Human opens vs. pre-opened emails

Industry changes, particularly iOS17 have introduced the concept of 'pre-fetched opens' or 'pre-opened email'.

Pre-fetched opens are cases where the recipient's inbox automatically opens an email as soon as it is delivered.

This results in higher than expected open metrics which are artificially inflated.

Human opens

To make sure you're getting the most accurate engagement data, Ometria has two open metrics:

• Opens - All opens, including those pre-opened by the recipient's email provider
• Opens (human) - Email opens which Ometria can verify were performed by your contacts and customers (humans), excluding pre-opened.

You can choose whichever metric you prefer. Human opens show real opens, but they only cover part of your audience. See ‘How reliable are human opens?’ below. 

The human metrics are:

• Emails opened (human)
• Email open rate (human)
• Email click to open rate (human)
• Email order to open rate (human)

You can find these in the following reports:

• Email performance in the Overview dashboard
• Campaign performance for automation and broadcast campaigns.
• Individual automation campaign reports - overview and flows tabs.
• Individual broadcast campaign reports - contact analytics and variants tabs.

How does Ometria know if an open is 'human'?

An email open is registered when a mailbox provider or email client, e.g. Gmail or Apple, download a specific image pixel embedded in the email.

When they download this image they also send information back to the sender about the device requesting the download. This information usually includes data about the device type, e.g. mobile, desktop, tablet.

When Apple introduced Apple Mail Privacy Protection they restricted the data being sent back, so that senders can only see that an Apple server is requesting this information and nothing else.

So, when Ometria’s email sender (SparkPost) sees that an Apple server with no other information has downloaded the image, it is identified as pre-fetched, i.e. non-human.

SparkPost then identifies any opens which are not pre-fetched as human.

What if a human later opens a pre-opened email?

Unfortunately, once an email is pre-fetched/pre-opened by the mailbox, SparkPost can’t tell if any subsequent opens are human.

This means that there may be pre-fetched opens that are also human opens. Ometria has no way of getting this data.

We recommend that you look at your ‘human opens’ as a trend, not an absolute number.

How reliable are human opens?

Human opens are reliable in one narrow sense. The opens in the human figure are very likely from real people.

But human opens only cover contacts whose device or mail app doesn't pre-fetch emails. For many retailers, that's a small share of the list.

This has three effects on how you read the figure:

• It undercounts real opens. Some people open emails that their mailbox already pre-fetched, and these opens don't count as human.
• It can move a lot. Because the base is small, a modest change in behaviour can shift the percentage noticeably.
• It doesn't say why it moved. A drop could come from a deliverability problem or lower engagement. It could also come from contacts changing their phone, mail app or privacy settings.

So use human opens to watch trends over time. Don't rely on them alone to measure engagement or inbox placement.

For a fuller picture, use these alongside human opens:

• Opens (total) for inbox placement. Pre-fetching inflates the number, but it only happens when an email reaches the inbox rather than spam. A sharp fall in total opens over time can point to a placement problem.
• Clicks and orders for engagement. Clicks can include automated activity, such as email security scanners. Orders are the strongest sign that people engaged with your email.

Why do human opens differ between mailbox providers?

Each mailbox provider handles email images differently. So the share of opens that count as human varies a lot by provider.

• Gmail loads images when the recipient opens the email, so Gmail opens are more likely to count as human.
• Microsoft (Outlook, Hotmail) often produces no trackable human open. Some Outlook setups block images by default, and many people read Outlook mail in Apple Mail.
• Apple Mail pre-fetches emails when Mail Privacy Protection is on, so these opens don't count as human.

Ometria works out the provider from the contact's email domain, not the app they read it in. A Gmail address read in Apple Mail counts as Gmail.

A company address hosted by Microsoft doesn't show as Microsoft either. This adds uncertainty to any breakdown by provider.

The figure for one provider can change suddenly without any change in how your contacts engage. This can happen when a provider changes how it loads images, or when contacts switch mail app.

If you've turned on open tracking consent for a country, Ometria doesn't track opens for contacts there who haven't consented. This lowers all open figures for those contacts.

Before you compare a provider's human open rate year on year, check whether total opens, clicks and orders moved too. If they didn't, the change is more likely in tracking than in engagement.

Send time optimisation

Our optimise send time feature in broadcast campaigns bases calculations on human opens only.

This is because pre-fetched opens are not useful when selecting a contact's preferred open time. 

If there isn't enough open data for a contact, the sent time optimisation model applies your account average.