Email Marketing KPIs: The Open Is a Download, So Track These Instead

The email marketing KPIs that still measure shoppers are the ones a person has to cause: bot-filtered clicks, placed orders, revenue per recipient, unsubscribes and spam complaints. Open rate no longer qualifies, because Apple Mail Privacy Protection downloads an email’s remote content, tracking pixel included, whether or not anyone reads it.

An open, as the platforms define it, is a download. Mailchimp’s help center says it plainly: “Mailchimp loads a tiny, transparent image into each email and counts how often the image is downloaded for delivered emails.” Apple’s privacy page for the feature says that, when an email arrives, Protect Mail Activity “downloads remote content in the background by default”, “regardless of whether you engage with the email”. Put those two sentences side by side and the open rate is counting Apple’s servers as readers. This guide reads Apple’s, Klaviyo’s, Mailchimp’s, Google’s and Yahoo’s own pages as served on October 12, 2026, and sorts the usual KPI list into the numbers that survive and the one that does not. It is written by Relvino, a company whose product decides per shopper across email, SMS and pop-ups and is measured on these same numbers, so treat it as informed but interested.

What an “open” actually counts

Every open-rate figure on every dashboard rests on the same mechanism. The sending platform puts a tiny image in the message, hosted on its own server. When a mail app fetches that image, the platform records an open. The method never observed a person reading anything; it observed a file request, and for most of email’s history a file request was a fair stand-in for a person scrolling down a message.

Apple broke the stand-in on purpose. Its June 7, 2021 announcement said: “Mail Privacy Protection stops senders from using invisible pixels to collect information about the user.” The privacy page explains what the tracking used to reveal (“Email senders may learn when and how many times you opened their email”) and how the feature ends it: instead of fetching remote content when a message is opened, Mail fetches it in the background when the message arrives. The iPhone user guide states the result as a design goal: “It also prevents senders from seeing if you’ve opened the email message they sent you.”

The email platforms describe the same event from the other end. Klaviyo’s email metrics reference says Mail Privacy Protection “changed the way that we receive open rate data on your emails by prefetching our tracking pixel”, and adds that “open rates will be inflated”. Mailchimp files it under bot activity. Klaviyo says it cannot tell a human open from an automated one when the feature is on, and Mailchimp’s answer is to filter those opens out of reports rather than recover the reads inside them, because the feature was built so that no sender could.

That is the distinction worth holding onto. A noisy metric can be cleaned. A metric whose signal has been deliberately removed at the source cannot, and the open rate for an Apple Mail reader with the feature switched on is the second kind.

Why filtering does not rescue the trendline

Platforms have responded by filtering, and the filtering itself changes the number. Mailchimp’s article on bot activity says “Bot filtering is enabled by default.” and warns: “Your open and click metrics will drop significantly when you turn bot filtering on.” It also publishes the dates from which filtered history exists: July 10, 2024 for Apple Mail Privacy Protection opens and SMS clicks, and August 10, 2025 for email clicks. A store that compares this autumn’s open rate with a pre-2024 autumn is comparing a filtered measurement with an unfiltered one, and the change in method can be larger than any change in the shoppers.

Filtering removes opens the platform can identify as automated. It cannot put back the human opens hidden inside them, because Apple’s design downloads the remote content when the message arrives, whether the shopper reads it that minute, next week or never. A filtered open rate is therefore a count of opens from people not using the feature, which is a slice of the list defined by mail app and settings rather than by interest. It can still be useful as a narrow signal. It is no longer the reach number the KPI list was built around.

There is a knock-on effect worth naming. Subject-line tests judged on opens, send-time settings tuned on opens and “engaged” segments built on opens all inherit the same problem. The send time optimization guide covers why the hour of an open and the hour of a purchase are different questions; Mail Privacy Protection makes the first one harder to observe at all.

The test a KPI has to pass

The practical rule follows from the mechanism. A KPI still measures shoppers when its numerator needs a human act that Apple’s background fetch cannot imitate: a click on a link, an order, an unsubscribe, a spam report. The iPhone user guide marks the boundary itself: “Mail Privacy Protection does not extend to links or email attachments such as PDF or other files.” Links still reach the sender when a person taps them.

Clicks are not perfectly clean either. Mailchimp says security tools “often click links or open emails before your contacts do”, and that its system “automatically tracks clicks as opens, even when a click is from a bot or spam filter.” So clicks pass the test with a condition: read them with bot filtering on, and treat a click rate that jumps across every segment at once as a scanner, not a sudden burst of interest.

