Optimise Your Online Shop with Ecommerce Conversion Tracking

Ecommerce tracking: what to measure, and what is a waste of time

Every ecommerce shop has tracking. Most have too much. Google Analytics, Facebook Pixel, Hotjar, Klaviyo, a CRM system, an email tool: each one collects data, but hardly anyone knows which data matters for decisions. The result: dashboards full of numbers nobody reads. Reports nobody understands.

And decisions based on gut feeling instead of data.

Good tracking means the right tracking, not more of it. Three metrics you check every week beat thirty you never look at.

The 3 metrics that really count

1. Revenue per session (RPS)

Not conversion rate. Not average order value. Not traffic. Revenue per session is the one metric that brings everything together: how much money does the shop make per visitor?

A fashion shop raised its conversion rate from 1.8% to 2.3%. The team celebrated. But the average order value fell from 87 euros to 64 euros, because the new customers only bought sale items. Revenue per session did not rise. It fell slightly.

The conversion rate had gone up. Revenue per session had gone down. The optimisation had improved the wrong metric.

Another shop raised its average order value from 45 euros to 62 euros through upselling. The conversion rate fell from 2.1% to 1.7%, because the upselling put some customers off. But revenue per session rose from 0.95 euros to 1.05 euros. The right metric showed the gain.

The rule: if you track only one metric, track revenue per session. It is hard to game and tells you the truth about the health of your shop.

2. Checkout abandonment rate by step

The overall checkout abandonment rate (usually 60–70%) is a vanity metric. It shows that a problem exists, but not where. What you need is the abandonment rate for each checkout step.

An electronics shop had a checkout abandonment rate of 68%. The team did not know where to start. The step-by-step analysis showed:

Basket: 100% reach it (starting point)

Address: 82% reach it (18% drop-off)

Shipping: 71% reach it (11% drop-off)

Payment: 52% reach it (19% drop-off)

Confirmation: 32% reach it (20% drop-off)

The biggest drop-off did not happen at the start. It happened on the confirmation page, where customers were meant to review their order one last time. The recordings showed: the page took 4.2 seconds to load. Many customers thought the process had frozen and closed the window.

The fix: the confirmation page was brought down to a 1.2-second load time. Drop-off at this step fell from 20% to 8%. Overall checkout conversion rose from 32% to 41%. Overall tracking would never have uncovered this specific fault.

3. Customer acquisition cost (CAC) by channel

Not total CAC. Not marketing spend divided by new customers. CAC per channel: what does a new customer cost through Google Ads? Through organic search? Through Instagram?

A cosmetics shop had a total CAC of 24 euros. That seemed acceptable. The channel analysis showed:

Google Ads: CAC 18 euros

Instagram Ads: CAC 31 euros

Organic: CAC 8 euros (content costs spread over time)

Affiliate: CAC 42 euros

The affiliate channel was losing money, but nobody knew because the total CAC hid it. The budget was shifted: more Google Ads, less affiliate. Total CAC fell to 19 euros, with the same marketing spend.

The rule: aggregated metrics lie. Segmented metrics tell the truth.

The vanity metrics you should ignore

1. Bounce rate

Bounce rate measures how many visitors view only one page and then leave. It says nothing about revenue. A blog article with an 80% bounce rate can generate revenue if the 20% who stay convert. A landing page with a 30% bounce rate can bring in nothing if nobody buys.

A shop cut its bounce rate from 65% to 45% with an interactive quiz on the homepage. Visitors stayed longer, but the conversion rate fell from 2.1% to 1.4%. The quiz distracted from buying. The lower bounce rate was a consolation prize.

2. Average time on page

Longer is not better. A customer who spends 5 minutes on a product page is no more convinced than one who needs 30 seconds. They are just slower, or more confused.

A furniture shop had an average time on the product page of 4.2 minutes. The team was proud. The recordings showed: customers scrolled back and forth because they could not find the dimensions. The long visits were searching, not reading. After the page was reworked, time on page fell to 1.8 minutes, and conversion rose by 12%.

