A shopper moves through product, cart and checkout steps on a phone. Harpy Guide

Find the Leaks — D2C Website Conversion Audit

9 min read · Harpy Media

A funnel is easier to improve once you know where buyers leave. This guide shows you how to measure the drop, check common friction points, and choose one change at a time.

Start with the numbers, not the page element you happen to dislike. The biggest drop is a better place to investigate than the loudest opinion in the room.

Part 1

Your revenue is hiding in your drop-offs

Before buying more traffic, look at what current visitors do on the site. A funnel can help you see where people leave and which step is worth investigating first.

This audit is the habit behind our Conversion Architecture work (Frameworks № 01): find the drop-offs, fix them in revenue order, and re-measure until the curve bends. It is deliberately unglamorous — five steps, no new tools, percentages instead of vibes.

🔑 Key insight. You don't have a traffic problem first. You have a leak problem first. Doubling visitors into a leaking funnel doubles the water on the floor; fixing the leak makes every existing visitor worth more — immediately, and forever after.

Part 2

Step 1 — pull the funnel numbers

Start with four stages: product-page views, add-to-cart, checkout and purchase. Compare the rate at each step over a consistent period, so a single unusual day does not drive the decision.

View → add-to-cart → checkout → purchase — each step as % of the previousThe funnel is a ratio chain: each leak compounds into the next

Sanity-check the chain before trusting it: a broken analytics event produces beautiful-looking funnels that don't connect to bank deposits. If purchase events and orders don't reconcile within a few percent, fix the measurement first — you'll otherwise optimize beautifully against a fiction.

Part 3

Step 2 — find the biggest drop

Look first for the step where the most shoppers leave. That is usually a better place to begin than a small issue that is simply easier to notice.

The arithmetic that makes it obvious: 10,000 monthly views. If 20% reach add-to-cart (2,000), 50% of those reach checkout (1,000), and 50% of those purchase (500) — the view-to-cart step loses 8,000 people; the checkout step loses 500. Even a modest 20% improvement on the biggest leak (8,000 → recover 400+ shoppers) outperforms a heroic 50% fix on the smallest one. Leaks are ranked by volume through them, not by how fixable they look.

💡 Tip. Write the leak as a sentence with a number in it: 'Sixty percent of shoppers who add to cart abandon at the shipping step.' If you can't write that sentence, you don't have a diagnosis yet — you have a hunch.
Find the step where intent thins outThis funnel is a map, not a benchmark. Fill it with your own rates, then investigate the largest meaningful drop.CONCEPTUAL FUNNEL · USE YOUR OWN RATESVISITProduct page viewsCARTAdded to cartCHECKOUTStarted checkoutORDERCompleted purchase
This funnel is a map, not a benchmark. Fill it with your own rates, then investigate the largest meaningful drop.

Part 4

Step 3 — the five usual suspects

One leak at a time, check the causes in order of how often they're guilty:

SuspectWhat it looks likeThe test
Slow loadEvery second of mobile load time taxes every step at once — it's the only leak that leaks everywhereField data, not lab scores: real-user mobile timings on your money pages
Unclear offer above the foldShopper can't say what this is, who it's for, and why it's better within two seconds of landingThe five-second stranger test on your top landing page
Weak proofClaims without reviews, numbers, or faces — trust asked for, not earnedIs your strongest evidence visible without scrolling?
Surprise shipping costCart total jumps at the last moment — the classic checkout killer, especially in price-aware marketsCompare cart-abandon rate before vs. after shipping displays
Payment trust gapsMissing familiar payment options, clumsy OTP flows, no visible security signals at the pay stepWatch five first-time shoppers attempt checkout, in silence
⚠️ Watch out. Fix the step you measured, not the step you enjoy. The biggest leak is often boring infrastructure — load time, a payment integration — while the fun fixes (hero redesigns, new copy) live at steps losing a fraction of the shoppers.
Check the friction buyers can feelThese are common places to look, not a diagnosis. Use customer behaviour and testing to confirm which one matters on your site.A QUICK FIELD CHECKLOADDoes the page feelslow?OFFERIs the value clearup front?PROOFIs trust easy tofind?SHIPPINGAre costs clearearly?PAYMENTDoes checkout feelsafe?
These are common places to look, not a diagnosis. Use customer behaviour and testing to confirm which one matters on your site.