Mailchimp points the same way. The bot activity article includes a recommendation in its list of things to know: “In addition to opens and clicks, we recommend that you use additional metrics such as bounces, unsubscribes, and conversions to measure the performance of your emails.” A platform asking its customers to read bounces, unsubscribes and conversions alongside its own open and click counts is a fair summary of where the open rate now stands.

The email marketing KPIs worth tracking for an ecommerce store

For a store, the KPIs that pass sort into two groups. Three measure whether email made money. Three measure whether email is wearing out its welcome, which is what decides whether it can keep making money next quarter.

Revenue per recipient

Klaviyo’s help center gives the definition: “RPR or Rev/Rec is determined by the revenue generated from your campaign divided by the number of deliveries in that campaign.” Its flow benchmarks use the same construction: “Revenue per recipient is the revenue attributed to flow emails divided by the number of delivered emails.” Both halves of the ratio exist without anyone opening anything: deliveries are recorded when the receiving server accepts the message, orders by the store. That is why it is the strongest single number on the list, and why Klaviyo’s published campaign and flow benchmarks are tables of revenue per recipient by store size and average order value rather than of open rates.

One limit travels with it. The numerator is attributed revenue, and attribution rules decide which orders count. The incrementality testing guide reads those rules in the vendors’ own words; revenue per recipient is only as causal as the attribution under it.

Placed order rate

Placed order rate is the share of delivered emails followed by an attributed order. It is revenue per recipient with the basket size removed, which makes it the better number for comparing a discount email with a full-price one: a discount can lift revenue per recipient by moving more units while the share of shoppers who bought stays flat. Read the two together. When revenue per recipient rises and placed order rate does not, the gain came from basket size, not from reaching more buyers.

Click rate, bot-filtered

Clicks are the earliest signal that still involves a person, so they are the right number for comparing two subject lines or two layouts, where waiting for orders would take too long. Keep the filter on, compare like with like, and do not compare a filtered click rate with one from before the platform began filtering.

Unsubscribe rate

An unsubscribe is a cost, but it is the cheap kind. Google’s sender guidelines put it bluntly: “Letting people opt out of your messages can improve open rates, click-through rates, and sending efficiency.” Yahoo’s requirements for bulk senders say “Honor unsubscribes within 2 days”. A store should watch the trend, not chase the rate to zero: an unsubscribe is a shopper using the exit the brand built, and the alternative exit is the spam button.

Spam complaint rate

This is the one KPI with a published ceiling set by someone other than the sender. Google’s guidelines say: “Keep spam rates reported in Postmaster Tools below 0.10% and avoid ever reaching a spam rate of 0.30% or higher.” Yahoo’s says: “Keep your spam rate below 0.3%”. Above those lines the cost is not one bad campaign but worse placement for every later one. The email list cleaning guide covers what lowers it; for KPI purposes the point is that it belongs on the same page as revenue, read weekly, because a campaign can win on revenue per recipient and lose on complaints in the same send.

Bounce rate and list growth

Bounces measure the list rather than the message, and a rising hard-bounce rate usually traces to how addresses are collected. List growth (new subscribers minus unsubscribes, bounces and complaints) is the KPI for sign-up forms and referral emails, and it is the only one on this list that measures email’s future audience rather than its current one.

• Open rate · What the number counts: Tracking-image downloads · Needs a person to act?: No, for Apple Mail readers with the feature on · What the source page says: Apple: content downloads “regardless of whether you engage with the email”
• Click rate (bot-filtered) · What the number counts: Link clicks, minus identified bots · Needs a person to act?: Yes, with a scanner caveat · What the source page says: Apple: protection “does not extend to links”
• Revenue per recipient · What the number counts: Attributed revenue per delivered email · Needs a person to act?: Yes (an order) · What the source page says: Klaviyo: revenue “divided by the number of deliveries”
• Placed order rate · What the number counts: Attributed orders per delivered email · Needs a person to act?: Yes (an order) · What the source page says: Counted from store orders, not mail events
• Unsubscribe rate · What the number counts: Opt-outs per delivered email · Needs a person to act?: Yes · What the source page says: An opt-out request, which a background fetch cannot make
• Spam complaint rate · What the number counts: Spam reports per delivered email · Needs a person to act?: Yes · What the source page says: Google: bulk senders must stay below 0.30%; it advises below 0.10%

