One hundred sends to find the Gmail template that works

Gmail will not tell you, because its templates keep no record of what happens after you press send. Google's Gmail Help page on templates covers four things: switching them on under Settings, Advanced; saving a draft as a template or overwriting an old one; inserting one from More options, Templates; and sending one as an automatic reply through a filter. That is the whole feature. To learn which template earns opens, you have to track the emails, and then you have to send enough of them for the numbers to mean something. The second part is where most people stop too early.

What Gmail templates are good at

They are free, they ship with every Gmail account, and they stay out of your way. A few details from Google's pages are worth knowing before you build a library on them.

• Templates are off by default. You enable them once in the Advanced tab of Gmail's settings.
• They work only in Gmail on a computer. Google's Workspace tips page says plainly that templates are not available in the Gmail mobile apps.
• Editing means writing a new draft and overwriting the saved template with it.
• A deleted template cannot be recovered, so keep a copy of your best ones somewhere else.

Extensions add speed rather than measurement. Streak's help article on snippets describes saving a draft as a snippet, giving it a shortcut code such as ^thanks that you type in the compose window, and inserting the subject along with the body. That is quicker than three menu clicks. The article says nothing about measuring which snippet works.

Measure the template, one send at a time

The useful number is simple: of the emails sent from this template, what share were opened at least once by a person. Count each email once, however many times it was reopened, or one forwarded thread will make a weak template look strong. Leave out opens your tracker flags as automated, because a security scanner fetching the image is not a reader. Send each template to the same kind of person, in the same week, so the template is the main thing that differs.

That is why replies are the number that settles it. Opens are early and plentiful, so use them to drop subjects that clearly fail, then judge the survivors by how many replies they bring.

Why twenty sends proves nothing

Say template A goes to 20 prospects and 12 open it, and template B goes to 20 and 9 open it. That reads as 60 percent against 45, and it is tempting to retire B. The standard check for two proportions is the z test described in the NIST Engineering Statistics Handbook. It pools both groups, divides the gap by its standard error, and compares the result with 1.96, the usual cut off for 95 percent confidence. Here the result is 0.95, about half of what it needs to be. For samples this small the handbook points to Fisher's exact test instead, and that gives a p value of roughly 0.53. In plain terms, two identical templates would show a gap this large or larger about half the time, by luck alone.

Twenty sends each tells you which template you happened to send on a good day.

Keep the same rates and grow the sample. At 86 sends each, the z score passes 1.96 for the first time. At 100 each, 60 opens against 45, it reaches 2.12, and Fisher's test gives about 0.047. A smaller gap needs far more: 50 percent against 40 needs around 190 sends per template before it clears the same bar. We ran these numbers ourselves with the handbook's formulas; you can repeat them in a spreadsheet in a few minutes.

A routine that fits real outreach

• Test one change at a time, usually the subject, so you know what the result is about.
• Alternate the two templates through the same list rather than sending A on Monday and B on Thursday.
• Decide the sample before you start, around a hundred sends each, and do not crown a winner early.
• If you send fewer than that in a quarter, pool your judgement across months, or accept that you are choosing by taste, which is fine as long as you call it that.

What Fast Mail Tracker shows today

You save templates in our dashboard with a name, subject and body. They appear in a Template button inside the Gmail compose window, and in the dashboard's own composer. Every template you save shows how many times it was used and its open rate, counted the way this article describes: the share of sends with at least one open, with opens our filters flag as automated left out. There are two limits you should know. The count covers emails sent from the dashboard composer; inserting a template from the Gmail compose window fills the draft but is not yet counted against that template. And we report opens per template, not replies, so judge the final round by the replies in your inbox.

Fast Mail Tracker is in early access while Google's security review completes, so places are limited; if you want each template's opens counted as you send, the install button below is where to start.