CRO checklist for Shopify: running the audit in the right order

The starting point is not a list of improvement ideas but the question of which funnel step loses the most. A store with a strong product page and a weak checkout needs different work than one whose category pages nobody scrolls through. Without that ranking you optimise wherever access happens to be easiest.

Aggregate numbers almost always hide the actual problem. Only a split by device shows whether a step is weak overall or breaks on the phone. The split between new and returning visitors matters just as much: returning customers buy differently, and their numbers flatter every average.

In practice that means defining your funnel steps before you open any report, and writing down which value describes each step. Only then do you look for causes. The table below maps the usual steps to the causes that most often turn out to be behind them in an audit.

An audit sensibly works backwards. Anyone who fails in checkout has already proven purchase intent, so every obstacle removed there turns directly into revenue. Further up the funnel the paths branch more, intent is fuzzier, and the effect of a single change is much harder to attribute.

This order is not a rule without exceptions. If your checkout runs on the Shopify standard and the numbers show nothing unusual there, tick off the first points quickly and put your time into the product page and range navigation. Conversely, detail work on category pages rarely pays off while checkout is still losing customers.

The product detail page is where purchase intent turns into a decision or gets stuck on an open question. Any question left unanswered here is either resolved in a new tab at a competitor or not at all. The audit view is therefore not aesthetic but factual.

Two pieces of information are very often given too late: delivery time and shipping cost. If both only appear in the cart, you postpone a rejection instead of avoiding it. In the same way, a variant selector that leaves unavailable combinations unmarked reads like a bug rather than like guidance.

For products that need explaining, images and sizing information carry more weight than any block of copy. Close-ups of material and finish, a size comparison against a familiar object, and fit information per product instead of one generic table answer exactly the questions that otherwise end up with customer service.

In checkout it is rarely design that decides, it is friction. If the payment method expected in that market is missing, a share of customers abandons at exactly that point without it showing up anywhere else. That is why drop-off rate per offered payment method is one of the first numbers you look at.

Forcing a customer account before the first order costs completed purchases. Guest checkout should sit next to account creation as an equal option, not as a small link underneath. Account benefits can be offered after the purchase, when the data already exists and the customer has a reason to keep it.

Check forms for three things: number of fields, behaviour of address validation, and clarity of error states. Validation that rejects correct addresses with additional lines blocks orders silently. Error messages have to sit at the field, name the actual problem, and preserve whatever has already been typed.

In German-speaking markets, price and shipping information is not only a legal topic but a trust signal. A missing unit price on bulk goods, an unclear VAT statement or a shipping note without reachable details read as sloppiness to experienced buyers, regardless of how well the store is designed.

Withdrawal rights, returns and contact details belong where people look for them in the moment of purchase: near the buy button, in the cart and in the footer. An audit checks whether that content exists and whether it is a few clicks away. The legal review itself belongs with your legal advisors, not in a CRO audit.

Payment expectations in German-speaking markets differ from other regions. Invoice and direct debit are still common here, PayPal is widely established, and in Switzerland many customers additionally expect local methods. Which methods your range and your margin can carry is a decision per market, not a global one.

Load time is often treated as a pure SEO topic, yet it acts directly on the purchase path. What matters is less a lab score than the question of when a visitor can perform the first meaningful action: apply a filter, pick a variant, add to cart. Those are the interactions you measure on real devices.

Layout stability is the underrated part. When a consent banner, a lazily loaded review widget or a late shipping note shifts the content, a tap lands on the wrong button. On the phone this produces abandonments that never show up in any error log, because technically nothing broke.

The most common lever sits with third-party scripts. Every additional widget brings load time, potential shifting and one more source of failure. An audit lists every third-party script that loads, assigns a purpose to each one, and flags those that nobody owns any more.

Beyond a certain catalogue size, internal search is the most important entry point, and its results are rarely reviewed. Take the most frequent search terms from your reports and look at what the store returns for them. Zero results for terms that do exist in the range are a data problem, not a search problem.

In broad categories, filters decide whether anyone reaches a product page at all. What counts are filters that match the buying decision, not the ones that happen to be derivable from existing fields. If filter values are maintained on only part of the catalogue, customers filter out relevant products without noticing.

Check your own analytics for the device split. In many D2C stores the weight sits on mobile, yet changes are signed off on a large screen. That gap creates faults nobody sees: buttons below the visible area, on-screen keyboards covering input fields, and image galleries that swallow the page scroll.

Accessibility is not a separate chapter in a CRO audit, it is the same path with a different input device. Walking through a complete purchase with the keyboard alone surfaces focus traps in overlays, unlabelled fields and error messages marked by colour only. Every one of those also costs sighted mouse users time.

The benefit is twofold: clearly labelled fields, visible focus and sufficient contrast help every customer, especially in poor light and in a hurry. That is why this pass sits high in the order rather than as rework behind the visual design.

Every point from an audit is a hypothesis at first. To turn it into a decision it needs a metric that was recorded before the change. Without that baseline you can prove neither an improvement nor a regression afterwards, and arguments about taste take the place of evaluation.

For event tracking in Shopify checkout, the web pixels interface is the intended route; custom scripts inside checkout are not. What matters is that your measurement setup respects consent and that you know what share of sessions is missing because of it. Otherwise you later compare two datasets of different completeness.

Not every finding needs a test. Clear defects, such as a blocking validation error or a button that cannot be reached on one device, get fixed rather than debated. Testing belongs where there is a genuine choice between two plausible variants and where the volume of orders can actually carry an evaluation.