How to Run a Conversion Audit That Finds Your Real Leaks

Here’s a scene I’ve watched play out a hundred times: a team stares at a flat sales chart, someone says “the homepage feels stale,” and suddenly three weeks disappear into a redesign that changes… nothing. Not because the team isn’t smart — because nobody checked where people were actually leaving. If you want to know how to run a conversion audit, the whole trick is this: measure where visitors leak out of your funnel first, diagnose why second, and only then fix anything. An audit is a flashlight, not a redesign. You map your money path, find the biggest leak with your own analytics, examine it through five lenses — data, a real walk-through, friction, clarity, and trust — and leave with a prioritized fix queue backed by evidence instead of hunches.

That ordering — measure, diagnose, then fix — is the entire difference between conversion work that compounds and conversion work that just burns weekends. Most teams do it backwards: they guess at fixes on whatever page someone complained about last. An audit flips the order so your effort lands where the money is. Okay, let’s walk through the whole thing, start to finish, so you can run one this week.

What is a conversion audit, and why does funnel-first beat page-first?

A conversion audit is a structured review of your funnel that identifies where and why visitors drop out before they complete the action that makes you money. The key word is funnel, not page. A page-first audit asks “is my homepage good?” A funnel-first audit asks “where do people who intend to buy actually stop?” — and those two questions produce wildly different to-do lists.

So before you open a single tool, write down your money path. For most businesses it looks like: entry → key page → action → completion. For an online course, that might be: blog post → course sales page → checkout started → purchase complete. For a SaaS app: landing page → pricing page → signup form → activated account. For a service business: Instagram profile → services page → contact form → booked call.

Here’s the part nobody tells you: you probably have more than one funnel, but you don’t have ten. Name your one to three core funnels — the paths that produce most of your revenue — and audit those. Everything else is noise for now. If you can’t name your core funnels in two minutes, that’s your first finding, and honestly it’s a valuable one.

The leak math: how to read step-to-step rates

Once your funnel is mapped, pull the numbers from your own analytics — not from a blog post about “typical” rates, yours. For each step, count how many people reached it, then divide each step by the one before it. Those step-to-step rates are where leaks hide.

Here’s a fictional worked example so you can see the shape of it. Meet Priya, who sells an online pottery course (she’s invented, but her math is how yours will look):

Funnel step	Visitors (30 days)	Step-to-step rate
Landed on site	6,000	—
Viewed course page	2,400	40%
Started checkout	240	10%
Completed purchase	132	55%

Priya’s instinct was to polish her checkout, because “checkout” sounds like where conversions live. But look at the middle: only 10% of course-page viewers start checkout at all. That’s her widest leak — the course page isn’t convincing people to act. The checkout, at 55% completion, is leaking too, but it sees a tenth of the traffic.

This is the biggest-leak-first doctrine, and it’s the heart of how to run a conversion audit: a modest improvement at your widest leak moves more revenue than perfection at a narrow one, because more people flow through it. (Those numbers are illustrative, not benchmarks — your funnel will have its own shape, and “good” is defined by your own baseline, not anyone else’s.)

The data-quality gate: is your tracking even trustworthy?

Before you believe any of those numbers, run one gate check: is the tracking itself healthy? I’ve seen teams panic over a “catastrophic drop at checkout” that turned out to be a tracking tag that never fired on the confirmation page. A broken funnel report is a false alarm that looks exactly like a five-alarm fire.

The gate is simple: complete your own funnel once, then confirm each step you took shows up in your analytics within a day. Check that your key events fire on mobile as well as desktop, and that a recent site update didn’t quietly break a tag. If the data fails the gate, fixing tracking is the first audit finding — don’t diagnose a funnel you can’t see clearly.

How to run a conversion audit: the five lenses

With your funnel mapped and your biggest leak identified, you diagnose it through five lenses. Each lens catches problems the others miss, and together they’re the spine of the audit. Work through all five for your biggest leak before moving to the next one.

Lens 1: The data lens — where and for whom does it leak?

