First of all, what are we talking about exactly?
On September 20th, Apple has released the latest update of its mobile OS, iOS15. This update includes new features favoring the confidentiality and the data privacy protection of users, especially in Mail, the email application of Apple. Here is how Apple describes this feature called MPP,
Mail Privacy Protection: “the Mail Privacy Protection hides your IP address so senders can’t link it to your other online activity or determine your location. It also prevents senders from seeing if you’ve opened the email they sent you.”
What have we observed since the update?
As this was a new update, it was unclear how much impact it would have on open data. Since the release, this is what we have seen:
1 / For users who have updated their Apple Mobile devices and accepted
MPP, all of their emails will be reported as having been opened because they will have been preloaded directly by Apple. Thus, you will not be able to distinguish the real opens from false ones. Likewise, the opening data transmitted to Tinyclues will also be falsified.
=> The first consequence that you will have is a false (artificial) increase in the open rate which you probably have already noticed.
2 / The speed at which you observe this change will vary depending on your customer database, the proportion of Apple Mobile device usage, and the proportion of Apple Mail application usage (users who open their email on other applications, as Native Gmail App, on iPhone, are not affected). We have observed that our clients who looked into these metrics received very different results on usage proportion depending on the marketer. One of the North American luxury marketers had 70% openers on Apple Mail; while around 14% for a French female garment marketer.
That being said, we often find that CRM teams don’t know this information. We strongly recommend you turn to your ESP to further investigate this question.
For example, here is the open rate trend of a marketer who has 70%
customers on Apple Mail.
Eventually, the impact on your open rate will depend on the speed of iOS15 deployment.
3 / According to the information of Mixpanel, the iOS15 adoption rate on November 15th was around 43%. The iOS15 adoption at the beginning of its release was much slower than iOS14, but has picked up in recent weeks.
Thus, if the first consequence of this update will be a false increase in the open rate for you, the speed at which you will observe this change will vary depending on your customer database and on iOS15 adoption rate. However, whatever your situation is, the more time passes, the less reliable your open rate will be.
How important is open data for Tinyclues models?
As you already know, Tinyclues' models rely on your first-party data: user table, purchasing records, reactions to the email campaigns (open, click, unsubscription), and, in some cases, website navigation.
The Tinyclues predictive models seek to predict purchasing, in other words, who is going to buy what today or at the moment when you send your campaigns. Also, logically the data that are the most important for the models are the purchasing records such as transaction records and information linked to purchased products (brand, category, etc.), followed by anonymous data of customers.
=>Thus, we found that the email open data is less important for the quality of purchase prediction in the models.
However, these data are still essential in predicting engagement in an email campaign. Actually, when you choose the campaign channel on the platform, Tinyclues adapts its predictive models accordingly to take the channel’s appetency into account. Nevertheless, opening and clicking data provide strong signals for engagement in the email channel.
Our ability to predict who is going to buy what is thus hardly impacted by the email data, on the other hand, this data remains important for email campaigns because they allow us to ensure the appetency to the email channel, meaning to have a higher chance for future buyers to be exposed to the campaign.
In other words, we are able to know, among a crowd of passers-by, who has a higher possibility to purchase a product promoted in your store, but the email data also allows us to predict who is going to enter the store.
What is the impact of those changes on Tinyclues?
Now we know that the impact of the update is reporting false open data for a certain number of customers and that on Tinyclues’ side, those data are in particular used for optimizing the email campaign engagement.
For now, the update of iOS15 has limited impact because the adoption rate is still lower than 50% at the moment, and we use record-based email data. Also, only a small part of the opening data is false for now.
What about tomorrow though? We have experimented in recent weeks with several scenarios on our side to predict what will be the impact on the model if a majority of opening data is false.
At this stage, our conclusion is that it will be more efficient to continue to use the opening data rather than ignore them. Extracting all or a part of opening data from the models actually risks too much excluding the email-engaged customers, and therefore sacrificing campaign performance.
Consequently, we can expect that a small proportion of Apple Mail users might be contacted because they are mistakenly considered as openers.
In the short term, we are working on new predictive models specializing in email campaigns that will be served as a bridge between the use of opening data and clicking data in order to limit the impact of soliciting none openers on Apple Mail.
We will let you know about these developments at the start of the year.
What are our short-term recommendations?
For now, we recommend some immediate solutions that are to adapt your
campaign KPIs and your rules for evaluating email activity, and to inform
internally about the impact on indicators based on the opening.
Regarding the indicators that you should be particularly vigilant:
Open rate – Keep track of the trend of your open rate to observe the progressive impact of iOS update. For this, we think that a scatter graph can
be particularly helpful to avoid the effects of average.
Conversion rate – With the open rate falsely increasing, your conversion rate is likely to drop. In the same way, start to follow the trend of this indicator and inform internally about the impact so that the changes can be expected.
Click rate – In the future, it will be your best indicator for your email campaign engagement. Also, ensure the quality of your tracking and optimize your call to action
Attribution – If your attribution model is based on opening, also keep track closely of your RPM and think about adding a second attribution model based on click to your reporting in order to compare the trends of those two different models. This will allow you to identify the gap and when it will take place (More than 20 different attribution rules are available on your Tinyclues Campaign Performance
Finally, the opens are regularly used for defining active customer databases, meaning those who will receive most of your communications. If you are using criteria like “customer has opened an email in the last 6 months”, be careful with the change of size of your active customer database, which will grow within several months because of the false increase in opening data. It would be great to start to add criteria defined by click, especially for Apple Mail users. You can also use the generic scoring of Tinyclues to sort your entire customer database and no longer rely solely on opening criteria for your large volume