iOS 15 & Email Marketing: What is really happening post release?
It seems a lifetime ago that Apple dropped the bombshell that they were killing off open tracking for anyone using the iOS default mail client in their new release iOS 15.
The method for doing so was claimed as pre-fetching all remote content in the email at the point of delivery, and as this would include the little pixel that is used to determine if the email was opened, all iOS 15 users would look like they have opened your email.
There were other claimed consequences, namely as the opens would then be cached on the device any real-time images such as countdown timers would be inaccurate, while those who use open data for predictions such as subject line optimisation, or for sending follow-up campaigns based upon whether they opened or not were going to be ‘dead’. Finally, the need to filter lists by overall engagement to assist with deliverability was going to be problematic.
But all of that was gleaned from press releases and our own little experiments on a handful of devices. Now it’s been released what is really happening with real-world data behind it?
At Reignite we capture large volumes of data from logs of our personalised images that are inserted into client campaigns, complete with what device requested the image. We’ve started using that data to drill into some of these key questions.
How big is iOS 15 right now?
OK we are only a week in at the time of writing so while iOS users might equate to 40-50% of your readership it’s going to take some time to roll out. Right now, we are seeing 6.63% of opens coming from iOS 15. Note: All of the data shown is UK centric – the scale of iOS 15 adoption may differ in other countries
This has been growing daily:
We would expect iOS 15 take-up to be around half of iOS users by the new year, with 90% coverage in 12 months’ time.
However, remember not all of these will be real opens but where Apple has pre-fetched the images.
Working out what the true impact on open rates is tricky as we don’t know who has received the email on an iOS device, just those whose device have opened on one. To understand if Apple is pre-fetching images every single every single time we’ve looked at a few things.
Does Apple pre-fetch every time?
We have a client that sends the same email nearly every single day (the content differs daily but the audience is the same each day). We’ve looked at all of those who opened one of these emails on iOS 15, and then was also sent the email again on each of the next 4 days.
If the assumption is that Apple will always pre-fetch images then you would expect all of these users to show an ‘open’ on each of the subsequent sends, with a tiny bit of leeway for those that have turned the device off.
Yet only around 1/3 of those recipients opened all 5. Clearly there is a lot of fetching as only 3.7% failed to open another email after that first one:
So, this tells us pre-fetching isn’t that consistent, and what it also tells us is there will still be a large percentage of iOS 15 users who aren’t giving us the open signal so you will be able to see when it’s 100% the case that they didn’t open the email. This can then help with those who wish to re-send with confidence to non-openers, or with engagement rules for deliverability.
When does pre-fetching occur?
During testing in beta I’ll be honest I personally was never able to get pre-fetching to happen, no matter how long I left it, whether I had wifi or plugged into a charger. So I was skeptical that once this went live you would see huge spikes of open volume at the point of delivery.
And that is what I’m still seeing with real data at scale.
To visualise this we have taken some client campaigns sent at a known time, and looked by the hour when opens occurred for both iOS 15 and non-iOS 15 opens. We only counted the first open by a recipient to avoid bias of re-opens on non-iOS openers.
If Apple was pre-fetching at the moment of delivery you would expect 2 things:
1) The percentage of the iOS 15 opens would be seen to primarily occur as the same hour of delivery
2) The percentage of all opens to be higher for iOS 15 in that first hour compared to normal opens
It’s actually the complete opposite we see.
In the first hour of the send only 13.6% of the total iOS 15 opens occurred. Compare this to non iOS 15 and 15% of all of those opens happen in the first hour.
This email was sent at 7am UTC (Which is currently 8am in the UK), and by around 4pm we had seen more than 2/3 of the non-iOS 15 opens that would occur, but only 50% of the iOS 15 opens.
Normally there would be little reason why emails sent at the same time would have the proportion of its total opens at different times – the patterns should match. And if Apple was doing pre-fetching at delivery you would see the proportions higher in the first hour or so for iOS 15 compared to others.
Where a lot of these ‘pre-fetches’ are occurring is actually later in the day when there is a high probability that the recipient has already ignored the email. These are where the majority of additional opens Apple generates that aren’t real opens.
We can also see this by looking at per hour, what proportion of the opens are from iOS 15.
In a world where all email clients operated the same you would expect the proportion of opens from iOS 15 to remain consistent over time. Yes, opens would be higher overall but the proportions should be the same. In a world where pre-fetching occurs at the point of delivery you would expect the proportion of iOS 15 to be highest in the first hour.
Yet we see the opposite. There is even a spike at 10pm UK time through to midnight UK time where a much higher proportion of opens are from iOS 15. My bet is this is when we all head to bed and the first thing we do is plugin our phone as the wonderfully researched article by Justin Khoo suggests is a key reason why Apple begins the pre-fetch.
But why is this important you might ask?
Well, it fits with my original experience that true pre-fetching (that being before I open an email) doesn’t happen that often. If you open the email in close proximity to when the email is sent there is a very high likelihood this is a real open.
Most of the additional opens generated are likely to be those who aren’t going to open the email at all anyway.
On the back of this finding I think there are 4 key areas to clear up.
1) Real-time Email Content is still valid
For real-time personalisation or optimisation of content the chances this means the vast majority of iOS 15 users who really open your email will get served content very close if not real-time to when they are viewing.
2) Send-time Optimisation is not dead
For a send-time optimisation point of view I believe the patterns in data can be managed to still make effective predictions. As with all send time optimisation it’s the quality of the algorithm, and the people managing the algorithm. With some adjustments and attention I still see this being a viable method using open data in conjunction with time of send.
3) The impact on engagement segmentation
I’ve seen enough now to have hope that there is enough signals in the data sets for someone in the deliverability industry to arrive at a prediction model for whether an open is really an open. When using this data at scale, over say 6 months worth of campaigns there should be enough gaps in open data to show non-opens, and an analysis of the open times to come up with an overall prediction score that equates to engagement. Yes, this is more advanced than just using ‘last open date’ but that is too simplistic by far anyway. This might just give you the reason you need to evolve that model anyway.
4) Comparing Open Rates
While there is an inflation of open rates these still make open rates comparable. Sure, you might not want to compare an open rate in 12 months time to now, but for making decisions like ‘did this subject line perform better than this one?’ or for general trends this will not be negatively impacted as its just a sample of your opens that are unreliable, but will be consistently unreliable.
While only a week into this new world we should be generally optimistic about the outcome of iOS 15 for email marketers. Yes, there is a lot of noise and complexity generated, and yes life would be easier if there wasn’t change but we are lucky that some of the innovations for email marketers such as real-time content, send time optimisation and engagement segmentation are not dead in the water.
As email marketers we should be striving for excellence anyway and all iOS 15 does is kick out some of the more simplistic offerings and force us to go that extra mile for the optimum approach.