Your overall step-to-step rate is an average, and averages hide stories. Split every funnel step by device and by traffic source before you conclude anything. The mobile-vs-desktop split is the classic reveal: a funnel that looks mildly leaky overall is often fine on desktop and disastrous on mobile, which points you at a very specific problem (and, given how much social traffic is mobile, a very common one).

Same with sources: visitors from a how-to blog post, a social profile, and a paid ad arrive with completely different intent, so a “bad” page rate might just be a mismatch between one source and one page. The rule to inherit here: segment before concluding. Never diagnose an average.

Lens 2: The walk-the-funnel lens — do it yourself, cold, on a real phone

You have founder blindness. We all do — you’ve seen your own site so many times that your brain autocompletes it. The cure is to walk your entire funnel yourself, on a real phone, over mobile data if you can, as if you’d never seen it: tap the link in your Instagram bio, read the page, find the button, fill the form, finish the purchase or signup.

The protocol is note everything: every hesitation, every surprise, every tiny annoyance gets written down, no matter how small it feels. Here’s a fictional labeled example of what a walk log looks like — this is from an imaginary walk, but yours will read eerily similarly:

• “Hero image took forever to load on 4G — stared at a blank screen for a bit.”
• “Hesitated at the price — couldn’t tell if it was one-time or monthly.”
• “The form asked for my phone number and I almost quit. Why does it need that?”
• “Hit an error on submit: ‘Invalid input.’ Which input? No idea.”
• “Confirmation page didn’t tell me what happens next.”

Every line in that log is an audit finding, and the walk costs you twenty minutes.

Lens 3: The friction lens — how much effort does each step demand?

Friction is everything that makes the next step feel like work: forms with too many fields, flows with too many steps, pages that load slowly, error messages that scold without explaining. For each funnel step, build a small effort inventory: How many fields? How many taps or clicks? How long does it take to load on a phone? What happens when something goes wrong — does the error say how to fix it?

Ask of every field and every step: “What do we actually do with this?” If the honest answer is “nothing yet,” it’s friction wearing a lanyard. Checkout and signup flows are where friction concentrates, and they deserve their own deep pass — I’ve written a whole companion piece on how to optimize a checkout flow that goes step by step through exactly that.

Lens 4: The clarity lens — can a stranger answer four questions?

On every key page in your funnel, a first-time visitor should be able to answer four questions within seconds: Where am I? What is this? Why should I care? What do I do next? That’s the four-questions pass. Run it on each funnel page and be ruthless — “they’ll figure it out” is not an answer.

Practically, you’re checking three things are present and findable: a clear value proposition (why should I care), visible proof that real people use and like this (more on proof in the trust lens), and one obvious call to action (what now). Vague buttons are a clarity leak all by themselves — “Submit” and “Learn more” tell a visitor nothing about what happens next. If your buttons are mumbling, my guide on how to write button copy covers how to make them speak up.

Lens 5: The trust lens — what doubts are you leaving unanswered?

People don’t just leave because a page is confusing; they leave because something made them nervous. The trust lens is a doubt inventory: walk the funnel asking “what would make a careful stranger hesitate right here?”

• Is the price hidden until deep in the flow? Surprise is the enemy of trust.
• Are policies buried? Refunds, cancellation, shipping — if people have to hunt, they assume the worst.
• Is proof thin? Testimonials and reviews build trust only when they’re real and shared with permission — fabricated praise isn’t a conversion tactic, it’s a liability in every sense.
• Is security unclear at the exact moment you ask for personal or payment details?
• Does the page look maintained? A copyright year from three years ago whispers “is anyone home?”

Each unanswered doubt is a finding. The fix is almost always honest visibility: show the price earlier, link the policy where the doubt occurs, put real proof next to the ask.

How do you gather evidence beyond your analytics?

The five lenses run on analytics plus your own eyes. Three adjuncts add other people’s eyes — use them honestly and they’ll sharpen every lens.

Session recordings and heatmaps. Watching anonymized replays of real visitors moving through your leaky step shows you hesitation and rage-clicks that numbers can’t. But privacy comes first, always: mask anything personal, never record sensitive fields like passwords or payment details, honor consent requirements in your region, and check your tool’s settings rather than assuming the defaults protect people. A conversion insight is never worth a privacy violation.

