
How to measure email marketing after Apple Mail Privacy Protection
Open rates stopped meaning much in 2021. Which email metrics to trust now, what good looks like by email type, how attribution works and how to test properly.
The Wizel Team
Email marketing strategist · · 8 min read
On this page
- What Apple Mail Privacy Protection changed
- The metrics to trust instead
- Rebuild engagement without opens
- Benchmarks by email type, as rules of thumb
- Attribution, and why email revenue won't match finance
- A/B tests that give real answers
- How to set up email measurement after Apple Mail Privacy Protection
- Frequently asked
Since Apple Mail Privacy Protection arrived in 2021, a recorded open no longer means someone read your email. Plenty of stores still pick test winners, build engaged segments and judge campaigns on opens. This chapter covers what to measure instead, what good looks like by email type, how email revenue squares with what finance sees, and how to run tests that give real answers.
50-80%
the open rates many lists jumped to after Apple's privacy change, read or not
What Apple Mail Privacy Protection changed
Apple Mail loads images, tracking pixel included, through a proxy, so an open gets recorded whether or not a person saw the email. It depends on the mail app, not the address: a Gmail or Outlook address read in Apple Mail is affected too, so you can't assume a fixed share of your list. The symptoms are open rates that jumped and stayed high, opens that no longer move with clicks or revenue, and “opened” segments full of people who never did anything else.
Four common setups broke with it:
- A/B tests that pick the winner by open rate.
- Flows and conditional splits triggered by opens, which now let in people who never read.
- Engaged segments built on opens.
- Send-time optimization based on when opens happened.
Opens keep one job: a sharp, lasting fall can point to a deliverability problem, and a campaign open rate below 20% suggests spam filtering. A very high one, 70% and up, more often means the audience is too narrow than that the copy is brilliant.
The metrics to trust instead
Rank your metrics. Decide on total revenue and orders, and on contribution margin if you know it. Explain the result with revenue per recipient, placed order rate and order value. Diagnose with bot-filtered clicks. Read opens last.
| Metric | What it measures | How to read it |
|---|---|---|
| Total revenue and orders | What the send or flow sold | The outcome. Compare with your own history at the same attribution settings. |
| Revenue per recipient (RPR) | Revenue divided by emails delivered | Compares sends and flows of different sizes. Watch total revenue beside it: shrinking the audience can raise RPR while total revenue falls. |
| Placed order rate | Orders divided by emails delivered | Klaviyo's default conversion rate, and small by design for campaigns. Always say which denominator you mean. |
| Click-to-conversion | Orders divided by unique clicks | Whether the email and offer persuaded the people it reached. |
| Click rate | Share of recipients who clicked | Your best read of attention once bot clicks are filtered out, since security scanners click links too. |
| Unsubscribe rate | Unsubscribes per email delivered | Keep it below 0.3% per send and look into anything over 0.5%. Track buyers who unsubscribe separately. |
| Spam complaint rate | Complaints per email delivered | Keep it below 0.1%. Even 0.2% causes deliverability problems. |
| Open rate | Recorded opens, many automatic | A rough deliverability signal. Never a test winner, a trigger or a target. |
Decide on revenue and orders. Diagnose with clicks. Treat opens as a smoke alarm, not a scoreboard.
Watch buyer unsubscribes closely. Most unsubscribes come from people who never bought, often in their first week on the list. A steady headline rate can hide a rise among buyers, who are far costlier to lose. The guide to list health numbers covers these guardrails in more depth.
Rebuild engagement without opens
Define an engaged subscriber by what they do: a bot-filtered click, site activity or a purchase within a window that suits how often your customers buy. 60-90 days is a common starting point; a store people buy from once every few years needs longer. Keep filtered opens as supporting evidence only, and never suppress people just because they lower your average click rate.
Two Klaviyo details matter. It filters bot clicks, but not the same way in every report, so a custom report may need an explicit Bot Click = false filter. And when you do look at opens, the Apple Privacy Open flag lets you separate the machine ones. Swap any open-based flow split for behavior: clicked, viewed a product, placed an order.
For deliverability without opens, watch complaints, bounces and unsubscribes against comparable sends, check that SPF, DKIM and DMARC pass, and use Google Postmaster Tools for Gmail's spam rate (check what it currently offers; Google keeps changing it). Seed tests are supporting evidence from a sample, not proof for every recipient. The deliverability chapter covers the rest.
Benchmarks by email type, as rules of thumb
Your own history is the first comparison. A peer figure only means something when the vertical, email type, period and metric definition match. Use two reference points at most, the median and the top 10%, and treat a gap as a question to investigate, not a verdict.
| Email type | Placed order rate (per recipient) | Revenue per recipient |
|---|---|---|
| Campaign to the whole list | 0.04-0.08% | $0.10-0.50 |
| Campaign to a tight engaged segment | Up to 0.3-1% | Varies |
| Welcome flow | 2-5% | $2-5 |
| Abandoned cart flow | 3.33% average, 7.69% top 10% | $3.65 average, $28.89 top 10% |
| Post-purchase flow | 1-3% | Varies |
Campaign click rates tend to sit at 0.5-2%. Flows convert far better per recipient because they reach people at a moment of intent, so judge each against its own kind, never a flow against a campaign.
Separate how good a result is from how it changed. A profitable campaign can still be down on last year's big sale. And when a blended number moves, check the mix underneath: average order value can fall overall while rising for both new and returning customers, if a campaign brought in more new ones. The guide to comparing email results over time goes deeper.