Reading the KPIs as pairs, not a list

A KPI list invites ranking. The surviving numbers are more useful read in pairs, because each pair answers a question a single number cannot:

• Revenue per recipient with spam complaint rate. Is the money coming at the expense of next month’s inbox placement? Revenue up with complaints up is revenue borrowed from later sends.
• Revenue per recipient with placed order rate. Did more shoppers buy, or did the same shoppers buy more? The two call for different next campaigns.
• Click rate with unsubscribe rate. Did a new format interest people or annoy them? A click lift with a matching unsubscribe lift is a format that works on a smaller list.
• List growth with bounce rate. Is the sign-up source adding real shoppers? Growth that arrives with bounces is growth that will have to be cleaned out.

The open rate can still sit at the bottom of the sheet for one job: spotting a deliverability failure. When filtered opens from non-Apple mailboxes fall sharply across every campaign in the same week, the problem is placement, not content. That is a smoke alarm, not a performance measure, and it should be labeled as one so nobody grades a subject line on it.

The question to ask about any report column is the one this guide has been applying throughout: what event created the number, and could that event have happened with nobody there? If the answer is yes, the column describes the infrastructure. If the answer is no, it describes shoppers, and those are the columns that belong in a KPI review.

Where Relvino fits

Relvino decides per shopper, in under 80ms, whether to send anything at all and on which of email, SMS and pop-ups. A message it declines to send cannot earn a complaint or an unsubscribe, which is why the cost KPIs matter as much as the revenue ones when judging it. A brand evaluating it in a 14-day pilot can keep the KPI sheet above unchanged: revenue per recipient and placed order rate for the money, complaint and unsubscribe rates for the cost. For the wider argument about whether the channel itself is worth the effort, the batch’s companion piece asks whether email marketing is dead, and the customer retention rate guide covers the store-level number these email KPIs eventually feed.

A KPI review that starts by deleting the open-rate column, or by moving it to the bottom as a deliverability alarm, is not losing information. It is admitting what the column was measuring after September 2021, and making room for the numbers a person has to cause.

Frequently asked questions

What are some examples of marketing KPIs?

Beyond email, the common ones for an ecommerce store are customer acquisition cost, return on ad spend, conversion rate on the site, average order value, repeat purchase rate and customer lifetime value. The useful habit carries over from email: for each one, ask which event creates the number and whether it is counted by the store (orders, refunds) or by a platform reporting on itself (impressions, attributed conversions). Store-counted KPIs are harder to inflate.

Which KPI can tell you about the performance of the referral email designed to build your subscriber list?

The referral sign-up rate: new subscribers who joined through the referral link, divided by referral emails delivered. Clicks show interest, but only a completed sign-up shows the email did its job. Check the same group again a month later, because referred subscribers who never buy, or who mark later emails as spam, add list size without adding audience.

What is a good open rate after Apple Mail Privacy Protection?

There is no meaningful general answer, because the share of Apple Mail readers with the feature on differs from list to list, and filtering methods differ by platform. The share is also not the same as the share of Apple addresses on the list: Klaviyo notes that a Gmail account read in Apple Mail with the feature on still produces affected opens, so a list’s domain mix does not predict how inflated its open rate is.

Can Apple privacy opens be separated from other opens in Klaviyo?

Partly. Klaviyo’s help center describes custom reports that add Total Apple Privacy Opens and Unique Apple Privacy Opens, and a single metric report filtered on an Apple Privacy Open property set to true or false. Klaviyo notes the property only exists for opens on or after November 20, 2021, and that it cannot tell a true human open from an automated one on those devices, so the false group is the cleaner open signal.

How much data is needed before email KPIs mean anything?

Klaviyo recommends at least a month’s worth of campaigns to get an accurate baseline for revenue per recipient. Complaint rate works on a shorter clock: Google’s guidance is to monitor it regularly in Postmaster Tools, so a weekly check is the floor for any brand sending at volume. Compare each KPI with the store’s own baseline first and a published benchmark second.

How long does it take to migrate to Relvino?

The technical cutover takes about 30 minutes: connect the Shopify store, point the sending domain, connect SMS, and shopper data ingests automatically. Revenue is then measured in a 14-day pilot. Export the old platform’s campaign and flow reports before switching, so revenue per recipient, placed order rate and complaint rate have a baseline to compare against.