User-testing-lite. You don’t need a lab. Ask five people who’ve never seen your site to complete your funnel while thinking out loud — friends-of-friends, a coworking neighbor, a small testing panel. Five strangers will surface stumbles you’re blind to, and five strangers beat zero strangers every single time. Give them a task, not a tour, and resist the urge to help when they struggle. The struggling is the data.

The support and sales question mine. Go read what people ask right before they buy — support tickets, DMs, emails, sales-call notes. Those questions are objections your pages failed to answer: “Does this work with…?”, “Can I cancel anytime?”, “How long does shipping take?” Every recurring pre-purchase question is evidence of a clarity or trust leak, delivered free of charge, in your customers’ own words.

How do you turn audit findings into a fix queue?

An audit that ends in vibes was just a long walk. The output of a real one is a findings log — one row per issue, each with the evidence behind it. For every finding, record: the issue, the evidence (the number, quote, or observation that proves it), the funnel step it lives at, severity, reach, and rough effort to fix.

Then prioritize. I like a simple table ordered by one principle: big leak + low effort goes first.

Finding	Funnel step	Evidence	Severity × reach	Effort	Queue
Error message doesn’t say which field failed	Checkout form	Walk log + recordings show repeated resubmits	High × high	Low	1 — fix now
Price ambiguity (one-time vs monthly)	Course page	Walk log + 14 pre-purchase emails asking	High × high	Low	2 — fix now
Headline may undersell the offer	Course page	Four-questions pass felt weak; opinion	Unknown × high	Medium	3 — test
Footer link color on one page	Blog	Noticed in walk	Low × low	Low	Backlog

(That’s an ICE-style framing — impact, confidence, effort — without the framework worship. The table is a thinking aid, not a religion; if a scoring debate takes longer than the fix, just fix it.)

The fix-vs-test split

Not every finding deserves an experiment, and this split keeps you honest:

• Obvious breakage gets fixed, immediately, no test required. Broken links, forms that error out, pages that lie (a price that doesn’t match checkout), missing mobile buttons, inaccessible contrast. You don’t A/B test whether broken should be unbroken.
• Judgment calls get tested — where your traffic allows it. New headline, restructured page, different offer framing: these are hypotheses, and a proper split test is how you avoid fooling yourself. But statistical honesty matters here: a reliable test needs enough conversions that most small sites can’t run one on most pages in any reasonable time. If that’s you, don’t fake it — make big, principled swings, measure before and after on the same metric for comparable periods, and hold your conclusions loosely. An honest “it looks better and nothing got worse” beats a fake “statistically significant win.”

And either way: one change at a time per funnel step. Ship five changes at once and you’ll never know which one helped and which one quietly hurt. The one-change discipline feels slow and is actually fast, because you accumulate knowledge instead of noise.

How often should you run a conversion audit?

An audit isn’t a one-time event; it’s a cadence. The rhythm that works for most small teams:

• Quarterly mini-audit (about half a day): re-pull the funnel numbers, re-check the data-quality gate, walk the funnel once on your phone, and refresh the top five rows of the findings log. Funnels drift — a plugin update, a new page, a shifted traffic mix — and the mini-audit catches drift early.
• Annual full pass (a day or two): all five lenses on all core funnels, fresh recordings, a new round of five strangers, and a re-read of the support-question mine.

And here’s the step almost everyone skips: the post-audit retro. Before each new audit, look back at the last one. Which fixes shipped? What did each one actually do to its funnel step’s numbers? Which “sure thing” did nothing? That accountability loop is what turns auditing into a skill — you learn which of your instincts to trust, and your next audit’s priorities get sharper. Without the retro, you’re not auditing; you’re just periodically worrying with a spreadsheet.

One more cadence note: your funnel’s entries matter as much as its steps. If the top of your funnel is organic social — and for a lot of small businesses it’s the biggest door — keep an eye on which platforms and posts actually send people into your money path, so you’re auditing the funnel your real traffic walks, not the one you imagine.

What should a conversion audit never do?

A few traps dress up as auditing. Name them so you can walk past them.

