How to clean analytics before analysis?

Raw email analytics are almost never clean enough to analyze at face value. Before you draw conclusions from open rates or click data, there's usually a layer of noise to remove, because some of what looks like engagement isn't human engagement at all.

The main sources of dirty data

Apple Mail Privacy Protection. Since 2021, Apple pre-fetches tracking pixels for Apple Mail users, registering opens even when the recipient hasn't read the email. If a significant share of your audience uses Apple Mail, your open rate is inflated. It may be meaningfully inflated.

Security gateway bot clicks. Corporate email filters, like Proofpoint and Microsoft Defender, scan links before delivery. These show up as clicks in raw event logs, sometimes milliseconds after send. Clicks before an open is recorded, or clicks on every single link in a campaign simultaneously, are almost certainly machine activity.

ISP probing. Some ISPs retrieve email content automatically when evaluating it. This can trigger opens and occasionally clicks without any human involvement.

How to clean it

For opens: filter or discount Apple MPP signals. Some ESPs let you exclude known Apple MPP events. If yours doesn't, use click rate as your primary engagement metric instead of open rate, since clicks require an actual action.

For clicks: filter out clicks that happened before a corresponding open, clicks that fired on every link in one session, or clicks from known security scanning user agents. Your ESP may offer bot filtering as a setting. If not, look for these patterns in your event-level data and exclude them before analysis.

For comparisons across campaigns: apply the same filters consistently. A cleaned open rate from one campaign compared against an uncleaned rate from another tells you nothing useful.

If you're not sure what's noise and what's signal in your data, the SOS hotline is free. Getting this wrong means optimizing for bot behavior, which isn't a useful direction.