“Is a 0.4% unsubscribe rate bad?” is one of the questions I am like 'What do you mean?' too, and it is almost never the question somebody is really asking. What they want to know is whether they are allowed to stop worrying, and whether the number they are about to put in front of their boss is going to get them into trouble.

Which is fair enough, but it means the answer people want is a single figure, and a single figure does not exist, is not coming, and would not help you very much if it did.

Go looking for a benchmark this afternoon and you will find the average unsubscribe rate reported as roughly 0.15% by one enormous study, around 0.22% by another widely republished source, and 0.89% by a third, with anything under 0.5% described as healthy. Those are not small differences; that is a spread of nearly six times, and every one of those numbers is published confidently by people who are not lying to you.

They are averages taken across businesses that share nothing with each other except a send button. A daily deals list and a quarterly B2B newsletter sit in the same dataset, as does a brand acquiring through discount pop-ups and one acquiring through a paid community. Blend all of that together and you get a number that describes nobody at all.

So the post does three things. It tells you which numbers are worth respecting and which ones I would ignore, it explains why my thresholds are consistently lower than the ones you will read everywhere else, and it answers the question sitting underneath all of it, which is when you should stop emailing somebody.

The four kinds of churn

Churn is four different things wearing one word

Most reporting collapses all of it into a single unsubscribe percentage, which is a bit like measuring your health by your shoe size. The four behave differently; they are caused by different things, and they need completely different responses from you.

• Voluntary churn; they unsubscribed. Visible, measurable and, for the most part, healthy, because somebody removing themselves has told you the truth and used the front door to do it. A programme with almost no unsubscribes is usually not a beloved programme, it is a quiet one that nobody is reading.

• Involuntary churn; the address stopped working. Hard bounces, abandoned mailboxes, and people who left their job and took their inbox with them. B2B lists decay considerably faster than B2C for that reason alone, and none of it says anything about your content.

• Silent churn; they are still on the list and they stopped caring. The expensive one, because there is no signal, no event and no line in any report. They quietly stop, they keep sitting in your denominator making everything look worse, and they slowly damage your sender reputation while they do it.

• Hostile churn; they marked you as spam. The dangerous one, and the only kind with an enforcement consequence attached to it. Less churn than damage, and it is the one number in this whole post with a hard ceiling set by somebody other than you.

If you take one thing from that list, take the relationship between the first and the last. An unsubscribe is a good outcome and a spam complaint is a bad one, so every decision you make about frequency, about how prominent your unsubscribe link is, and about who you keep emailing, is really a decision about which of those two exits you would rather somebody used. Make the unsubscribe easy, because you are choosing the better of two departures.

Benchmarks

What the internet says, and what I would work to

I should say at the top that I do not believe in benchmarks, and anybody who follows my work will have heard me say it before. Not because measurement does not matter, but because a benchmark borrowed from somebody else's business is a number with no context attached, and context beats best practice every time.

That said, people need somewhere to start, so if you want numbers to work to, these are mine. You will notice immediately that they are lower, sometimes a lot lower, than anything you will read elsewhere.

The reason my complaint and bounce thresholds sit so far below the published ones is that the published ones are ceilings, not targets. 0.3% is the point at which mailbox providers start taking action against you, so treating it as a goal is a bit like treating the speed limit as a minimum. By the time you are approaching it you have already been sending unwanted email to a lot of people for a long time.

Why I will not give you an unsubscribe benchmark

Some of my own emails have an unsubscribe rate that would horrify a benchmark chart, and I am pleased about it, because I asked for it. If I send something designed to clear out people who signed up for one thing and are getting another, or an email that says plainly that the next few months are going to be about a topic that will not suit everyone, a high unsubscribe rate is the email working exactly as intended.

So a single unsubscribe benchmark across a whole programme tells you almost nothing, because it averages an email that was meant to filter people with one that was meant to serve them. What you want instead is a benchmark for each type of email you send, built from your own history.

A welcome email, a monthly newsletter, a promotional send, a re-permission campaign and a deliberate clear-out should each have their own expected range, and the only useful comparison is that email against the last twelve versions of itself. Everything else is noise dressed up as insight.

What to compare yourself against instead

• Yourself, over time. Your own trend line is the only comparison where every other variable is held still. A rising unsubscribe rate on a stable programme means something; a 0.4% unsubscribe rate sitting on its own means almost nothing at all.