• Redesign-as-audit. If the conclusion was “rebuild the site” before the data was pulled, that wasn’t an audit — it was a redesign looking for a permission slip. An audit is a flashlight, not a renovation crew; let findings drive scope, never the other way around.
• Page-vanity audits. Polishing whichever page is most visible (usually the homepage) while ignoring the funnel. Your homepage might not even be in your money path.
• Copying competitor flows blind. Their checkout has one step and yours has three, so you copy theirs? You can’t see their data, their constraints, or whether their flow even works for them. Steal questions from competitors, never answers.
• Benchmark-chasing. Hunting for “the industry average conversion rate” and grading yourself against it. There is no universal number — rates vary so much by offer, price, traffic source, and audience that cross-site comparisons are noise. The only benchmark that matters is your own funnel last quarter versus your own funnel now.
• Dark-pattern “fixes.” An audit never recommends tricks: no fake countdown timers, no invented “only 2 left!” scarcity, no guilt-trip opt-outs, no hiding costs until the last screen, no making cancellation a maze. These aren’t conversion tactics with an ethics asterisk — they’re trust-destroyers that trade long-term customers for short-term clicks, and the doubt they create is exactly what your trust lens exists to remove.

Your conversion audit workbook

Everything above, condensed into the four artifacts you’ll actually use. Copy these into a doc and you’re ready to run.

The five-lens checklist set

• Data: Funnel mapped? Tracking passed the gate? Step-to-step rates pulled? Segmented by device and source? Biggest leak named?
• Walk: Full funnel completed on a real phone, cold? Every hesitation, surprise, and annoyance logged?
• Friction: Effort inventory per step — fields counted, steps counted, load felt on mobile, every error message read aloud?
• Clarity: Four-questions pass on each key page? Value prop, proof, and one clear CTA present?
• Trust: Doubt inventory done — price visible early, policies findable, proof real, security clear at the ask?

The findings-log template

One row per finding: Issue | Evidence | Funnel step | Severity | Reach | Effort | Fix or test? | Status. If a row has no evidence, it’s an opinion — it can stay, but label it one.

The prioritization table

Sort the log by severity × reach, break ties with effort, and draw a line under the top five. Big leak + low effort ships first; everything below the line waits for the next cycle.

The walk-the-funnel protocol card

Real phone. Mobile data if possible. Start where your customers start (often a social profile or a search result). Complete the entire funnel. Note everything, no matter how small. Twenty minutes, once a quarter, minimum.

FAQ: how to run a conversion audit

How long does a conversion audit take?

A focused first audit on one core funnel takes roughly a day: an hour or two mapping the funnel and pulling segmented numbers, twenty minutes walking it on your phone, a couple of hours across the friction, clarity, and trust lenses, and an hour building the findings log. Quarterly mini-audits take about half a day once the templates exist.

What tools do I need to run a conversion audit?

At minimum: your existing analytics (for funnel steps and segments), your own phone (for the walk-through), and a spreadsheet (for the findings log). Session-recording and heatmap tools are helpful adjuncts, not requirements — verify any tool’s current features and privacy settings yourself before relying on it, because capabilities change.

What’s a good conversion rate to aim for?

There’s no universal number, and anyone selling you an “industry average” is selling noise — rates vary enormously by offer, price point, traffic source, and audience. The honest benchmark is your own baseline: measure your funnel’s step-to-step rates today, fix the biggest leak, and compare against yourself next quarter.

Should I A/B test every change from my audit?

No. Obvious breakage — broken links, failing forms, misleading information — gets fixed immediately without a test. Reserve A/B testing for genuine judgment calls, and only if you have enough conversions for a trustworthy result. Smaller sites are better served making one bold change at a time and honestly comparing before-and-after periods.

How is a conversion audit different from an SEO audit?

An SEO audit asks whether the right people are finding you; a conversion audit asks what happens after they arrive. They’re complementary: SEO fills the top of the funnel, and the conversion audit makes sure the funnel doesn’t spill what SEO (and your social channels) pour in. If traffic is healthy but revenue is flat, the conversion audit comes first.