Ecommerce analytics: trace one problem through checkout

Ecommerce analytics is the practice of connecting what shoppers do to what the store actually delivers: product discovery, checkout, payment, fulfilment and repeat questions. A dashboard full of visits and conversion rates is useful only when it helps you locate a specific obstacle and test a change. This guide starts with one purchase journey rather than a list of isolated metrics.

Define the decision before opening a report

Write a question that could change an action: “Why do visitors reach checkout for this product but rarely complete payment?” or “Do people who land on this category find a product?” Avoid beginning with “How can we improve our conversion rate?” because the answer may differ by device, product, language and traffic source.

Choose a comparison period that makes sense for the store. A holiday week should not be compared casually with an ordinary week. Record major campaigns, price changes, stockouts and site releases. Without those notes, a graph can suggest a false cause.

Decide which result matters: completed orders, gross margin, enquiries, successful downloads or repeat purchases. A lower conversion rate during a campaign that reaches new audiences does not automatically mean the site became worse. Look at absolute orders and order quality too.

Map the path from landing page to order

Start with a particular landing page, collection or product. Observe sessions reaching it, product views, add-to-cart actions, checkout starts and completed purchases. If the platform reports a funnel, read its definitions: open and closed funnels can count journeys differently. Shopify's behavior reports describe its session-based funnel.

In GA4, ecommerce events such as view_item, add_to_cart, begin_checkout and purchase need to be sent with appropriate product data. They are not automatically created for every site; Shopify integrations may send some of them. Check implementation and event counts before trusting a funnel. Google's ecommerce setup guidance explains the event model.

Sessions are not people, event counts are not orders, and gross sales are not profit. Keep the measurement unit next to every number in a report. Reconcile purchase events against the order system over a sample period. If tracking says 80 purchases and the store says 63 orders, diagnose duplicate events, missing consent, test orders or timing before optimising a percentage.

Segment only where it changes a decision

Split the journey by mobile and desktop when the interface differs. Separate new and returning visitors if the offer or trust level differs. Compare traffic sources when an ad promises one thing and the landing page shows another. Avoid dozens of tiny segments that produce dramatic percentages from three orders.

For each segment, record its sample size and period. A mobile conversion rate of 2% from 50 sessions and 1% from 50 sessions is a difference of one order; it is not a firm diagnosis. Use the numbers to choose a page to inspect, then observe actual page behavior and customer questions.

Add qualitative evidence. Read support messages, failed search terms, return reasons and abandoned checkout feedback where available. These may explain what a funnel cannot: uncertainty about shipping, compatibility, returns or payment methods.

Diagnose a specific drop-off

Suppose 1,000 sessions land on a product page, 80 add the product to cart, 35 start checkout and 20 purchase. These are fictional numbers. The strongest initial question may be why 45 carts do not reach checkout. Check the cart on a phone, delivery cost visibility, discount fields and stock state. Do not change the hero image just because the overall conversion rate is 2%.

If the break is between checkout start and purchase, inspect payment errors, unavailable delivery options, unexpected total cost and account requirements. If the break happens before product view, look at category navigation, filters and whether the landing page matches the search or ad.

Write down one hypothesis and a test. Example: “Shipping cost is first revealed in checkout, so some buyers stop in the cart.” The intervention might be showing a reliable delivery estimate earlier. The validation is not “the team likes the new copy”; it is a comparison of cart-to-checkout progression and complaints over comparable periods, while watching margin and orders.

Build a report that an owner can use

A weekly report needs a small set of linked measures: sessions to priority landing pages, product views, cart additions, checkout starts, orders, refunds and support issues. Add a short note for anomalies. Assign an owner to each proposed action and date the change. A single chart with no decision attached is decoration.

Retest without claiming certainty too soon

After a change, verify that the public page and analytics events still work. Compare the same stage and segment with a suitable previous period. Document concurrent campaigns or stock changes. If traffic is small, expect noisy rates and look for repeated qualitative signals; do not declare a winner from a handful of orders.

Make the next decision explicit: keep the change, revise it, roll it back, or collect more evidence. Ecommerce analytics earns its place when it shortens the path from a customer problem to a verified improvement.

Check the data before a major decision

Before spending money on a redesign, open a small sample of real orders and check whether their dates, values, products and transaction IDs appear correctly in the analytics system. Exclude test orders from commercial interpretation while preserving a record of them for debugging. Inspect whether refunds are reported and whether consent settings affect measured sessions. If the site has multiple currencies or languages, verify that the reporting view does not silently mix them. A corrected event setup can change a dashboard even when customer behavior is unchanged, so annotate the date of tracking fixes.