Purchase journey analysis: how to find where shoppers leave your store

Which events to measure

A funnel is built from events – actions your analytics records when a shopper sees a product, adds it to the cart or starts checkout. Google Analytics 4 has a set of recommended ecommerce events; if they aren't sent, or aren't sent everywhere, your funnel will show gaps in measurement rather than shopper behaviour.

For revenue, Google's documentation says to send the currency together with the value parameter, and a list of products (items) with each event. Without them the funnel shows the steps, but not which products and how much money get stuck in them.

In Shopify and WooCommerce stores, some of these events are sent by the platform or its extensions – but not always all of them, and not always correctly. That is why a test order is not a formality but the first step of the analysis.

Consent and data quality

In the EU, analytics cookies need the visitor's consent. Google's consent mode lets tags adjust their behaviour depending on whether a visitor agreed to analytics (analytics_storage) and advertising (ad_storage) storage. The practical consequence is simple: your analytics does not see every shopper.

• Compare periods in which neither the consent banner nor the measurement setup changed.
• Take absolute numbers (orders, revenue) from your store, and step-to-step proportions from analytics.
• Don't collect personal data in event parameters: products, totals and steps are enough for journey analysis.

How to build and read a funnel

GA4 has a funnel exploration for this. You define the steps – events or pages – and see how many users move from one to the next, how many leave and what they do afterwards.

1. Add the main steps: view_item → add_to_cart → begin_checkout → add_shipping_info → add_payment_info → purchase.
2. Choose a closed funnel to see only those who started at the first step, or an open one if shoppers enter the journey at different points (from an ad straight to a product, say).
3. Turn on elapsed time between steps: if it takes long to get from cart to checkout, shoppers are looking for something or hesitating.
4. Review the next action after each step – it shows where leavers went: back to the category, to the delivery terms, or nowhere.

Read the progression rate of each step, not the overall conversion. If only a small share get from the product page to the cart but nearly everyone who starts checkout finishes it, redesigning checkout won't change anything – the problem is on the product page.

Where to look for the biggest loss

Averages often hide the problem. If checkout works well on desktop but the fields are too small on mobile or the payment window won't open, the overall rate will look only “slightly low”. So break the same funnel down.

Look for the biggest loss, not the lowest rate: the step where a weaker progression costs the most orders. A low rate on a rarely visited page can cost less than a small gap on a step everyone passes through.

Numbers show “where”, watching shows “why”

A funnel will tell you that some shoppers leave the cart. It won't tell you whether the delivery cost put them off, they couldn't find the discount code field or they were simply comparing prices. For that you need qualitative data.

• Session recordings and click maps: watch dozens of sessions from the step with the biggest loss, not random ones. Only record visitors who consented, and mask form fields.
• Searches with no results: they show what shoppers look for and can't find.
• Customer service questions: every repeated question by email or phone is information missing from the site.
• Your own test purchase on a phone: from search to payment, with a real payment method and address.
• A short survey for people leaving checkout: one question, “What stopped you?”, with a few answers.

How to write a hypothesis

A finding is not yet a solution. So that a change can be checked, write it down as a hypothesis: at which step, for which shoppers, what gets in the way, what you will change and which metric you will watch.

Each hypothesis is worth assigning to one of two types. If the shopper is left with an unanswered question – about delivery, price, choice or trust – you fix it with information. If the journey itself gets stuck – speed, forms, payment or errors – it needs technical work. We cover the most common causes of both in why visitors don't buy from your online store.

Change one thing, then check

Change several things at once and a month later you won't know which helped and which hurt. So change one thing at a time and judge the result by the measure you wrote in the hypothesis.

• In a high-traffic store, test important changes with a controlled test: only some shoppers see the change, the rest form the control group.
• In a low-traffic store, compare longer periods before and after the change, avoiding holidays and big promotions.
• After every platform or plugin update, place a test order – a fixed journey can break silently.

To stop a fixed journey breaking after the next update, it needs ongoing monitoring. That is what website maintenance is for: updates, daily backups, monitoring and a monthly report.