Why outbound isn't working, and the three layers to build before you scale it

Outbound stops working when nobody has heard of you, and the fix is not more sequences. It is the three layers of go-to-market that make a cold message land warm: something that creates demand, something that makes you discoverable when a buyer looks, and a presence in the conversations where buying decisions actually form. Most teams skip all three and then blame the channel.

I ran an outbound product before this one. I got to watch, account by account, who won with it and who churned. The pattern was so consistent it eventually became the reason I stopped building outbound tools.

Why does cold outreach have such low reply rates now?

Volume is the obvious answer, and it is true as far as it goes. Every inbox is saturated. Deliverability is a full-time job. Buyers have learned to smell a merge field from the subject line. Reply rates that were 8 to 10 percent a few years ago are, in my experience, closer to 1 to 2 percent for a typical cold list, and a lot of that 1 percent is "unsubscribe."

But volume explains why the channel got harder for everyone. It does not explain why it still worked for some of my customers and not for others, on the same tool, with similar lists and similar copy.

The difference was recognition. When a prospect got a message from a company they had seen before, in a thread, in a post, in a ChatGPT answer, the reply rate was a different animal. Not because the copy was better. Because the message was a follow-up in the buyer's head, not an interruption. The customers who churned were asking the sequence to do the recognizing and the selling at the same time, and no sequence can.

What has to exist before outbound converts?

Think of go-to-market as a pyramid with five layers. Outbound is layer four. Closing is layer five. Almost every AI tool on the market automates four, and the pitch quietly assumes one through three are handled.

The bottom layer creates demand: posts on a cadence, short-form video, a brand people bump into. This is the slow, compounding part. It is also the part founders skip because it does not show up in a pipeline report for months.

The middle layer gets you discovered. Historically that meant search. Now it also means the AI engines. When a buyer asks Perplexity or ChatGPT "what should I use for X," the answer names three or four vendors, and buyers treat that list as a shortlist. If you are not in it, a competitor is, and you will never see the lost deal in any dashboard. I covered how to measure this in share of model.

The third layer is listening and participating. Public threads where someone with budget says "we're evaluating tools for X" or "our current vendor just tripled pricing." These are the highest-intent moments on the internet and they are visible to anyone paying attention. Show up there as a helpful person, not a pitch, and you build the recognition that layer four later cashes in. Finding those live conversations is a timing problem more than a search problem.

Build those three and outbound becomes the last mile: reaching people who already recognize you, at a moment you chose because you saw them raise a hand.

How do you build the base without a marketing team?

This is the honest objection. A seed-stage company with two founders and one engineer does not have anyone to post daily, monitor six platforms, and run GEO checks. That was the actual constraint that made "just buy an outbound tool" so seductive.

A few things I would do, roughly in order.

Start with listening, because it pays back fastest and teaches you the market's vocabulary. Write one sentence describing the conversation you care about, the way you would say it to a coworker, and go find where it happens. Reddit and Hacker News for developer and founder audiences, LinkedIn and X for sales and marketing buyers, Bluesky and Threads for the people who left X. Reply like a person. Track who you talked to so you never message them cold later.

Next, check the AI engines for your category's ten or so buying questions. Ask each one in ChatGPT, Perplexity, and Gemini. Write down who gets named. Then look at which sources the answers cite, because those are usually threads you could have been in. That list is your content plan for the next month.

Then produce on a cadence you can actually keep. Two posts a week beats a daily plan you abandon in three weeks. Pull the topics from what you heard while listening. The phrasing people used in threads is the phrasing that ranks, and it is the phrasing the engines quote.

Only then scale outbound, and when you do, feed it the context you built: the scored conversation, the person, whether you already talked. An outbound agent handed that is a different product from one handed a CSV.

Is this just "do marketing first"?

Partly, yes, and I understand why that lands badly on a founder who needs pipeline this quarter. The nuance is that the three layers no longer need three hires. Listening across platforms, checking the engines, drafting in your voice, and logging outreach are all things one operator can run continuously now, and that operator can be software.

Disclosure: that operator is what I am building at Oetzi. It exists because I watched outbound fail for a very specific reason, and the reason was never the outbound.