Personyze Wiki Personyze Wiki docs
English
Open Personyze
Docs/ Email Campaigns/ Subject Tests — Reading the Result
Email Campaigns

Subject Tests — Reading the Result

How to read an email’s subject test on the Performance step: the table, the verdict line, the automatic winner and resuming the test.

4 min read Updated 6 hours ago

A subject test sends one email with two or three subjects, split evenly between its recipients, and shows which subject gets more clicks. You set it up in the email editor — see Writing the email. This page is about reading the result.

Judge on clicks, not opens. Apple Mail opens every email it receives, read or not, so open rates overcount. The verdict and the automatic winner both compare clicks per email sent.

Where the result is

On the campaign’s Performance step, General tab, inside By email, under the rows: a Subject test card for each email that has had one, named after the email. The same card is on Analytics › Email & SMS with a single campaign picked.

The subjects are lettered: A is the email’s own, B and C the ones added with Test another subject and Add a third subject. Each person always gets the same subject, on a resend too.

The table

Column Shows
(letter) A, B or C.
Subject The subject. One taken out of the test reads (removed from the test).
Sent Messages sent with this subject in the range.
Click rate Messages with at least one click, as a share of messages sent. One per clicked message, so it is not the same figure as the campaign’s Click rate card, which counts every link click.
Open rate Messages opened, as a share of messages sent. Its header warns: “Overcounted: Apple Mail opens every email it receives, read or not, so the winner is judged on clicks.”
Conversions People who clicked through and then reached a goal within a day.

The leading row is highlighted with Most clicks once a subject leads with enough confidence, or Winner once one has been picked.

A person whose test subject comes out empty — for example a subject that is only ${first_name}, for someone with no first name — is sent subject A but still counted under the letter they were dealt.

The verdict line

One line over the table says where the test stands. It uses the same statistics as the site’s A/B tests.

Line Means
Too early to call: wait until each subject has been sent at least 100 times. A subject still in the test has fewer than 100 sends in the range.
No subject is ahead on clicks yet. The subjects are level on clicks.
B leads on clicks – 99% sure The leader has reached the email’s confidence threshold.
B is ahead, 71% sure – not enough to call yet A subject is ahead, but below the threshold.

The threshold is the confidence set under Pick the winner automatically — 98% unless you changed it — and it applies even when automatic picking is off. With three subjects, the leader is compared with the other two together, and the confidence allows for having picked the best of three.

Picking the winner automatically

With Pick the winner automatically switched on in the email editor, everyone after the pick gets the winning subject. Its line reads When one subject is [98% (as A/B tests)] sure, after each has been sent [1,000] times:

Setting Choices Default
Confidence 90%, 95%, 98% (as A/B tests), 99% 98%
Sends of each subject 250, 500, 1,000, 2,500, 5,000 1,000

Until it picks, the report adds: The winner is picked automatically once it is 98% sure and each subject has been sent 1,000 times. Every subject still in the test must reach the sends, and the leader must be ahead and at least that sure.

Once it picks, the verdict line becomes Winner: {letter} (picked automatically on {date}, {pct}% sure) – everyone now gets {letter}.

Resuming the test

Resume the test, beside the winner line, splits recipients between the subjects again. It also switches Pick the winner automatically off, so the same subject is not picked again at once; switch it back on in the email’s editor when you want it. Resuming does not create a new version of the email.

Reading it well

  • Wait for the sends. Below 100 sends of each subject the difference is mostly chance, and a small lead can turn around.
  • A test compares subjects of one email. To test whole emails against each other, use the campaign’s A/B test.
  • Only email types with a preheader offer a subject test; each test subject can have its own preheader.
  • The table counts by send time within the range. Change the range on the Performance step to see an earlier or later stretch of the test.

}

Did this page answer your question?
Thank you — that goes to whoever maintains this page.