Using analytics to find where people drop off

Analytics is where most teams start, and it’s a decent place to start as long as you know what it can and can’t do. It tells you what happened and where. It’ll tell you that 63% of people who reach the basket leave before paying. It won’t tell you they left because the delivery cost only shows up on the final step and was more than the item. The number points at the problem, and you have to go and find out what it is.

The funnel is the basic shape. You line up the steps of a journey, landing to payment or signup, and read the drop-off between each. The total is rarely the interesting bit. What you want is the step where the drop is biggest relative to the others, because that’s where the friction is and that’s the step worth investigating. A page that loses 40% of the people who reach it is a finding. A page that loses the usual 10% is noise. You’re looking for the spike in the line.

Segmentation stops the average misleading you. The same journey looks completely different for a first-time visitor on a phone and a returning customer on a laptop, and the aggregate blurs them into a number that describes nobody. Split by device, by source, by new or returning, and the picture gets clearer. Usually one segment carries the problem and the others are fine, which also gives you a first clue about who to go and watch.

This is where the numbers meet the other kind of research. The numbers tell you where and who, and the qualitative work tells you why. On its own, analytics makes it easy to mistake correlation for cause. Conversion is up this month and we changed the homepage, so the homepage did it. Maybe. Maybe it was the bank holiday, or the sale, or the weather, or any of the hundred other things that changed. The number can’t separate those. That’s what a controlled test is for, and it’s why analytics sits upstream of testing and can’t replace it.

Watch out for vanity metrics, the numbers that are easy to read and mean nothing, like sessions, page views and time on site. They move around the outcomes you care about without being outcomes themselves, and optimising for them makes the dashboard look busy while the business stands still. Pick the one number that is the goal, a sale, a signup, a booked call, and read everything else in service of it. Choosing that number is a decision, and if you don’t make it, the dashboard will make it for you, usually by picking the one that flatters you.

Put it to work

Pull your highest-value journey and find the single step with the largest drop-off. Write down three possible reasons, then note which one you would need to watch people to confirm.

Your answer

Show the worked example

There is no right answer. Write yours first, then compare.

Sources

1. NN/g, Analytics and User Experience
2. CXL, Analytics for Conversion Optimization