Improve Your Conversion Rate with Behaviour Analysis

Behaviour analysis in ecommerce: what the data reveals about your customers, and how to turn it into tests

Every visitor to your shop leaves traces. Where they click. How far they scroll. Where they drop out. What they search for. Most shops collect this data, but few use it systematically. Analytics shows the what. Behaviour analysis shows the why.

Only the two together lead to tests that bring in revenue.

We start every project with a behaviour analysis. Not because it is exciting, but because it has the highest ROI. One week of analysis can replace three months of guessing. This article shows which methods we use, what we learn from them, and how we turn those insights into concrete tests.

The four pillars of behaviour analysis

1. Quantitative analysis: what is happening?

Quantitative data shows patterns. It tells you WHERE a problem is. Google Analytics, Shopify Analytics, Adobe: they all show the same metrics: traffic, conversion rate, bounce rate, time on site, pages per session.

The most important quantitative analysis is segmentation. Not “The conversion rate is 2.1%”, but “The conversion rate on mobile is 1.4%, on desktop 2.8%. For organic traffic it is 2.5%, for paid traffic 1.2%. For returning visitors it is 3.8%, for new visitors 1.5%.”

A fashion shop had an overall conversion rate of 2.0%. Segmentation showed desktop at 2.9% and mobile at 1.1%. The mobile gap was enormous. The analysis went deeper: mobile conversion on the home page 1.8%, on the category page 1.2%, on the product page 0.9%, at checkout 0.4%.

The problem was not the home page. It was the product page on mobile. Heatmaps showed the CTA button sat outside the thumb zone. 60% of mobile visitors scrolled without seeing it. After moving it into the thumb zone, mobile conversion on the product page rose to 1.4%. Total revenue rose by 12%.

2. Qualitative analysis: why is it happening?

Quantitative data shows the problem. Qualitative data shows the cause. Heatmaps, session recordings, surveys, usability tests: they all answer the question of why customers do what they do.

An electronics shop had a product page bounce rate of 64%. Analytics showed the problem. Session recordings showed the cause: visitors clicked the main product image, expected it to enlarge, got nothing, and left.

The fix was trivial: a zoom feature. The bounce rate fell to 48%. The conversion rate rose by 9%. The investment: one developer, 4 hours of work.

Another example: a home and living shop had a checkout abandonment rate of 68%. The recordings showed customers entering their address, clicking “Continue”, and nothing happening. The button was not broken. It was loading.

But there was no loading animation. Customers thought the button did not work and clicked again. Then twice more. Then they gave up. After adding a loading spinner, abandonment at that step fell by 18%.

3. Funnel analysis: where do customers get lost?

The funnel shows the path from visitor to buyer. Every step loses customers. The question is: where do you lose the most, and why?

A furniture shop analysed its funnel: 100% home page → 45% category page → 18% product page → 8% cart → 3% checkout → 2% purchase. The biggest loss was between the home page and the category page (55% drop-off).

Heatmaps showed the main navigation was overloaded: 12 categories, 8 of them with submenus. Customers were overwhelmed. After cutting to 5 main categories with clear subcategories, the flow from home page to category rose to 62%. Total revenue rose by 15%.

4. Cohort analysis: who comes back?

Cohort analysis groups customers by the date of their first purchase and shows how many buy again after 30, 60 and 90 days. It is the most important metric for long-term growth.

A beauty shop was strong at acquiring first-time buyers, but the cohort analysis showed that only 8% of them bought again within 90 days. The analysis went deeper: customers who came via paid social had a 90-day repurchase rate of 4%.

Customers from organic search had a rate of 14%. The cause: paid social customers bought on impulse, often products they did not really need. Organic customers researched carefully and bought products that suited them.

The consequence: the shop cut its paid social budget by 20% and invested in SEO content offering buying advice. The number of first-time buyers fell by 8%, but the 90-day repurchase rate rose to 16%. LTV doubled. Total revenue rose by 23%.

From analysis to test: the workflow

Behaviour analysis on its own makes no money. It has to be built into a testing workflow:

1. 1
   

   Collect data.

   Analytics, heatmaps, recordings, surveys. Identify patterns.

2. 2
   

   Write a hypothesis.

   “Mobile conversion on the product page is 50% lower than on desktop. We suspect the CTA button sits outside the thumb zone. Test: move the button into the lower half of the screen. Expected result: +15% mobile conversion.”

3. 3
   

   Prioritise.

   Not every hypothesis matters equally. Prioritise by impact (how much revenue is at stake?), confidence (how sure are you?) and ease (how easy is the test?).

4. 4
   

   Test.

   A/B test with statistical significance. At least one week. At least 1,000 visitors per variant.

5. 5
   

   Evaluate.

   Primary metric: RPU. Secondary metrics: conversion rate, AOV, bounce rate.

6. 6
   

   Roll out or iterate.

   The winner is rolled out. The loser is analysed: why did it not work?

Tools for behaviour analysis

We use a standard tool stack:

Quantitative: Google Analytics 4 (segmentation, funnels, ecommerce tracking), BigQuery (for complex queries over large datasets).

Qualitative: Hotjar or Microsoft Clarity (heatmaps, recordings), UserTesting.com (for targeted usability tests).

Analysis: Excel or Google Sheets (for cohort analyses and manual segments), Tableau or Looker (for dashboards).

The question you ask matters more than the tool. An Excel sheet with the right question is worth more than a Tableau dashboard with the wrong one.

Frequently asked questions about behaviour analysis

How long should I collect data before I test?

At least one week for quantitative data (to capture weekday effects). At least 100 recordings for qualitative data (to spot patterns). For seasonal products: at least one full season.

Should I analyse all the data myself?

Not necessarily. Many shops collect data, but nobody has time to analyse it. A weekly 30-minute review is better than a monthly 4-hour marathon. Or hire an analyst for the setup and the first analyses, then carry on in-house.

How do I avoid drowning in data?

By tying every analysis to an action. If an insight does not trigger a concrete test or change, it is not valuable. Data is not an end in itself. It is the starting point for decisions.

What is the most important analysis?

Funnel analysis. It shows you where you lose the most customers. A shop that loses 60% on the home page has a different problem from one that loses 60% at checkout. Funnel analysis prioritises your tests.

How often should I analyse?

Daily for operational metrics (revenue, conversion, abandonment). Weekly for qualitative data (recordings, heatmaps). Monthly for strategic analysis (cohorts, channel performance, LTV). Quarterly for comprehensive reviews.

Behaviour analysis as a culture

Behaviour analysis is not a department. It is a culture. Shops that get value from it put a fixed slot in the calendar, weekly or fortnightly, in which the team looks at the last period’s data and agrees one decision.

Training matters as much as the cadence. When marketing, sales and support can read the data too, not just the analysts, the quality of the hypotheses rises and the analysis stops being a bottleneck.

You will find more in-depth material in the Academy track A/B-Testing Mastery ↗.