Attribution, and why email revenue won't match finance
Klaviyo credits an order to the last email opened or clicked inside a lookback window: 5 days for email on accounts created since October 2024. You can count clicks only, shorten the window, and exclude bot clicks and Apple Privacy opens. The Model comparison tool previews the effect, and changes apply to past data too. The guide to email-attributed revenue covers the mechanics.
The setting moves the answer a lot. In one practitioner's comparison, email was about 40% of store revenue on the default open-and-click window and 5-10% on a 1-day click window. Neither is what email caused. Attributed revenue is captured revenue: some of those buyers would have ordered anyway. If email's attributed revenue climbs while store revenue stays flat, email is mostly touching people who were buying already.
Only a holdout measures what email adds. In one practitioner's replenishment flow holdout, 15.5% of recipients bought against 15% of the group that got nothing: half a point of real lift, for a flow credited with far more. Flows that reach people already on their way to buying capture more than they create.
To square email with finance, use finance's numbers. Finance counts contribution margin: revenue net of discounts, minus cost of goods, shipping, payment fees and returns. Reconcile every tool against Shopify, and Shopify against the bank. Show retention as Shopify reports it, in second and third orders and lifetime value, rather than as an attributed share. A subscription store should report against a Placed Order metric that excludes renewals, built as a Klaviyo custom metric, because renewals bill whether or not anyone opened an email.
If you report to clients, put the attribution window beside every revenue figure, keep the brand's existing settings so this month compares with last, and report a month only after its window has closed (after the 5th on a 5-day window). Judge each client against its own history, never another brand's share. When a client insists on a 1-day window, warn them the share will drop and show the default equivalent in the Model comparison tool.
A/B tests that give real answers
A test gives a real answer when it changes one thing, is judged on orders or revenue, and runs until enough people have bought. In Klaviyo, set the winning metric to revenue, not opens, and keep the offer, content and timing the same across variants.
Count orders, not recipients: 100-200 orders per variant for a moderate read, 300-500 for a strong one. On recipients, 1,000 per variant is a floor and about 10,000 is comfortable. Small lists rarely separate on cosmetic changes like button color, so spend their tests on big swings: the offer, or plain text against a designed email. Do the arithmetic first. A flow with 100 entrants a day converting at 3% makes three orders a day, so a test that needs 150 orders per variant, 300 in all, takes 100 days. If that's too slow, test in campaigns, which reach far more people in one send, and carry the winners into flows.
A common campaign setup sends the variants to about 25% of the audience, waits about 4 hours, and sends the winner to the rest. On a smaller list, 4 hours often holds too few orders to separate the variants, so lengthen the window or treat the result as directional. Don't stop early on a hunch, and check the winner didn't raise unsubscribes or complaints. Test these first:
- The welcome offer, its type and size, judged on first purchases, not just sign-ups.
- The first emails of your welcome and abandoned cart flows, which earn most of each flow's revenue.
- Offers and formats in campaigns. A winning subject line dies with its campaign, but a winning pattern, like dollars against percent, carries over.
To learn whether a whole flow earns its keep, run a holdout: split entrants 50/50 at the top, send one side, tag both with a profile property, and after about 30 days compare all store spending per person, not attributed revenue. Aim for about 5,000 people per side; 1,000 is workable. Once any test has an answer, rebuild on the winner and remove the split. Klaviyo's Experiments tab lists every active and ended test, which is the quickest way to find ones left running for months.
This month, write down your attribution settings, switch every test's winning metric to revenue, rebuild your engaged segment on clicks, site activity and purchases, and start one holdout on your biggest flow. Then work through the 15 questions to ask your email data each month. Measured this way, the figure that matters most is what customers spend over time, which is the next chapter: email retention and lifetime value.
How to set up email measurement after Apple Mail Privacy Protection
- 1
Fix your attribution settings
Choose the window and what counts (opens and clicks, or clicks only), write it down, and keep it the same month to month.
- 2
Filter out machine activity
Use bot-click filtering in every report you rely on, and separate Apple Privacy opens when you look at opens at all.
- 3
Rebuild engaged segments
Base them on bot-filtered clicks, site activity and purchases within a window that suits your purchase cycle.
- 4
Judge tests on revenue
Set the winning metric to revenue and wait for enough orders per variant before calling a winner.
- 5
Report against your own history
Lead with revenue and orders, then revenue per recipient and placed order rate, after the attribution window has closed.
- 6
Run a holdout
Split your biggest flow 50/50 and compare all store spending per person after about 30 days.
Frequently asked questions
Are email open rates still accurate?
No. Since Apple Mail Privacy Protection in 2021, Apple Mail can record an open without anyone reading the email. Treat open rate as a rough deliverability signal, never as a test winner or a target.
What should I track instead of open rate?
Total revenue and orders first, then revenue per recipient and placed order rate, with bot-filtered clicks to diagnose. Keep unsubscribe and spam complaint rates as guardrails.
What is a good revenue per recipient for email?
As a rule of thumb, $0.10-0.50 for a campaign to the whole list and $2-5 for a welcome flow. Your own history at the same attribution settings is a better comparison than any benchmark.
How many people do I need for an email A/B test?
Count orders rather than recipients: about 100-200 orders per variant for a moderate read and 300-500 for a strong one. On recipients, 1,000 per variant is a floor.