Growing an Email List Is a Leaky Bucket

Four terms, and only three of them get modelled

Net list growth in a period is additions minus unsubscribes minus bounces minus silent churn. The first three appear in every email platform's dashboard because each one is an event the platform can see: someone clicked a link, a server returned a permanent failure, a form was submitted. The fourth is not an event at all. It is the absence of events — an address that still exists, still accepts mail, and has not opened anything in a year.

Because it is invisible, silent churn is the term that gets left out of the model, and it is usually the largest one. A list that reports a 0.2% monthly unsubscribe rate and a 0.1% bounce rate looks almost watertight. Add the addresses that have gone dormant without saying so and the real monthly loss can be several times that. The rest of this article treats all four terms as a single proportional loss rate, because for the arithmetic that follows only the total matters.

Constant additions and proportional loss converge on a ceiling

Write the month down. If N is the list size, a is the number of addresses added each month and L is the fraction lost each month, then next month's size is N plus a minus L times N. The list stops changing exactly when the addition and the loss cancel: a equals L times N. Solve for N and you get the ceiling, N equals a divided by L. Nothing else is needed — no simulation, no cohort model, one division.

The consequences are unforgiving. At 2,000 new subscribers a month, a 0.5% monthly loss rate gives a ceiling of 400,000 addresses; 1% gives 200,000; 2% gives 100,000; 3% gives 66,667; 5% gives 40,000; and 8% gives 25,000. Halving the loss rate doubles the ceiling, and doubling the additions doubles it too — at 5,000 additions a month the same six loss rates cap out at 1,000,000, 500,000, 250,000, 166,667, 100,000 and 62,500. Every list you have ever seen is somewhere on that grid.

Approach is slow, which is why so few teams notice the ceiling before they hit it. Starting from zero, the size after t months is the ceiling times one minus (1 minus L) to the power t, so reaching half the ceiling takes 34.3 months at a 2% loss rate and reaching 90% of it takes 114.0 months. At 5% those become 13.5 and 44.9 months. A list three years into its life is usually still climbing, which makes the flattening look like a marketing failure rather than the arithmetic arriving on schedule.

What the ceiling costs you just to stand still

The ceiling has a running cost, and it is the part that never appears in a growth plan. A list of 100,000 addresses losing 2% a month loses 2,000 addresses a month. If you are paying $2 to acquire a subscriber, holding that list steady costs $4,000 a month, or $48,000 a year, before a single new subscriber is added on top. That is not a growth budget; it is maintenance.

It also tells you the price of a target. Growing from 100,000 to 150,000 at the same 2% loss rate means the ceiling has to move to 150,000, which means additions have to rise from 2,000 to 3,000 a month — permanently, not for one campaign. Half of that new spend is buying growth and half is paying for the extra leakage the larger list creates. Any plan that adds subscribers for a quarter and then stops has bought a temporary bump that decays back to the old ceiling.

A list can grow in size while shrinking in usefulness

Leaving a list and losing interest in it are different events with different rates. An address is removed slowly — unsubscribing takes a decision — but it stops being opened quickly. Run both processes on the same additions and you get two curves. Take 2,000 additions a month, a 1% monthly rate at which addresses actually leave the list, and a 6% monthly rate at which subscribers stop engaging while staying on it.

Each process has its own ceiling by the same division. The list converges on 2,000 divided by 0.01, which is 200,000 addresses. The engaged part converges on 2,000 divided by 0.06, which is 33,333. So at equilibrium the engaged share of the list is exactly the loss rate divided by the decay rate — 1% over 6%, or 16.67% — and no amount of additional signups changes that ratio, because both ceilings scale with additions identically.

The trajectory is what makes this deceptive, because the two curves diverge slowly and in the wrong direction for morale. After 12 months the list holds 22,723 addresses of which 17,469 are engaged, a healthy 76.9%. After 24 months it is 42,864 and 25,783, or 60.2%. After 36 months, 60,717 and 29,740, or 49.0%. After 60 months, 90,569 and 32,519, or 35.9%. After ten years, 140,124 and 33,313 — 23.8%. The headline number roughly doubles between years three and five while the number of people who actually read anything grows by less than three thousand.

