On the use of “Amazon Jails”

Ever since Amazon removed their navigation from their checkout screen, it has been said that transactions (or other critical tasks the business would prefer the customer to complete) should not have “distracting” navigational elements on the screen. This is because those elements could take people away from the task at hand and erode conversion.

Not only does this idea not stand up to scrutiny (despite apparent quantitative evidence for it), but makes no sense from the point of view of human behaviour.

The case against

For an Amazon Jail to be effective, it must make someone complete a task against their will.

Conversely, without the “jail” they must arrive on a page with the intention to complete an action only to permanently lose that intention on seeing a link to some other part of the site.

If anyone reading this has personally experienced a theft of their free will in this way, it would be good to know. More likely if they think they have, it’s that they didn’t have a strong enough intention to complete the task in the first place.

Obviously, this is not an argument for putting wantonly distracting advertising on a checkout page. But simple hygiene navigation deserves to be there. Predictable navigation is simply good design, and our users should not be deprived of it or led to expect that it might disappear at any time without warning (or that recourse to the browser’s back button should be normalised).

But what of quantitative tests that show removing navigation improves the rate of conversion?

Firstly, consider the rather obvious point that if a link isn’t there, you can’t click it.

Now consider the above scenario where a user apparently intends to transact but they click a link that takes them away from the page instead. But their intent to complete their initial task may still there, so they return later and complete it. That is, unless you think that people can be made to act against their will.

So the immediate rate of abandonment is not a holistic measure of the effect because the impact of a Jail can be measured in a number of ways:

  • Rate of simple abandonment from the checkout process i.e. what % started checkout processes that are not completed.
  • The overall rate of purchases within sessions i.e. are sessions with the Jail enabled more likely to complete a purchase than those who do not.
  • Visitors’ total spend over a given (possibly quite long) period.

Each method has strengths and weaknesses; there is usually a trade-off between directly measuring a behaviour attributable to a change (abandonment) and measuring something meaningful to the business (revenue per visit).


You can use statistics to argue for or against the use of Amazon Jails (and many other things besides), but if you consider the actual UX of what is happening, then you have to conclude it doesn’t make sense.