• Your acquisition sources, against each other. The most useful cut available to you and one that hardly anybody runs. Churn by source will tell you, without any ambiguity, which of your growth channels is producing subscribers and which is producing numbers.

• Cohorts rather than the whole list. Everybody who joined in March, tracked as a group. A blended list figure is an average of a healthy new cohort and a dying old one, and it successfully hides both of them.

• The same email against itself. A promotional blast and a lifecycle email should never share a row in a report, because putting them together is how programmes get judged unfairly in both directions at once.

Engagement

Engagement is a risk question, not a marketing one

Every conversation about churn eventually walks into this, because you cannot say who has churned until you have said what counts as being alive. Most businesses have never written that down, so they inherit ninety days from a blog post, and some platforms will even start throttling your sends to “disengaged” contacts using a definition they made up on your behalf.

Ninety days is not a finding, it is a round number that got repeated until it started to sound like one. Applied to the wrong audience it will have you cutting people who were about to buy, and I would go further: as a blanket rule it is complete b*llocks.

The reframe that fixes it is to stop treating engagement as a marketing metric and start treating it as a risk measure. The question you are really asking, every time you decide whether somebody stays on the send, is a two part one. If we send this email to this person, does it put our deliverability at risk, and does it create friction or a bad experience for them?

Somebody who never opens and never clicks is not automatically a risk on either count. Plenty of humans simply do not interact with email and never will, even when they signed up on purpose and want the product, and that is human variety rather than a failure of your copy. What you are hunting for is not quietness, it is total disengagement, which is a different thing entirely.

The two kinds of engagement, and why counting one gives you the wrong picture

Direct engagement is what happens inside the email and what your ESP can see, which comes down to opens, clicks and unsubscribes. Most of what people do with an email is not trackable at all, so you are working with a narrow slice of reality even at your best.

Opens are the weakest signal in that slice. Apple's privacy protections and image pre-fetching mean an open can fire without a human ever looking, so building an engagement decision on opens alone is building on sand. The click is the reliable direct signal, because a click is somebody deciding to do something.

Influenced engagement is everything that did not happen inside the email but was prompted by it, or that shows intent towards your business somewhere else entirely. Site visits, baskets started, webinars attended, forms filled, purchases made. Real engagement, sitting somewhere your ESP cannot see it.

Take my own audience as an example. Plenty of people read the blog regularly and appear stone dead in my newsletter reporting, and there are half a dozen explanations before I get anywhere near “they have gone off me”. The open never fired because of how their client handles images. They read it in Apple Mail. Gmail clipped the email and they read the top half without ever loading the rest. Any assumption I make from the email data alone is an assumption made with one eye shut.

You have to understand your deliverability before you can do this

Working out what engagement means for your audience is not a standalone exercise you can do in an afternoon with a spreadsheet, because the whole point of it is risk, and risk is a deliverability concept. If you do not know where your mail is landing, what your complaint rate looks like by segment, or how your reputation is currently sitting, you cannot judge whether sending to a quiet group is dangerous or completely fine.

Mailbox providers are deciding placement on one underlying question, which is whether the people you send to want what you send. They read the answer through engagement signals, so a large group of people showing nothing at all generates the worst possible answer, and it drags down placement for everybody on your list, including the people who love you.

So your engagement make-up is not list tidiness, it is deliverability infrastructure, and treating it as a marketing housekeeping job is why so many programmes get quietly worse for years without anybody being able to say why.

Sorting people by risk rather than by activity

Once you have a window that suits your sending frequency, and a list of the meaningful actions that count as influenced engagement for your business, you can sort your list into four groups. I teach the full method, including how to size the window properly and how to score the database, inside the deliverability certification programme, but the shape of it is worth having now:

• Highly engaged. Several direct actions inside your window, clicks especially. Usually a small slice, and do not panic when it is smaller than you hoped.

• Engaged. At least one direct action, or one meaningful action outside the email, inside your window. Active, just not loudly.

• Disengaged. Nothing direct inside the window, but a real action before it. John lives here, and binning this group without looking is the most expensive mistake in the whole exercise.

• Dormant. Nothing at all, inside the email or outside it, across a long period. The group carrying your risk, and the one your reputation is paying for.

Notice what changes when you sort this way. You stop asking who is ignoring you, which is a slightly wounded question, and start asking who is costing you, which is a decision you can act on without any feelings attached to it.

The sunset

When do you stop emailing somebody?