Why buying a list is arithmetically self-defeating

The temptation is obvious: if the ceiling is additions divided by loss rate, buy 50,000 additions at once and jump. The reason it fails is that the loss rate is not a constant you inherit — it is partly a function of what you send and to whom, and buying addresses raises it for everything else you own.

Model it. Buy 50,000 addresses; assume 12% are invalid, leaving 44,000 delivery attempts. Assume 0.8% of those recipients mark the message as spam, which is 352 complaints. Google's published sender guidelines tell you what that means: senders are told to keep spam rates reported in Postmaster Tools below 0.3%, and Google recommends staying below 0.1% and never reaching 0.3%. At 44,000 delivered, the 0.3% ceiling is 132 complaints. You are at 352, nearly three times over, on the first send.

Now the feedback. Suppose inbox placement halves on each subsequent send while complaints stay above the threshold — the exact decay is an assumption, the direction is not — so placement goes 95.0%, then 47.5%, then 23.75%, then 11.875%. Reputation attaches to the sending domain, not to the list, so this happens to your real subscribers too. A genuine, opted-in list of 30,000 on the same domain sees inbox placements of 28,500, then 14,250, then 7,125, then 3,563 — a fall of 87.5% in four sends, or 24,938 lost placements every time you press send. Meanwhile the purchased list, at a generous 0.2% click rate, returned 44,000 times 0.95 times 0.002, which is 84 clicks. You traded your deliverability for 84 clicks.

The honest measure is engaged subscribers, not total addresses

Everything above points at one reporting change: publish the number of addresses that have opened or clicked something in a defined recent window, alongside the total, and treat the first as the real list. It is the number that predicts revenue, it is the number that predicts deliverability, and it is the number that stops rising when the acquisition work stops — which is exactly the feedback a growth team needs.

Two practical additions. Define the window and never move it silently, because widening it from 90 days to 180 days grows the engaged count without a single subscriber doing anything. And retire the addresses that fall out of it: suppressing a dormant segment lowers the total, which feels like a loss, but it removes the addresses most likely to generate a complaint and it makes the ceiling arithmetic honest again. A smaller list you can actually reach is worth more than a larger one the mailbox providers have stopped believing in.

Frequently asked questions

Why does my list stop growing even though signups have not slowed?

Because losses are proportional to size while additions are not. A bigger list leaks more addresses per month at the same loss rate, and growth stops when the leak equals the intake — at a size of additions divided by the loss rate. At 2,000 additions a month and a 2% monthly loss that ceiling is 100,000, and it will be reached whatever the signup form does. To move it you must either raise additions permanently or lower the loss rate.

How long does it take to reach the ceiling?

Strictly, forever — the approach is asymptotic. Practically, starting from zero you reach half the ceiling in 34.3 months at a 2% monthly loss rate and 90% of it in 114.0 months. At a 5% loss rate those drop to 13.5 and 44.9 months. The higher your loss rate, the lower and the sooner your ceiling, which is why a leaky list feels like it plateaus early: it does.

Is a bigger list always better?

No, and the arithmetic shows why. If addresses leave at 1% a month but stop engaging at 6% a month, both quantities have ceilings — 200,000 addresses and 33,333 engaged subscribers on 2,000 additions a month — so the engaged share settles at exactly 1 over 6, which is 16.67%. Along the way the total keeps rising while the engaged share falls from 76.9% at one year to 35.9% at five. Growth in the headline number is compatible with no growth at all in people who read anything.

Can I just buy 50,000 addresses and skip the wait?

The arithmetic says no, because the purchase raises the loss rate it is meant to outrun. Google's sender guidelines ask senders to keep spam complaint rates reported in Postmaster Tools below 0.3%, and recommend staying under 0.1%; on 44,000 delivered messages that ceiling is 132 complaints, which a non-consenting audience passes easily. Reputation attaches to your sending domain, so the damage lands on the subscribers who did opt in — in the modelled case, inbox placements for a genuine 30,000-address list falling 87.5% across four sends.

Should I delete subscribers who never open anything?

Suppressing them is usually the right call, and the total-size number is the only thing that suffers. Dormant addresses cannot buy anything, they are the ones most likely to generate a spam complaint against the 0.3% ceiling Google publishes, and removing them makes your engaged count and your ceiling arithmetic describe the same list. Define the engagement window explicitly, keep it fixed, and try a reactivation sequence before suppression so the decision is based on a real last attempt rather than silence alone.