3. Pages per session

More pages do not mean more interest. They often mean more searching. A customer who visits 12 pages before buying is no more engaged than one who buys after 2 pages. They are just less sure where to go.

A fashion shop cut average pages per session from 8.3 to 4.1 with a better search function and clearer navigation. The conversion rate rose by 9%. Customers found what they were looking for faster. Fewer pages, more revenue.

Tracking setup: the minimum that has to work

You do not need a 20,000-euro analytics stack. You need three things that work reliably:

1. Ecommerce tracking in Google Analytics 4. Not just page views. Product views, add-to-basket, checkout steps, purchases, all as events. Without ecommerce events every other analysis is worthless.

2. One working conversion pixel. One. Not five. Every extra pixel slows the page down and complicates consent management. Our recommendation: GA4 for analysis, one retargeting pixel (Meta or Google) for advertising. Anything more is overhead.

3. One system for qualitative data. Heatmaps or recordings, not both. Pick one tool and use it with a clear purpose. Most shops have 5 analytics tools and use none of them properly.

A sporting goods shop had 8 tracking tools running. The page loaded 2.3 seconds slower than without tracking. The mobile bounce rate was 58%. After cutting down to 3 tools (GA4, Meta Pixel, Hotjar), load time dropped by 1.8 seconds. The bounce rate fell to 44%. The conversion rate rose by 7%, just from less tracking.

How we use tracking for CRO decisions

Tracking is not an end in itself. It has to lead to decisions. Our workflow:

1. 1
   

   Weekly review of the 3 core metrics: RPS, checkout abandonment by step, CAC by channel.

   Any deviation of more than 10% from the previous week’s average triggers an investigation.

2. 2
   

   Form a hypothesis.

   “Checkout abandonment on the payment page has risen by 15%. We suspect a problem with the payment provider.”

3. 3
   

   Collect data.

   Do not analyse everything, only what fits the hypothesis. In this case: payment provider performance, error rates, recordings of the payment page.

4. 4
   

   Act.

   If the hypothesis is confirmed, you act. No further meeting. No “let’s keep an eye on it”. A test or a direct change.

A home & living shop followed this workflow. The weekly review showed: RPS had fallen by 12%. The checkout analysis showed: drop-off on the shipping page had risen from 18% to 31%. The hypothesis: a new shipping provider had been added and the shipping costs were being calculated wrongly.

The data confirmed it: customers who chose this provider suddenly saw 12 euros more in shipping costs. The action: the provider was removed temporarily and the fault fixed. Within 48 hours RPS was back to its normal level.

Without a weekly review, the fault would have gone unnoticed for weeks. Without a segmented checkout analysis, nobody would have known where the problem was. Tracking is only as valuable as how regularly you evaluate it.

Frequently asked questions about ecommerce tracking

Do I need Google Analytics 4, or is Universal Analytics enough?

Universal Analytics has been shut down. GA4 is a must. But GA4 is not intuitive. The reports are structured differently and real-time data is limited. Invest time in GA4 training, or work with an analyst who knows the system. Misreading GA4 data is worse than no tracking at all.

What is the maximum number of tracking tools a shop should use?

Three: one analytics tool (GA4), one ad pixel (Meta or Google), one qualitative tool (Hotjar or Clarity). Every extra tool slows the page, adds data protection complexity and makes it less likely that the team actually uses the data.

Should I use server-side tracking?

From 50,000 euros in monthly revenue: yes. Server-side tracking is better for data protection and less exposed to ad blockers. But the setup effort is high (10,000–20,000 euros). Below 50,000 euros in revenue it is not worth it. Client-side tracking is enough.

How often should I evaluate my tracking data?

Weekly for the 3 core metrics. Monthly for deeper analysis (channel performance, cohorts, lifetime value). Daily evaluation leads to overreaction: small swings get read as trends. Quarterly evaluation is too rare: problems are spotted too late.

What is the biggest tracking mistake you see?

Tracking everything, and using nothing. Shops with 40 metrics on the dashboard that never base a decision on them. Better three metrics checked weekly than forty that get ignored.

You will find more on this in the Academy track Data-Driven Decision Making ↗.