2026: Emails

Here's a dirty little secret about Buttondown, and almost all the other services like it: we're not actually in charge of sending your emails. We use a downstream software vendor for that. These vendors are called MTAs, and you can read more about them here.

The reason behind this is fairly simple, even if it's not obvious. Sending email—the literal software-level act—is quite difficult, annoying, and arcane. By using a vendor to handle this, we've been able to focus on what we think is the most important part of leverage that we can apply (or, to use a slightly smarmier term, our secret sauce): subscriber management, archives, and more.

Over the past year, we've embarked on a pretty successful quest of insourcing things like analytics (you can read more about that here), data, and more. In 2026, we're going to tackle the true dragon: we're going to shift to sending all of our emails—or at least the majority of them—ourselves, without relying on a downstream vendor.

The natural question here is why.

First off, if you're reading this and you have a technical background, you might be a little terrified on our behalf. I think that's correct. I think the usual guidance of "don't roll your own email" is absolutely the right one. However, the value proposition changes once you hit the scale that we have.

From a purely selfish perspective, sending emails is our second-largest operating cost (first if you exclude payment processing, which is almost worth ignoring given that there's no world in which we try to roll our own version of Stripe). A lot of our decisions around how to price Buttondown and what features around emailing to allow or disallow are dictated by the fact that, very literally, sending emails ain't free. This changes the chessboard of our options if we're able to relax that constraint—though obviously it's easier said than done, given that we'll have to send these emails from a server somewhere, and that server will itself cost money, albeit much less than what we're paying currently.

The second reason is reliability. We've pioneered a really novel and really valuable multi-processor approach that allows us to jump between multiple MTAs depending on their own day-to-day reliability. However, even with that work, we've had a number of issues this past year that stem from vendor deliverability that we can't fix ourselves. I don't mean this to say that we pin the blame on them—if anything, it's the opposite. We need to be in a position where we have ultimate agency over being able to send emails reliably and with an extraordinarily high level of reputation.

The third and final bit is about performance. One of the things I mentioned when we wrote about insourcing analytics is that the vertical integration sped up the entire process because we're no longer having to deal with quite as much translation and separation that comes from munging data from a vendor. We believe that we can substantially increase the speed at which we send emails and ingest events when it's us in an entirely closed-loop system.