You are probably under-emailing your list

The newsletter has been written for nine days. It is good — better than the last one — and it is still sitting there, because every time you go to send it you think we just emailed them. So you check, and the last send was the nineteenth of last month, which is to say five weeks ago, which is to say most of the people on that list have forgotten they are on it. The draft is not waiting on a decision about content. It is waiting on a fear nobody has ever put a number to.

We see this far more often than the opposite. The founder worried about sending too much has usually sent four emails this year. The instinct behind it is decent — you do not want to be the company that shows up in the inbox three times a week with nothing to say — but it gets applied to a situation that is nothing like that one. A list is not a fixed asset you protect by leaving it alone. It decays on its own: people change jobs, filters get stricter, and the sender they cannot place gets swiped away without opening. Silence does not preserve the relationship. It just makes the loss quieter.

Put rough numbers on it, since the fear never has any. Say a list of two thousand, and each send produces about eight signups worth roughly forty dollars each — call it three hundred dollars of value per send. Say each send also costs you about six unsubscribes, or 0.3%. Send monthly and that is twelve sends a year: around thirty-six hundred dollars, seventy-two people gone. Send every two weeks and it is twenty-six sends: around seventy-eight hundred dollars, a hundred and fifty-six gone. Doubling the cadence cost you eighty-four subscribers and returned roughly four thousand dollars more. Those figures are invented and yours will differ, but run your own through the same shape and you will usually find the same asymmetry: the unsubscribes are the number you feel, and the sends you skipped are the number you never see.

the cost of the email you did not send is invisible, which is why you keep not sending it.

Frequency is also not usually what burns a list. Sameness is. People will happily take a weekly email that tells them something they did not know, and will drift out of a monthly one that is the same three product updates rearranged. So watch the signals that actually mean stop, and they are specific: open rate falling send over send across a run of five or six, not one bad Tuesday; clicks concentrating in a shrinking group of the same people; replies going from questions to silence. And one hard stop that overrides all of them — spam complaints. If those rise, cut immediately and ask what changed, because that number affects whether your email arrives at all, not just how it performs.

The cadence you should pick is the fastest one you can hold for a year. This matters more than the exact interval. Fortnightly forever beats weekly for five weeks and then nothing until March, because the second pattern trains people not to recognise you, and every send restarts from cold. Pick the interval by looking at how much you genuinely have to say, halve your optimism, then say it out loud in the first email — people who know a monthly email is coming are far less startled when a monthly email arrives. That is one of the quiet jobs of the first email anyone gets from you: it sets the expectation the rest of the year gets measured against.

Then the rule that makes a fixed cadence survivable: a scheduled send has to be allowed not to go. A calendar slot is a commitment to consider sending, not to send regardless. When the slot arrives and the honest answer is that this week's email would be filler, the right outcome is a skip with a reason written down — and the reason is the important half, because "nothing to say" three slots running is not a scheduling problem, it is a signal about the whole plan. We built ours to work exactly this way, and it is worth reading on its own terms: [a send that can say no](/blog/when-to-skip-a-scheduled-send) is what stops a cadence from becoming a treadmill.

This is also the honest boundary of what AI drafting does for you. It takes the blank page away — hand it your notes and a rough angle and you get something to react to in a minute, and reacting is much easier than starting. It does not give you something to say. If you have nothing this fortnight, a generated draft is a fluent way to send nothing, and readers can tell. The fix is unglamorous: keep a running note through the month of what you learned, what a customer asked twice, what you changed. Then the draft has raw material, and you are approving something real rather than approving prose. Check every number in it before it goes, always.

One measurement note, because the wrong denominator keeps this fear alive. Unsubscribe rate per send rises when you send more often, mechanically and forever, so if that is your headline number you will always conclude you should send less. Look at revenue or signups per subscriber per quarter instead, and at whether the list is growing net of departures. Those answer the actual question. And a departure is not a failure — [the person who leaves](/blog/the-unsubscribe-you-should-be-glad-about) has told you something useful for free and stopped skewing every rate you read.

So the practical version, in one move: go and look at the real date of your last send, not your memory of it. If you cannot recall roughly when it was, your subscribers certainly cannot, and the email you have been protecting them from is the one they were most likely to open.