A cold email tech stack is only worth what its parts do together. Ours is seven tools. For one client we averaged 30 to 35 qualified leads a month, and after rebuilding around better enrichment and automation we averaged 70 to 80, on the same outreach volume with the same team.

The difference was not a better tool. It was working out what each platform was built for, and stopping ourselves from asking it to do the rest. Here is the whole stack, what each part costs, and the limits we ran into.

What actually changed

We used to verify email addresses by hand. Copy an address, paste it into a checker, wait, copy the result back into a spreadsheet, 300 times over. It took days, and the work was worthless the moment the list changed.

The stack now processes over 1,000 contacts a day. The speed is not really the point. The point is that each tool covers a gap the others leave, which is the only reason the whole thing holds together.

Apollo, where every list starts

Apollo is the front of the operation. Its B2B search is the best we have used: millions of contacts, with filters granular enough to isolate a role, a company size, an industry, a tech stack, or a funding stage.

It also keeps list building, verification, engagement tracking, and sequencing in one place, which removes a lot of switching. Data accuracy sits around 85 to 90% in our testing. Good, not perfect, which is why verification and enrichment sit downstream of it.

We start in Apollo and usually finish there too, coming back to enrich qualified leads once the rest of the stack has done its work.

Clay, and the credit trap

Clay handles enrichment and anything that needs data pulled from several sources at once: firmographics, technographics, social profiles, funding. More sources mean richer profiles and better angles to write from.

The flexibility is what sold us. We clean Apollo lists in it, build new ones from scratch, connect other tools’ APIs, and run a different workflow for every client without the tool fighting us.

Running Clay enrichments through the OpenAI and Anthropic APIs is what took that client from 30 to 35 qualified leads a month to 70 to 80.

Two warnings. Garbage in, garbage out: bad domains and wrong titles produce enrichments just as bad, so cleaning is not optional. And credits vanish. We once enriched 5,000 contacts with every data point Clay offered, spent $300, and never used 90% of it. Now we only enrich what changes the message.

Smartlead, sending and deliverability

Smartlead sends. Whitelabel matters when you run an agency, the API connects the rest of the stack, and unlimited mailboxes mean campaigns scale without hitting a plan ceiling.

Deliverability is the part that is not negotiable. The infrastructure keeps mail out of spam, which matters more than any clever subject line.

The frustration is campaign management. The sending side is solid, but basic contact and campaign handling has rough edges that linger longer than they should. Juggling dozens of campaigns across clients, that friction adds up.

DeBounce, the insurance policy

Every bad address damages your domain reputation. A high bounce rate tells providers you do not maintain your lists, and it pushes your legitimate mail into spam alongside the rest.

DeBounce catches invalid addresses, catch-alls, disposables, and role accounts before send. Most of our campaigns hold bounce rates under 1%.

It takes weeks to rebuild a burned domain. DeBounce starts at $10 a month. That is the whole argument.

GPT and Claude, research to copy

GPT does the research: reading sites, pulling company detail, finding angles that would take a person far longer to reach.

Claude helps with the copy, but it does not write the emails. It writes one or two sentences that carry the personalisation. The argument is still ours.

Approval happens at the template level, not per email. A sequence might have three steps with four variations on the first, three on the second, and one on the third, and we review every variation. Nobody reads every individual send, and no process should pretend otherwise.

The real check happens earlier, during enrichment in Clay. Keyword filters flag outputs carrying words we avoid, a safety model reviews the AI output, and anything either one flags gets rewritten by hand. Tone is what we watch hardest.

Make, the connective tissue

Make connects everything. Every qualified lead runs through it on the way to the client.

When a positive reply lands in Smartlead, a label triggers a Make workflow that enriches the lead from Apollo, writes it to Notion, updates the client’s Google Sheet, posts to their Slack, and emails a summary. The whole thing takes seconds.

If Make goes down, lead handoff stops. That is how central it is.

Notion, the command centre

Everything that is not sending lives in Notion: tasks, lead databases, client reports, internal documentation. Shared pages give clients somewhere to check progress and read strategy docs without an inbox full of attachments.

How the tools fit together

Apollo finds prospects matching the ICP. The list exports to Clay for validation, enrichment, and personalisation. It uploads to Smartlead and the campaign sends. A positive reply triggers Make, which distributes the lead to Notion, Google Sheets, Slack, and email.

Quality control runs alongside that. Clay pulls company data, GPT drafts the angle, keyword filters and a safety model review it, anything flagged gets rewritten by a person, Claude refines the tone, and a human signs off on every variation before the campaign goes live.

What it costs

The most common question we get is what this actually costs to run. Here is the honest breakdown at three levels.

Beginner: $279 a month

Everything you need to start properly. If you are new to outbound, start here.

That gives you capacity for 2,000 leads and 6,000 emails a month. Reply rates of 1 to 5% are typical. At the industry average of one qualified lead per 250 contacts, that is eight to ten qualified conversations a month if everything is done well.

Explorer: $650 a month

Our recommended tier for a growing business, and the point where outbound starts moving pipeline.

30,000 leads, 150,000 emails a month, and enough volume to test properly and scale what works. With good execution, reply rates above 5% become achievable.

Expert: about $10,820 a year

Agency scale. At this point decisions get made on what the work needs, not what the tools cost.

What we learned

Start simpler than you think. You can run basic campaigns on Apollo alone, since it has sequencing built in. If you care about data quality and deliverability, start at the beginner tier instead.

Reply rate does not tell the whole story. We have seen 8% reply rates generate no revenue, and 2% reply rates close six figures. Count qualified conversations, not replies.

Tools are not the only cost. Domains run $10 to $15 each and you need several, plus email hosting, plus the two or three months it takes to get genuinely good at any of this.

Data quality beats everything else. Perfect copy, perfect timing, and perfect deliverability still come to nothing if you are writing to the wrong people at bad addresses.

The honest version

This stack works for us, at our scale, with our clients. Six months ago it looked different, and in six months it will have changed again.

The tools are not the point. Knowing what problem each one solves, and what to ignore in each of them, is the point. Working that out took months, and it matters more than knowing every feature.