Movable Ink Research: Analysis of iOS 15 Content Caching

A report from Movable Ink analyzes the early results of Apple iOS 15's recent-time caching changes now that iPhone users have access to Mail Privacy Protection.

Last Updated: April 26, 2022

Abstract
On June 7, 2021 Apple announced new privacy-oriented features in iOS 15 with the launch of Mail Privacy Protection. One of those features was the prefetching of content for users of the Apple Mail client. There has been active conversation within the email marketing industry with little data available to inform the discussion. With the General Availability of Apple Mail in iOS 15, Movable Ink conducted a data analysis of image opens on Apple’s new mobile operating system to understand the differences between iOS 14 and iOS 15.

Data Analyzed
• Emails opened from September 20th, 2021 (the iOS 15 GA release date) to 28th, 2022 for all Movable Ink clients. This totaled over 3.6 trillion content impressions.
• Analysis of Apple Mail with Mail Privacy Protection enabled (iOS 15) vs iOS 14 across every day in that same time period.
• Image opens have been divided into a ventile distribution (defined as 5% chunks) with a mean recency delta calculated for each one.
• Movable Ink will continue to review the data and update this post with the latest findings.

Methodology
For each user interaction with email content, Movable Ink logs timestamped events: opens (user loaded the content in the email), clicks (user clicked/tapped the content and was redirected to a web page), and other types - such as conversion events.

With Apple Mail with Mail Privacy Protection enabled (iOS 15) sometimes prefetching content on a user's behalf, it is not possible to directly determine whether an open event was prefetched by the iOS 15 mail client or the result of direct user interaction. However, by looking at the duration between each user's "open" event and "click" event it's possible to infer which type of open it was. On iOS 14, these represent the amount of time between when the user views the email and when they click on it. This analysis is based on the following assumptions:

• Users with iOS 14 and iOS 15 devices behave similarly. Prior to the iOS 15 release date, users could only obtain iOS 15 by enrolling in the beta program. Following the release, iOS 15 traffic has increased by an order of magnitude from 30M impressions/day to over 1.2B impressions/day which indicates that the iOS 15 traffic represents real users.
• Users who click on content are representative of the entire iOS 15 user base. While these users might exhibit higher engagement levels, there is no data to suggest that their devices would have different fetching characteristics than other users’ devices.

An individual user’s open and click events are not naturally paired: a user may view email content multiple times and click multiple times. For the purposes of this analysis, all pairs of opens followed by clicks were considered, and the one with the shortest duration was used. The intention was to minimize the effect of outliers, such as long-running sessions with multiple clicks.

With these assumptions in mind, Movable Ink examined the difference in open-to-click time between iOS 14 and iOS 15 users and concluded that they were due to Apple Mail iOS 15 sometimes loading content prior to the user interaction (ie: prefetching).

Results

The analysis yielded some surprising results. Movable Ink can validate that content prefetching is occurring across a limited percentage of iOS 15 Apple Mail users who have enabled Mail Privacy Protection.

Here are the findings of recency for all images:

Average iOS 15 Apple Mail Recency Analysis

• 85% of Apple Mail users see content rendered less than 2 minutes prior to open
• 5% of Apple Mail users see content 2 to 21 minutes old
• 5% of Apple Mail users see content 21 to 144 minutes old
• 5% of Apple Mail users see content over 144 minutes old

Since no communication is sent to a full file of just iOS 15 recipients, Movable Ink extrapolated the above Apple Mail data across an average all-mail client file. The proportion of mail clients in this calculation were based on Movable Ink content impressions in the two-week period 9/4 to 9/17 leading up to the iOS 15 release. Of the total content impressions, 43% were from iOS devices and 10% were from desktop Apple Mail, for 53% total Apple audience.

Note: Apple’s upcoming desktop MacOS version (Monterey) will also feature Mail Privacy Protection but has not yet been released. We expect that desktop Apple Mail, traditionally representing 10% of email traffic, may have higher rates of pre-fetching, due to being permanently connected to power and wifi. We will update this post after MacOS Monterey has been released.

Extrapolating to the full all-mail client file list:

• 92% of all recipients see content less than 2 minutes old
• 2.6% of all recipients see content 2 to 21 minutes old
• 2.6% of all recipients see content 21 to 144 minutes old
• 2.6% of recipients see content over 144 minutes old

Conclusions

1. Opens will not be a reliable metric going forward on Apple Mail. Because Apple will make image requests for all emails, regardless of a user-initiated open, opens cannot be relied upon as a behavioral indicator (performance metric, deliverability indicator, trigger for a journey split etc).
2. For the vast majority of Apple Mail users, “recent-time” is far closer to “real-time” than expected. 85% of iOS 15 Apple Mail recipients with Mail Privacy Protection enabled (or 92% of a typical all-mail client email list) will view images that are somewhere between real-time to less than two minutes old when compared to other email clients.
3. Fewer than 5% of iOS 15 Apple Mail openers with Mail Privacy Protection enabled (or 2.6% of a typical email list) will view images older than 144 minutes.

[Update Oct 18, 2021]: In distinguishing iOS14 traffic from iOS15 traffic from all other traffic, we use the device’s User-Agent header. We detect User-Agent set to “Mozilla/5.0” as iOS15 and use the presence of “OS 14_” for iOS14. In April 2021 we observed approximately 2,000 “Mozilla/5.0” impressions per day, where as of October 2021 we regularly observe billions of “Mozilla/5.0” impressions per day. In order to better control for “Mozilla/5.0” impressions that may not be iOS15, we have updated our reporting to only include open/click pairs where the click event contains “OS 14_” / “OS 15_”. This removed approximately 10% of open/click pairs from the dataset, which we believe were likely cross-device and not representative. This methodology change accounts for the difference in results as of Oct 18, 2021.