I wrote two subject lines for notes that were otherwise the same length. One sounded like a person talking. One was tighter, with the interesting word moved to the front. I wanted the plain line to win, which is already a problem, and then the results refused to flatter the preference. That refusal is the useful part. The test is inconclusive. This page exists so I do not quietly turn a preference into a finding.
The hypothesis was simple. A subject line written the way I would say it out loud will be opened more often than a compressed, clever version of the same note. "More often" assumed a comparison I could trust. I did not earn that comparison. Writing the failure cleanly matters more than picking a style.
The plain line won on one send and lost on the next. The sample is too small, and the topics were not the same kind of note.
What I thought I was testing
Subject lines are a tiny room where attention is obvious. The reader gives you a glance. The glance either earns the open or it does not. I suspected that cleverness spends the glance on the writer's skill, and that a spoken sentence spends it on the topic. If that were true, the plain line would open more often when the note underneath was the same.
"The same" is the whole experiment. If the notes differ, you are testing topics, not lines. I knew that and still let the topics differ on the second pair because I had two notes ready and did not want to wait. Impatience is how inconclusive tests are manufactured. I am documenting the manufacture so I recognize it sooner.
I also did not hold the list still in any way I would call serious. The groups were small. Small groups can be swung by a handful of habitual openers. A win in a handful is a story. I like stories, and I do not get to call them results. The bias note is partly about this: availability will offer you the send you remember, not the pair.
What happened
The first pair favored the plain line. The second pair favored the tight line. I do not have a result I trust. The topics differed, the lists were small, and I changed more than the subject. Any one of those would have been enough to stay quiet. Together they mean I should not choose a house style from this test.
I can still describe the lines without pretending the description is data. The plain line named the note the way I would mention it to a colleague. The tight line was shorter and put the concrete noun first. Both were specific. Neither was a puzzle. I am not testing clickbait against honesty. I am testing two honest shapes. That limit matters. A test of trickery would be a different, worse experiment.
| Pair | Plain line | Tight line | What else moved | What I can say |
|---|---|---|---|---|
| First | Favored | Lost the comparison | Less than the second pair, still not clean | Not enough to adopt a style |
| Second | Lost the comparison | Favored | The topic was a different kind of note | The topic may be the result |
| Both | No stable winner | No stable winner | Sample stayed small | Inconclusive, on purpose |
No result yet
"Inconclusive" is a result if you let it be one. It means do not update the strategy. It does not mean do not update the method. The method failed to isolate the line. The next test has to be more boring. Same note. Same list logic. One difference. A size large enough that a handful of people cannot crown a winner. Until then, the honest sentence is the one in the frontmatter: the plain line won once and lost once.
There is a temptation to average them into "it depends" and sound wise. "It depends" without a depends-on is a shrug. If I learn what it depends on — topic type, how well the reader knows me, whether the noun is rare — that will be a later hypothesis. I do not have it yet. Inventing it now would be the test editing itself after the fact.
What I would repeat
What I would repeat is the discipline, not the conclusion. Write both lines. Send them only when the note underneath is the same. Do not declare a style the winner because it matches a preference. Preference is allowed to choose when evidence is absent, as long as you do not dress the preference as evidence. Right now I am keeping subject lines specific and skipping the cleverness. That is a working rule, not a finding. Cleverness is a second experiment, not a substitute for this one.
The working rule is conservative. Specific beats puzzling, because a puzzle spends attention the note then has to repay. I still write the spoken version first, out loud, and then a tighter version beside it. I pick one. I write down that I do not know which kind wins. The picking is editorial. The ignorance is scientific. Mixing those two sentences is how people become sure too early.
How the next pair has to look
One note, duplicated. Two subject lines. The body identical, including the first sentence, so an open is about the subject and not about a preview that leaked the difference. A list split that I can describe without hand-waving. A number of sends I am willing to call small in the write-up if it is small. No second change hiding in the from-name, the send time, or the preview text. Preview text is a second subject line. If it changes, you have two variables again.
I would also log my preference before the numbers, in a sentence, so I can see whether I were rooting. Rooting is human. Hidden rooting becomes a biased read of a tie. A tie should stay a tie.
Until that pair exists, this page should not be cited as "plain language wins" or as "tight lines win." Cite it as a method note about a dirty test. Dirty tests are common. They become harmful when the dirt is washed out of the retrospective. I would rather keep the dirt visible. The experiment template has a result field for a reason, and the result is allowed to be "I do not know."
That sentence is the one I need the practice of writing. Marketing work is full of near-tests that get promoted to principles because a principle feels better in a meeting than a shrug. The meeting can survive a shrug. The next send can be cleaner because the shrug was allowed. That is the only win I am taking from this one.