Stop Treating Cold Email Like a One-to-One Channel. Run It Like Ads.
By Michael Saruggia, Fractional GTM Engineer ·
In 2015 you needed about 30 cold emails to book one meeting. By 2020 it was closer to 100. Michael Saruggia puts today's number somewhere between 500 and 1,000. Most teams read that as "cold email is dead." Michael reads it differently: email stopped being a one-to-one channel, and the teams still running it that way are the ones it stopped working for. In this session he walked through the cold email machine he runs for clients with Claude skills, from sourcing the full TAM to keeping hundreds of domains out of spam, and then spent the second half on the part he says decides whether any of it works: the offer. This recap covers each stage, the costs and thresholds he uses, and how he repackages offers for cold traffic.
Why did cold email stop working?
Michael opened with what he hears from almost every company he talks to: pipeline is down. Every channel got more expensive at once. Paid social, SEO, outbound and events all cost more, events got more crowded, sales cycles got longer, and so getting a meeting costs more too.
Tactics also wear out faster than they used to. A line like "I noticed you have 1,000 followers on LinkedIn" worked when it was new. Now anyone can generate it, so it doesn't land. Michael puts intent data in the same bucket. It may have worked six months ago, but people got used to it.
The usual complaint from SDR managers, sales leaders, CROs and CMOs is that LinkedIn and cold calling still work and email doesn't. Michael says that makes sense, because email changed. When reply rates were high enough, an SDR could reach out manually to a handful of people and it paid off. That version of the channel is gone.
What does it mean to run cold email like ads?
Michael's reframe is to treat cold email the way you treat Meta ads. It is still very effective, but it is a volume channel: you put a message in front of your decision makers' inboxes at scale, test creatives, and keep the ones that convert. It needs far more volume and much better infrastructure, and it changes what a "good" email even is.
His working floor is at least 30,000 cold emails per month, against a TAM of 30,000 or more reachable people. Below that, this playbook isn't the right fit.
What are the stages of the cold email machine?
| Stage | What happens | Tools and thresholds Michael uses |
|---|---|---|
| 1. TAM sourcing | Map every company you could sell to | Clay for LinkedIn data; niche directories elsewhere |
| 2. ICP verification | An agent scrapes each website and returns true or false | Claude skill, one prompt per check |
| 3. Email enrichment | Cost-ordered waterfall across five providers | About $0.012 per enrichment |
| 4. Email verification | Two verifiers, one of them catch-all capable | MillionVerifier and LeadMagic, bounce rate under 2% |
| 5. Copy testing | Test offers like ad creatives, scale the winner | 1,000 emails per test, scale at 1 interested per 300 |
| 6. Deliverability | Unbranded domains, rotation, spintax, health checks | SmartLead, retire domains under 1% reply rate |
He is tool-agnostic. At each step he picks whatever is most efficient for the price, which is why the stack mixes so many vendors.
How do you source your full TAM?
Michael wants the whole market mapped before anything else. His example: if you sell a social media tool to agencies, there are probably 50,000 US agencies that manage social media for clients, and he wants all of them in the list. Without the TAM number, he says, it is very hard to operate.
Where you pull the list from depends on where your TAM lives:
- LinkedIn: Clay is still the most efficient option when your buyers are on LinkedIn, with LinkedIn data included from about $150 per month. Agencies are a good fit because they want to be found there.
- Off LinkedIn: plenty of markets aren't on LinkedIn at all. Restaurants are on Google Maps, ecommerce brands are in Store Leads, funded companies are on Crunchbase, real estate agents are on realtor directories, and schools are on Niche.com.
So the first question is always where your TAM is listed and which API can pull it. Finding the right source mostly comes from experience, trying many and seeing what works. Michael's estimate is that about 10 different sources cover roughly 90% of cases.
How do you clean a dirty TAM list with an agent?
A raw TAM list, Michael warns, is bad. Search "agency" in Apollo or Clay and you get insurance agencies, brand agencies, cold email agencies and everything else that uses the word. You need to check each company against your actual ICP, and at 20,000 to 50,000 rows, that check has to be cheap.
