The most expensive setting in your sequence is the day gap

Think about the last time you built a cold email sequence. You agonized over the first line. You rewrote the subject four times. You argued with someone about whether the CTA should be a question or a soft ask. And somewhere in the middle of that, you clicked a dropdown that said wait 2 days, because it was already there, and you never thought about it again.

That dropdown might be doing more work than the subject line.

A study of 100,000 paired cold emails, 50,000 AI-generated and 50,000 human-written, run from October 2025 through April 2026, measured inbox placement against follow-up spacing. One-day intervals landed 71% of mail in the inbox. Three-day intervals landed 93%. That is a 22 point swing from a setting most teams treat as scheduling furniture.

The setting nobody revisits

Here is what I find genuinely strange about outbound as a discipline. We have made copy into a craft. There are people who do nothing but write first lines. There are frameworks with names. Meanwhile the operational settings underneath, the ones that determine whether any of that copy gets seen, get configured once during onboarding and inherited forever by everyone who touches the account afterward.

Interval is the purest example. It gets set by whoever built the first sequence, usually to whatever the platform defaulted to, usually tight, because tight feels aggressive and aggressive feels like effort. Then it propagates. Every sequence cloned from that one carries the same gap. Two years later a team is running forty sequences at a spacing nobody ever chose on purpose.

Notice where the curve flattens. Going from one to two days buys you ten points. Two to three buys twelve more. Three to four and beyond buys two. That shape matters more than any single number, because it tells you there is a real cost to being tight and almost no additional benefit to being patient past a point. Three days is where the return stops.

What the cold email follow-up interval data shows

The methodology is worth a paragraph, because it is better than most of what circulates in this space. Each AI email was paired with a human-written one matched on persona, ICP firmographic, sequence stage, sender domain age, and sender domain authority. Deliverability came from Gmail Postmaster Tools and Microsoft SNDS rather than from open tracking. Industry mix ran SaaS 28%, agencies 18%, financial services 12%, healthcare 9%, manufacturing 8%, retail 7%, other 18%.

The headline finding people will quote is the reply rate gap: 4.1% for AI-written versus 5.2% for human-written, with positive replies at 1.4% against 2.1% and meetings booked at 0.7% against 1.1%. That gap has narrowed, from 2.0 points in 2024 to 1.1 now, entirely because AI improved while human performance stayed flat. Fine. We have written about that comparison before and the picture has not fundamentally changed.

The cadence finding is the one nobody is quoting, and it is the more useful one, because unlike the AI versus human question it does not require you to change how your team works. It requires you to change a number in a dropdown.

Why spacing behaves like a deliverability control

The mechanism is not mysterious once you say it out loud. Filtering systems evaluate sending patterns, not just content. A burst of messages to the same recipient in rapid succession looks like a pattern. Volume compressed into a short window looks like a pattern. Spacing dilutes both.

There is also a second-order effect that I suspect is doing more of the work than people assume. Tighter intervals mean more total sends in the same calendar period from the same infrastructure. A five-step sequence at one-day gaps compresses five sends into a week. The same sequence at three-day gaps spreads them across two and a half. Same volume of intent, half the daily sending pressure per domain. This is the same relationship we traced when looking at how sending volume interacts with deliverability, arriving from a different direction.

I want to be careful about one thing, because the conclusion is tempting and slightly too clean. This is observational data from paired campaigns, not a controlled experiment where interval was the only variable manipulated. Teams that run three-day gaps may differ systematically from teams that run one-day gaps in ways that also affect placement: list quality, domain maturity, general operational care. The correlation is strong and the mechanism is plausible. It is not proof that changing your dropdown moves your number by 22 points.

Which is the honest caveat, and I would rather state it than let the chart do the arguing. But consider what the caveat costs you if you act on the finding anyway and it turns out to be partly confounded. You extended your sequence calendar by a few days. That is the entire downside. Compare that to the downside of ignoring it if it is real: every message you send from now on gets judged on a 71% sample, forever, while your team runs copy tests against noise. Asymmetric bets with a cheap wrong side are the easiest decisions in marketing, and this is one of them.

The other thing worth saying about the flattening curve is what it implies about the aggressive-cadence school of outbound. There is a genuine argument for tight follow-up, and it is not stupid: attention decays, a prospect who half-registered your first email is more likely to respond to a second one soon after, and momentum is real. That argument assumes the second email arrives. At one-day gaps, on this data, nearly three in ten do not arrive anywhere the prospect will see. The aggressive approach defeats itself before the psychology ever gets a chance to work.

The vertical spread nobody plans around

The other thing buried in this study is how differently the same tactics land by industry. Reply rates for AI-written mail ranged from 6.1% in SaaS down to 1.9% in financial services. That is more than a threefold spread across verticals in one dataset.

SaaS being the one vertical where AI-written mail outperformed human-written mail is a genuinely interesting result, and I would not over-read it. SaaS buyers get more cold email than anyone, are more habituated to its conventions, and are probably the population whose expectations most closely match what a model trained on cold email produces. That is not a compliment to the model. It is a comment on how formulaic the category's inbound has become.

If you sell into fintech or financial services, that 1.9% deserves to be in your planning assumptions rather than discovered in month four. Every benchmark you have been handed is blended across verticals, which means for a financial services team the industry average is not a target, it is a mirage. This is the same distribution problem we unpacked in how reply rate benchmarks hide their own spread, showing up along a different axis.

Changing the interval without breaking the sequence

So what do you actually do on Monday. Not a rebuild. A test, on the setting that costs nothing to change.

The reason I like this as a first test is that it is nearly free. It costs no creative time, no new tooling, no headcount, and no argument about whether AI should be writing your emails. It extends your sequence calendar by a few days, which almost never matters in a B2B cycle measured in weeks or months, and it is fully reversible if the number does not move.

“You cannot optimize copy that never arrived. Placement is not a deliverability problem sitting next to your messaging work. It is the denominator underneath it.”

Here is the reframe I would offer, and it is not really about intervals. Outbound teams spend their attention where the craft is, and the craft is in the writing. Everything upstream of the writing gets treated as plumbing, configured once by whoever set up the account, then inherited by people who assume someone thought about it. Usually nobody did. The most valuable hour you will spend this quarter is probably not on another subject line variant. It is on reading your own configuration back and asking which of those numbers anybody actually chose.

Do this next: open your sequences, write down the configured interval for each, and find the tightest one running real volume. Widen it to three days, change nothing else, and watch placement rather than replies for four weeks. If the study's shape holds even partially, you will recover more inbox than a quarter of copy testing would have. That is the sort of unglamorous fix that makes an outbound program work, and it is sitting in a dropdown nobody has looked at since setup.

See where you are cited today

A free snapshot audit of your rankings and AI citations before we ever talk.