow that asks for a card before a free trial carries more friction than one that does not. It also produces users with far higher activation intent. The right answer depends on your model and the lead quality you need, not on friction minimisation for its own sake.

The test for good friction: does this step serve the visitor as much as it serves the business? A long intake form for a complex professional service is appropriate: the friction signals seriousness and sets expectations. The same form for a $9-a-month consumer tool is a conversion killer.

You can settle these debates with evidence. Place a short on-site survey right after a friction point: ask those who completed it whether it felt reasonable, and ask those who nearly bailed what gave them pause. The answers usually end the internal argument faster than any opinion.

Anxiety: the friction that hides #

Effort and confusion friction are visible: you can point at a twelve-field form or a baffling label. Anxiety friction is subtler, and routinely underestimated.

A visitor on a pricing page can be fully sold and still not convert, because the page left a doubt unresolved. “What if I want to cancel?” No answer. “Is my data safe?” No badge, no policy link nearby. “Is this company even real?” Thin footer, no social proof.

The Baymard Institute’s checkout research (and the same pattern holds across SaaS and lead-gen) shows trust and security signals directly influence whether a visitor who intends to convert actually does. The intent was there. The anxiety was the blocker.

So when you review your inventory, ask explicitly: what questions does this page raise that it never answers? Then commit to answering them in the copy, in the design, or in proximity to where the doubt arises. A guarantee beside the submit button outperforms the same guarantee buried three scrolls down.

Turning the audit into a programme #

A one-time friction audit is useful. A recurring one compounds.

Set a cadence (quarterly suits most teams) to re-walk your critical conversion paths. As you ship features, change copy, and run experiments, new friction accumulates. What feels obvious to the team that built the latest flow often baffles a first-time visitor.

The strongest conversion programmes alternate between two modes: generating hypotheses from data (recordings, heatmaps, surveys) and generating them from structured walkthroughs like this one. The friction audit is the structured complement to the data-driven work; neither alone is enough. It also feeds the wider CRO process as a steady source of high-confidence test ideas.

Your funnel is never finished. Expectations evolve, competitors raise the bar, and new traffic sources arrive with different mental models. The friction that was acceptable last year may be disqualifying this year. Keep walking the funnel. Keep asking what you are making people pay to convert. Convert more, guess less.

Frequently asked questions #

What is the difference between a friction audit and a heuristic UX review?

A heuristic review checks a page against general usability principles. A friction audit is narrower and outcome-focused: you walk the actual conversion path as a first-time user and catalogue every point of effort, anxiety, and confusion between the visitor and the goal. It is scoped to one thing (what is making conversion feel costly), which is what makes it actionable.

Do I need analytics or testing tools to run one?

No. The core method is a manual walkthrough and an honest inventory, and it costs nothing but disciplined attention. Tools make it sharper: session replay and heatmaps confirm where real users stall, and on-site surveys reveal the doubts behind anxiety friction. But the audit itself is the highest-leverage thing you can do before you ever instrument or A/B test.

How do I tell bad friction from good, qualifying friction?

Ask whether the step serves the visitor as much as the business, and whether removing it makes the experience genuinely better or just shallower. A field that routes someone to the right onboarding, or a card-on-file step that filters for real intent, earns its place. A field nobody acts on, or a step that exists out of habit, is pure tax. Cut it.

How often should I repeat the audit?

Quarterly works for most teams, plus a fresh walkthrough whenever you ship a significant change to a conversion path. New friction accumulates silently as you add features and copy, and the people who built the latest flow are the least able to see it. Treat it as a standing input to your prioritisation backlog, not a one-off project.