Safari ITP 2.3 kills Google Optimize & Google Analytics cookies as soon as 24 hours after being set with sessions referred from paid ad link clicks, but not longer than 7 days (a remnant of ITP 2.2), if there is no return visit within those respective timeframes to reset the clock.

Note that the new iOS 14 AppTrackingTransparency framework can also impact Google Optimize & Google Analytics cookies with in-app Safari browsing if users elect to not allow themselves to be tracked “across apps and websites owned by other companies“.

So what?

If a Google Optimize test lasts more than 7 days, users browsing with Safari may fall out of the experiment. A Safari browsing user included in an experiment could fall out after 24 hours if they are referred from a paid ad link click. As such, if / when a Safari browsing user that has fallen out of a test returns to the site, they will be added as a net new user.

When this happens you’ll see artificially high unique visitor counts in a test, disintegrating winning variant calculations. Also, if a user falls out of a test they will no longer see the variant they were previously presented, which could be confusing – and may lower conversions rates.

Personalizations can be impacted too if they are driven by cookie’d data. Personalizations based on environment variables (e.g., device category) and/or data layer variables would not be impacted.

What next?

You could exclude devices with Safari browsers from Google Optimize tests. But, that could be an “expensive” approach if you get a meaningful portion of traffic from Safari browsing users as experiment results would not consider that potentially sizable user base.

You can also run parallel tests, respectively including or excluding devices with Safari browsers. The test version including Safari browsers would allow you to see how a portion of those Safari browsing users falling out of the test impacts results.

Possibly the best long term solution is to use Google Tag Manager (GTM) server-side tagging to set the Google Optimize experiment ID (cookie name =_gaexp) and Google Analytics client ID (cookie name = _ga) as HTTP only cookies. This approach eliminates the related ITP shortened cookie expirations. More on that in another upcoming post…

Safari ITP impact on Google Optimize tests