6.5 Cookies and trackers

Avoid cookies, especially third-party cookies

Cookies

A cookie is a small file.
It is stored by a server on the terminal (computer, phone, etc.) of a user. It is associated with a web domain. This file is automatically sent back during subsequent contacts with the same domain, website.

Cookies have multiple uses. They can be used to remember your:

• customer ID with a merchant website;
• shopping cart;
• language for displaying the web page;
• pre-filled fields in a form;
• identifier for tracking your browsing for statistical or advertising purposes.

Some of these uses are necessary for the main functionalities of the service (requested by the user, so without the need for the user's agreement).
Others, which are not essential, require the user's agreement.

Third-party cookies

They are deposited by third parties, different from the main site visited by the user. They are mainly used by third parties to:

• know the pages visited by the user on the main site;
• collect information about the user, mainly for advertising purposes.

Why should you avoid them?

Cookies, especially third-party cookies, should be avoided for several reasons:

• Transparency issues: the user may not know to whom their browsing data is sold/transmitted;
• Hidden information: information about cookies is often hidden and not easily understandable;
• Ultimate goal of targeted advertising: targeted advertising is made possible by cookies that capture user information, and contrary to what one might think, we are easily recognizable on the internet. All plugins and settings in our workspace make us easily identifiable (see AmI Unique).

Best practices if you have to use them

If you still need to use them:

• Be as transparent as possible. Explain to the user simply:
  • that you use cookies;
  • who will benefit from this information;
  • the purpose of these cookies.
• Avoid third-party cookies.

Avoid A/B testing and mouse tracking

Navigation analytics tools allow you to know:

• Traffic (visitors, visiting periods, access);
• Behaviour (pages viewed, visit duration, bounce rate);
• Even more precise elements:
  • Areas where the user clicked;
  • Their journey.

They are not all necessarily to be avoided. It depends on:

• The information recorded;
• How the collected information is used.

A/B testing

A test comparing two similar pages to see which one is more effective (comparing form or content). The test is conducted on a sample of visitors. Out of 100 visitors, there will be approximately:

• 10 people who test version A;
• 10 people who test version B.

Testers are not aware that they are participating in a test.

The page that will be chosen is often the one with the lowest bounce rate or the highest conversion rate. The problem is that this method encourages capturing more and more user attention and pushing them to consume. Care should be taken with the success metrics that have been decided.

Example

A/B testing for New York Times headlines

The New York Times uses A/B testing for its article headlines.

In 2021, the TJCX blog author (Tom) created a script. The script, for 3 weeks, identified the different possible headlines for articles.

These tests showed that about 29% of articles had multiple headlines (up to 8). There were several significant differences:

• A majority with fairly minor changes (syntax, capitalization, etc.)
• More significant changes that evolve with the progress of the story
• A/B tests that measure and seek to increase the number of clicks.

The trend that emerges, in the case of A/B tests, is that headlines tend to become more dramatic over time.
(See TJCX website example on a Cuomo sex scandal).

Mouse tracking

This involves capturing (collecting data)  the user's mouse clicks and movements on the interface. This method helps guide interface development.

Using this feature seems justified in the testing phase to identify problems. It is not justified in production.

Like A/B testing, when used in production, this method encourages the service to capture more and more user attention and push them to consume.

An alternative is to occasionally conduct user tests. These tests will help understand why and how users interact with your service.