We burned our sending domain in a week

We came back from SaaStr with 125 ICP-fit demos and a follow-up list, did what you'd expect, and inside a week Google Postmaster Tools had helply.com sitting at "Bad."

Nobody writes about this part. Everyone posts the conference photos. Then a quarter later they're wondering why their email isn't landing and quietly buying a second domain.

So: what happened, what we did, and what I'd do differently.

What we did wrong

We sent everything from one domain.

Product email, transactional notifications, the newsletter, sales follow-up, and a sudden spike of post-conference outreach, all going out of helply.com. For a while that's fine because the volume is low and the mix is mostly wanted mail.

Then you 10x the volume in a week and shift the mix hard toward mail that gets ignored, and the domain's reputation is an average that includes your password resets.

The spike is what did it. Not the volume in absolute terms, the shape of the volume. Going from a steady low baseline to a sudden burst reads as exactly what a compromised domain looks like, because that's what a compromised domain does.

How we found out

Later than we should have. A few deals went quiet in a way that felt off, someone checked, and Postmaster had been telling us for days.

That's the actual lesson buried in here: nobody owned deliverability. It wasn't anyone's job, so it wasn't anyone's dashboard, so it went bad silently while 4 people were busy being pleased about the pipeline.

The fix

Three parts, none of them fast.

Separate the mail streams. Transactional, product notifications, marketing and sales outreach now go out on different subdomains with their own authentication and their own reputation. If sales outreach tanks, password resets still land. That separation should have existed from day one and it's cheap to set up before you need it and painful after.

Let the main domain rest. We pulled volume right down on helply.com and left it there while the reputation recovered. It takes weeks. There's no button.

Ramp back gradually. Small daily volumes on the new subdomains, increasing on a schedule, watching the numbers. Warming a domain is boring and there is no way to compress it, which is a very unpleasant thing to learn while you're sitting on the warmest list you've ever had.

What it cost

Weeks of degraded follow-up on the exact leads we'd spent 100k to generate.

I can't put a clean number on it because you can't see the emails that didn't land. That's the nasty part of deliverability problems. The failure is invisible and looks identical to disinterest, so you spend the same period concluding that your follow-up messaging is weak when actually it's in spam.

The real cost is that half the value of conference intent is gone in 3 weeks anyway. We lost a chunk of that window to a technical problem that had nothing to do with the pitch.

What I'd do before the next event

Set up the subdomain architecture now, before the volume exists. Warm the sales subdomain over the 6 weeks before the conference, so on day one after you land you've got headroom.

Put deliverability on someone's name. Not as a project. As a number they look at weekly, same as pipeline.

Spread the follow-up over more days than feels right. The instinct is to hit everyone in 48 hours because intent decays. Intent decaying is real and so is the reputation cliff, and you have to trade them off deliberately rather than by accident.

Where it stands

Reputation's recovered. Streams are separated. We're running the ramp schedule into Customer Success Summit in September and I expect to be fine.

I'd like to tell you we have this fully solved. What we actually have is one incident, a fix that appears to work, and a metric someone now watches. That's most of operations, honestly.

Check your Postmaster before your next conference. It takes 2 minutes and it'll be more useful than another look at the booth design.