The trend of anonymized email open data ramped up in 2021, especially with Apple’s launch of Mail Privacy Protection and Twilio SendGrid’s launch of the Apple Open Identifier. Since then, there’s been more interest in the anonymization of open data, specifically around Google prefetching opens. This post explains what Google prefetching is, how to detect prefetching, and what Twilio SendGrid’s plans are to flag prefetching in the future.
Prefetching is a technique for speeding up operations by beginning a fetch function that expedites the result before it's needed. This means sending an email to a recipient during an active/open web session and the inbox provider loading all images immediately before the web UI displays the email.
Google has been prefetching images for faster image loading for a while. We first reported this behavior in 2013, and you can read more about our initial findings here. Since then, we have not seen any significant increases in prefetching opens. Through our research of the past 6 months, we found that Gmail prefetching accounts for roughly 2% or less of total email opens month over month.
In September 2021, Apple updated its Mail app to allow users to have more control over their data privacy. You can read more about our findings and how to identify this behavior in your sending here. Ultimately, Apple’s Mail app similarly causes prefetch opens when Apple Mail users enable the security feature.
Both Apple and Gmail use prefetching, but do so to serve mailbox users differently. Therefore, we advise senders to look at the impact of prefetching from Gmail and Apple differently. Here's a quick look at both:
Gmail signals someone was active in the inbox, resulting in Gmail wanting to render the content quicker to provide a better user experience with shorter loading times.
Apple signals that the recipient has requested anonymity in engagement—they are seeking data privacy.
Even further, both Apple and Gmail may require you to better understand how people engage with your brand beyond email opens, so you’ll need to treat your next communications with them differently. Clicks, conversions, and app activity are good additional metrics to explore.
Since each organization and recipient list is different, the strategy you build to engage with your customers should reflect that. If you have questions around how to build and/or adapt your email strategy with items like Gmail prefetching in mind, reach out to our Expert Services team.
As of today, if you are looking to identify Gmail prefetch opens, you can find the user agent string below passed to you in our Event Webhook.
Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/42.0.2311.135 Safari/537.36 Edge/12.246 Mozilla/5.0
We are reasonably confident that opens generated from this string are indicative of the Gmail prefetch activity, as these happen within seconds of delivery. In testing, we also see these happen when someone has recently had an active web application. You may see continued genuine human open activity after seeing the Gmail prefetch open activity. However, this would come from a different user agent string.
We are committed to ongoing research to find the most effective way to flag prefetch opens when there is a considerable impact on conventional open data. Currently, our Apple Open Indicator is unique to Apple's MPP-related open events because those opens have inflated and anonymized a considerable percentage of conventional open data. Since we do not see a considerable impact, or increase in impact from Google prefetching, we have not prioritized a flag for the webhook unique to these opens at this time.
As this prefetching trend continues, we anticipate there may be other providers beyond Gmail that senders will want to identify. However, today, we do not see any significant or increasing amounts of prefetch opens coming from other providers. We will continue to provide information on these evolving trends and consider Deliverability Services for consultative advice unique to your email program goals.