You stop when the risk of sending outweighs the value of sending, and not before. Which sounds obvious until you notice that almost nobody has written that rule down, so it gets decided in a panic by whoever is most worried about the numbers that week.

The signals worth watching are not the same as the metrics on your dashboard. You are looking for whether risk is growing: a cohort where complaints are creeping up, a segment where nothing at all has happened inside or outside the email for a long stretch, a source that behaved badly from the day it started, addresses that have begun bouncing softly and repeatedly, and people who stopped both direct and influenced engagement at the same time. One quiet quarter is not a signal; quiet everywhere, for a long time, with complaints rising around it, absolutely is.

The other thing worth saying is that stopping is not one action, it is a ladder, and most programmes leap from the top of it straight to the bottom.

1. Change what you send. Different content, different ask, different angle, before you assume the relationship is over.

2. Reduce frequency. Fewer sends to that group rather than more, because the instinct to email harder at quiet people is almost always the wrong one.

3. Ask them directly. A preference prompt, one plain question, an easy route to less rather than none at all.

4. Stop the marketing. They stay on the list and keep getting transactional and service mail, they stop getting campaigns.

5. Suppress. Excluded from sends, retained as a record, still recognised if they come back.

6. Delete. Genuinely gone, and only for data retention reasons rather than performance ones.

The gap between suppress and delete is the one people get wrong and it is an expensive mistake. Suppressing protects your deliverability while keeping the history, so when somebody returns through a purchase or a form you still know who they are and what they did. Deleting throws away the relationship along with the record, and they come back to you as a stranger.

Who stays on the list, whatever your rule says

• Customers inside a long purchase or renewal cycle. Silence is the expected behaviour and you designed the cycle.

• Anyone with a live service, support or sales conversation running. Engaged, just not with your marketing.

• Recent subscribers who have not had a fair run. Somebody who joined six weeks ago and received three generic broadcasts is not disengaged, they have had a bad orientation flow.

• Segments where quiet is normal and documented. Seasonal buyers, contacts who are out of market for now, anybody whose cycle you already understand.

Every one of those is an exclusion rule rather than a judgement call, which means it gets written into the system once and applied forever without anybody having to remember it on a Friday afternoon.

Growth vs Churn

The number that matters more than either of them

Growth and churn tell you very little on their own, and reporting them on separate slides is how a business talks itself into believing everything is fine.

Net list growth, meaning what you gained minus everything you lost including the silent losses, gets closer to the truth. Even that has a hole in it though, because it treats every new subscriber as equivalent and they are very obviously not.

Add ten thousand people through a discount pop-up, watch eight thousand of them go functionally quiet within two months, and you did not grow your list by ten thousand. You grew it by two thousand, and you damaged your sender reputation on the way, because for those two months you were mailing eight thousand people who did not want you.

Which is the consequential opt-in problem in its purest form. Somebody who handed over an address to get money off did not sign up for a relationship, they signed up for a discount, and counting them the same way you count somebody who deliberately subscribed is how a list becomes a liability while the growth chart keeps climbing.

The cuts that tell you the truth

• Net growth by cohort. Track each joining month as its own group and you will see how long a cohort stays useful, which is usually a shorter and more sobering picture than the blended figure.

• Net growth by acquisition source. One source is quietly funding your churn problem and you will be able to name it within an hour of running the report.

• Engaged list size rather than total list size. The only figure that tracks revenue, and reporting it changes the conversation about volume almost immediately.

• Revenue or pipeline per thousand sent. Because a smaller, healthier list often earns more than the bloated one it replaced, and that is the metric that proves it.

What I would keep an eye on

• Spam complaint rate on every send, watching the absolute count too if your list is small.

• Hard bounce rate, treated as a data quality alarm rather than a deliverability statistic.

• Unsubscribe rate split by email type, compared only against previous versions of that same email.

• The size of your dormant group, because it is the part carrying your risk.

• Net growth by cohort and by source, which is where cause lives.

• Revenue or pipeline per thousand sent, which stops anyone arguing that bigger is better.

Where I would leave it

You did not come looking for a number, even though you thought you did. You came because something felt off and you wanted to know whether to worry about it.

So the reframe is worth sitting with. Churn is not something that happens to your list, it is the receipt for decisions you made much earlier, about who you let in, what you promised them, how often you turned up and whether the thing they received resembled the thing they were offered. The number at the end is just the arithmetic on all of that.

Which is oddly good news, because it means churn is not something you manage, it is something you cause, and anything you cause you can change.