Every time you press send, roughly fifteen percent of that campaign is landing in spam. Not because you did something wrong last Tuesday, not because of your subject line, and not because you are being punished. Across providers, across industries, across list sizes, about one in seven of your emails is not reaching an inbox, and that is the ordinary state of the channel in 2026.
Sit with the arithmetic for a second. Send to a hundred thousand people and fifteen thousand of them never saw it. Your reporting will show ninety nine percent delivered, your dashboard will look fine, and nobody in your business will ever know.
So this is my read on where deliverability sits right now, drawn from the published benchmarks, from my own client work, and from the several billion emails a quarter I have visibility of. Then, at the end, what I think happens next year, which you are welcome to disagree with and quote back at me in December.
Pulling together the Validity benchmark data, several other published reports and what I see across the accounts I work on, the picture is consistent enough that I am comfortable stating it plainly. Around fifteen percent of a typical send does not reach the inbox. Validity's own 2026 figures put global inbox placement in the low eighties, which lands in the same territory from the other direction.
Two things worth saying about that number before anybody panics or relaxes.
It is an average, so it is describing nobody exactly. Some senders sit at two percent and some sit at forty. The average is useful for calibrating your expectations and useless as a target.
It is invisible in standard reporting. Delivered means the receiving server accepted the message. It says nothing about whether a human ever had the chance to see it, which is why so many programmes spend years optimising things that were never the problem.
If you have never measured placement separately from delivery, that gap is the single most expensive unknown in your programme, and it is the first thing I go looking for on any audit.
Professionally, this has been my busiest year by a distance, and speaking to other consultants, tool vendors and platform teams, they are all saying the same. More problems, more emergencies, more businesses waking up to find their placement has collapsed with no obvious cause.
The instinct is to conclude that the filters have got harsher, or more sensitive, or somehow unfair. And the filters have got stricter, so that instinct is not wrong, but it misses what is really going on.
What has changed is that the tolerance disappeared. For years you could run a programme with a sloppy list, half-configured authentication and a habit of mailing everybody, and the providers absorbed it. Enforcement has now become mechanical, so the same behaviour that was quietly tolerated in 2022 produces a rejection in 2026.
Which means most of the emergencies I get called into are not new problems. They are old problems that finally arrived. Your practices caught up with you, and that is a much less comforting explanation than blaming Gmail.
Gmail moved from soft deferrals to hard rejections. Non-compliant bulk mail now gets permanently refused rather than quietly dropped into a junk folder where you might have noticed it eventually.
Microsoft brought its own requirements in and will not accept a monitoring-only DMARC policy. If your record is still sitting at p=none, Microsoft is the provider that will punish you for it first.
The three big providers converged. There is no longer a weak link you can lean on while you sort the others out. Whatever you build has to satisfy all of them at once.
Cold email is dying, and a great many B2B businesses have built their entire pipeline on top of it. Microsoft hosts most corporate mail and Microsoft is now the hardest inbox to reach, which means the tactic that used to work is failing precisely where it was aimed.
The response from a lot of the market has been to buy tools that promise to solve it. Inbox warming services, primary placement guarantees, rotation setups, deliverability scores that sit in a dashboard and reassure you. All of it is a sticking plaster over a wound that needs stitches, and none of it changes the underlying fact.
You cannot shortcut this. There is no product you can buy that substitutes for sending mail people want to a list that agreed to receive it, and every year the providers get better at telling the difference between the real thing and an imitation of it.
Ask most marketers what deliverability is and you will get a version of: how many of my emails land in the inbox rather than spam. Which is the outcome, not the thing itself, and confusing the two is why so many people cannot fix it.
Read that definition again and notice the two words doing the work: ongoing and behaviour. Deliverability is not a setting you configure, not a score you check, and not a problem you fix once. It is a condition, and conditions respond to how you behave over months rather than to what you did this morning.
