The Outbound and Data Stack We Run for Every Client (and Why Each Tool Earns Its Place)

If you are a founder or operator evaluating an outbound stack, you have probably noticed that most stack lists are written by vendors who sell one of the tools. That makes them useless for the decision you actually face: which pieces are load-bearing, which are interchangeable, and which you can skip until you have proof. We run the same core stack across every client engagement, and over 2.5M cold emails and 950K+ enriched contacts we have learned where each tool earns its place and where it does not.

This is a teardown, not a pitch. For each layer we name the job it does, the failure mode it prevents, and the honest tradeoff. By the end you should be able to evaluate any of these tools on your own, with or without us.

The sending layer: 52 mailboxes, dedicated domains, and why this comes first

Nothing else in the stack matters if your email lands in spam. This is the most common pain we hear, and it is almost always an infrastructure problem, not a copy problem. Most agencies and tools put you on shared sending infrastructure where you inherit other senders' reputation, which is why industry inbox placement sits around 60 percent.

We run dedicated sending domains with 52 warmed mailboxes spread across Google, Microsoft, and Azure. Spreading across three providers means no single provider's reputation swing can take a whole campaign down, and warming each mailbox before it sends real volume keeps reputation clean. The result is 98.5 percent average inbox placement and 0.15 to 0.9 percent bounce. The tradeoff is that this takes weeks to set up correctly and is not a button you press. If you want to understand the mechanics, see our breakdown of sending infrastructure, and run any draft through the free spam words checker before you send.

The data layer: Clay and waterfall enrichment instead of one stale Apollo export

The second most common complaint we hear is bad data: decayed Apollo lists, bounces, and contacts who left the company a year ago. A single data provider is a single point of failure, because every provider has gaps and every database decays. The fix is waterfall enrichment, where you query provider A, then fall to provider B for the records A missed, then C, and so on, taking the first verified hit.

We run this through Clay, which is the orchestration layer for the data, not a database itself. Across engagements we maintain 1,800+ production Clay tables that handle list building, multi-provider enrichment, and verification before a single send. That is how GearLocker ended up with a proprietary 66,000-school database that no off-the-shelf list could have produced. The honest tradeoff: Clay has a learning curve and credits add up if you do not design tables carefully. More on how we use it on the Clay enrichment page.

The signal layer: Perplexity and Claude turn triggers into relevance

Good data tells you who to contact. Signals tell you when, and that timing is what separates a reply from a delete. We watch for funding rounds, hiring spikes, tech-stack changes, and job changes, then route those signals into outreach so the message arrives when the prospect actually has the problem.

Two AI tools do specific jobs here. Perplexity does live research on a company or person so a message can reference something real and current. Claude writes the personalization from that research at volume, without the template smell that kills reply rates. This is not AI for its own sake: it is how we hit a 25 to 30 percent target positive-reply share, and how Chateau Constellation reached 177 interested wine importers by timing outreach around trade fairs. The tradeoff is that signal-based outbound needs the data and infrastructure layers working first, otherwise you are personalizing emails that land in spam. See how the triggers fit together on the signal-based outbound page.

The delivery and orchestration layer: HeyReach, n8n, and CRM sync

The last layer moves the message and handles what comes back. HeyReach runs the LinkedIn side using real profiles only, no fake accounts that get banned, so cold email and LinkedIn work as one sequence rather than two disconnected channels. CRM sync writes every contact, reply, and status into HubSpot, Salesforce, Pipedrive, Attio, or Monday so your team works from one record instead of a spreadsheet that drifts out of date.

Tying it together is self-hosted n8n, which handles the automation and AI reply classification: routing positive replies to a human, filtering out-of-office and not-interested, and keeping the CRM current. Self-hosted matters because you own the workflows and the data, not a vendor's account. This is the difference between a campaign that eats 15 to 20 hours of founder time a week and one that takes about an hour. Read more on sales automation and the full GTM systems overview.

Why we hand the whole stack over at day 90

The most important design decision is not a tool, it is ownership. The recurring frustration with agencies is "we own nothing when it ends." Our 3-month outbound pilot is built to invert that: domains, mailboxes, Clay tables, n8n workflows, and CRM are configured in your accounts, and at day 90 you keep all of it. The pilot proves the system with real campaigns first, like ATI's 78K emails and $300K+ CAD pipeline at a 37 percent positive reply rate, then you decide whether to run it in-house or keep us on.

Every tool above earns its place because it does one job the others cannot, and because you can take it with you. That is the test we apply, and the one we would suggest you apply to any stack a vendor proposes.