Part 5

Step 4 — fix in revenue order

Make one change at a time where you can. That makes it easier to see what changed and whether the update helped.

1. Rank the leaks by shoppers lost × proximity to purchase.A leak at checkout is worth more than the same-sized leak at first view — those shoppers already decided to buy.
2. Fix the top one. Ship it. Wait out the measurement window.Give each fix a fair trial — two weeks minimum on most stores, longer on thin traffic.
3. Re-measure the whole funnel.Not just the fixed step — fixes ripple. A faster checkout raises purchases; it can also expose a weak product page that was hiding behind it.
4. Keep or revert, then move to the next leak.The keep/revert discipline is what turns changes into learning instead of churn.

Part 6

Step 5 — re-measure weekly

Check the funnel on a regular schedule and compare like with like. A steady improvement can add up, but any example rate is a calculation, not a forecast for your store.

Weekly 2% improvement → ≈2.8× conversion in a yearCompounding is why 're-measure weekly' is a step, not a footnote

Same day every week, same funnel, one page of notes: each leak, its number, what shipped, what moved. When we run this discipline on client stores, the same traffic converting 28% better is a typical outcome of the first cycle — not because of any single brilliant fix, but because the leaks got fixed in the right order and nothing was changed faster than it could be measured.

That's the whole audit: numbers, biggest drop, usual suspects, revenue order, weekly re-measure. Run it until it's boring, then keep running it — boring compounding is what a healthy store feels like from the inside. Want a second pair of eyes on your funnel? The first diagnosis is free.

Part 7

What healthy looks like

Percentages mean nothing without a reference. Typical healthy ranges for a D2C store — treat them as orientation, not gospel, since traffic quality and price band move them:

StepTypical healthy rangeIf you're under
View → add-to-cart≈5–10%The offer is unclear or the product page isn't earning desire — suspects 2 and 3
Add-to-cart → checkout≈40–60%Something changes between wanting and committing — price reflection, trust, comparison shopping
Checkout → purchase≈50–70%The pay step itself — surprise costs, payment trust, clunky OTP or UPI flows

Two honest caveats. First, ranges vary by traffic quality: a store running heavy top-funnel ads will show a thinner view-to-cart than one living on branded search, and that's not a leak — that's the mix. Compare your funnel to your own history first, industry ranges second. Second, don't chase all three steps at once — the biggest-drop rule from Step 2 still governs.

💡 Tip. Benchmark against yourself in week-over-week trend form: a funnel whose every step drifts down 1% monthly is in worse shape than one step sitting under the range but stable. Direction first, absolutes second.

Part 8

When the leak isn't on the site

Sometimes you audit every suspect, fix every fix — and the funnel still doesn't move. Then the leak isn't in the store; it's in the match between the traffic and the page. Two versions of this:

Wrong audience. Ads promising one thing (price, novelty, urgency) landing on a page selling another (premium, utility, considered purchase). The page converts fine for people who meant to arrive; the bought visitors bounce because they were promised a different store. The tell: branded and direct traffic converts multiples better than paid, with the same page.

Right audience, wrong landing. An ad for one product landing on the homepage, a collection page, or a different hero — every step of which re-asks the question the shopper already answered. The tell: the specific campaign's funnel dies at view → add-to-cart while the rest of the site holds.

🔑 Key insight. Before redesigning a page, check who was arriving at it. Half of 'conversion problems' are actually promise problems — the ad, the listing, or the link made a promise the destination didn't keep. Fix the match before the page; it's cheaper, faster, and usually the actual problem.

Want this run on your account instead? The first diagnosis is free.

Book Free Consultation

Related terms

CRO (Conversion Rate Optimization)CVR (Conversion Rate)UX (User Experience)PDP (Product Detail Page)ATC (Add to Cart)

New guides, straight to your inbox.

Practical D2C playbooks as we publish them. No fluff, no spam — unsubscribe anytime.