I have surveyed around three thousand marketers globally, and more than eighty five percent still say they do not know enough about deliverability and do not feel equipped to deal with it. Not that they have never heard of it. That they understand it exists, could give you the one-line definition, and would not know where to start if their placement collapsed on a Monday.
The Email Marketing Health Check data says the same thing from a different angle. Across the seventeen areas of email marketing it assesses, deliverability comes out second lowest, consistently, month after month.
So we have an entire profession that knows the channel matters, knows this part of it is technical, and has quietly decided it belongs to somebody else. In a large organisation there might be a dedicated deliverability team, and almost nobody reading this works somewhere like that.
Not to become a specialist, and not to configure DNS records at two in the morning. To understand the ecosystem well enough to make good decisions, to spot a problem before it becomes an emergency, and to push back when somebody asks for something that will cost you the channel.
The alternative is what happens now, where the person who owns email cannot explain why results are falling, cannot evaluate the advice they are being given, and cannot tell whether the tool they just bought does anything at all.
Search volume for deliverability training is up, course enrolments are up, and businesses are starting to put budget behind it. Which is the most encouraging thing in this whole post, and about five years later than it should have happened.
A term I have been reaching for all year, so let me define it: deliverability washing is claiming deliverability capability you do not have, whether that is a tool with a score that measures nothing, a platform giving advice outside its competence, or a specialist whose expertise turns out to be a landing page.
The problem it creates is not that people are being lied to exactly, it is that everything sounds equally authoritative, so a marketer trying to learn has no way of sorting the useful from the confident.
The single most common trap, and it catches good people constantly. Because a platform sells email software, we assume it must be expert in email, and by extension expert in deliverability. Those are not the same discipline and the second one is frequently not where the platform's interest lies.
Klaviyo, for instance, gives you a deliverability score, and it does not tell you what it appears to tell you. Plenty of platforms do the same thing, presenting metrics labelled as deliverability that are not deliverability metrics in any meaningful sense. A number that goes up and down while your placement does something entirely different is worse than no number at all, because it manufactures confidence.
None of which makes those platforms bad products. It makes them products, and understanding the ecosystem your email travels through is your responsibility, not something you can delegate to a vendor whose incentive is for you to keep sending.
Which is the part that gets lost under all the noise. Deliverability is not complicated, it is just unglamorous and cumulative. Prove who you are, send to people who asked, stop sending to people who stopped caring, keep your complaints low, be consistent, and give more than you ask for.
Everything else is detail. The reason it feels overwhelming is not the subject, it is the volume of people talking about it.
Predictions are opinions with a date attached, so treat these as mine rather than as forecasts, and hold me to them.
Find out where you land. Placement measured separately from delivery, by provider, as a trend rather than a single reading.
Fix authentication properly and get DMARC to enforcement. A record at p=none is monitoring, not protection, and Microsoft will find you first.
Separate your streams. Transactional away from promotional, and different subdomains for different jobs, so one bad month does not take down your order confirmations.
Deal with your dormant group. The people showing nothing at all, inside or outside the email, are the ones costing you placement for everybody else.
Put a human back in the copy. If your programme has drifted into automated writing, the engagement decay is coming whether or not you have noticed it yet.
Learn enough to evaluate advice. Not to become a specialist, but so that when a vendor or an agency tells you something, you can tell whether it is true.
Build the smaller, better list. Particularly in B2B, and particularly if your pipeline currently depends on volume outreach, because the runway on that is shorter than people think.
The state of deliverability in 2026 is that the rules stopped being suggestions, the tolerance ran out, and a profession that was never taught this is now being held responsible for it.
None of which is a reason to be gloomy, because the businesses doing well are not doing anything exotic. They authenticate properly, they send to people who want it, they suppress the people who do not, and they give more than they ask. The advantage available to anybody willing to do the boring version of this has never been larger, precisely because so few people are doing it.
And if the baseline does move toward twenty percent next year, the gap between the senders who prepared and the senders who did not stops being a rounding error and starts being the difference between a channel that works and one that quietly does not.