His answer is a Claude skill. It takes a domain as input along with a prompt such as "given the domain name, check if this is really an agency, return true or false and nothing else." The agent visits the site, scrapes the basic information, applies the prompt and returns the answer. The output can be true or false, or a label like agency versus SaaS. In the agency example, he runs two checks: is it really an agency, and does it actually offer social media services.
Michael keeps a public library of about 200 skills, and the session covered roughly 20 of them. To use one, you copy the markdown file and install it as a skill in Claude. The skill then interviews you about the input and the prompt, so setup is mostly a conversation.
How does the email enrichment waterfall work?
Anyone can build a waterfall. Michael tuned his for cost, speed and accuracy.
- Ordered by cost. It runs five providers from cheapest to most expensive, including LeadMagic and Datagma, so the expensive ones only get called when the cheap ones miss.
- Built for speed. Lookups run in parallel batches. He has dropped a provider because its API rate limit was too slow for a 30,000-contact run.
- Small on purpose. He could add more providers, but each one costs time and money.
His current blended costs are estimates, since each contact can resolve at a different provider:
| Step | Approximate cost per contact |
|---|---|
| Enrichment | $0.012 |
| Validation | $0.039 |
Should you use a full data provider instead? Michael's view is that for cold email at scale, a full-coverage data provider is usually too expensive. For a typical sales team without a huge TAM, where reps want every field and SDRs manage their own lists, a full provider is the better choice. The waterfall also captures data at a point in time, and that data decays, so it shines for fast, cheap testing and high-volume runs rather than as a permanent system of record.
Why verify every email before sending?
Because bounces will wreck your sending infrastructure. Michael keeps bounce rates under 2%, and at 30,000 emails a month that means verifying everything.
He uses two verifiers. MillionVerifier is very cheap but can't handle catch-all addresses. LeadMagic can tell whether a catch-all email is valid. Running both lets him keep as many addresses as possible without letting bounces through.
The extra coverage pays off in replies. Pull 1,000 people from a classic provider like Apollo and you might find emails for about 200 of them, roughly 20%. That 20% isn't your whole TAM, and it's the same 20% everyone else is emailing. If you can reach people others can't find, you get more replies. So Michael would rather spend more on enrichment and verification and get a better result later.
How should you test cold email copy?
Here the ads comparison is literal. Michael sets up five to ten tests, and each one usually varies only the value proposition or the lead magnet. For a data tool, one test might offer to enrich 1,000 of your leads and another to verify your emails.
Each variant gets 1,000 emails. Any variant that produces at least one interested reply per 300 sends gets scaled.
The math behind that threshold:
| Input | Result |
|---|---|
| 1,000 emails per day at 1 interested per 300 | 3 to 4 interested replies per day |
| About 1 meeting per 4 interested replies | Roughly 1 meeting per day |
| A 30,000-contact TAM | 20 to 30 meetings per month |
In his words, it is like ads: you set up your creatives, push traffic and keep the one that's working.
Do intent signals still help?
Less than they used to, according to Michael. Most teams use the same handful of signals: open jobs, company growth, M&A, a new website, a product launch. Prospects have gotten used to messages built on them. He accepts that signals can still work in some situations, but his broader argument is about coverage. If you hit your whole TAM every 30 days, you reach the accounts showing signals anyway. And most people who are actually in market never show up on an intent platform.
How do you keep a high-volume sending machine out of spam?
"Google and Microsoft don't like us," as Michael put it. Staying out of spam takes several layers.
How should the domains and inboxes be set up?
- Never send from your primary domain. Set up unbranded domains instead. His example setup is around 25 domains with 200 addresses on top.
- Rotate domains. At high volume, a domain will eventually get blacklisted or blocked. When a domain's reply rate drops below 1%, something is wrong: retire it, buy a new one and warm it up for 15 days.
- Build double the capacity. If you plan to send 30,000 emails a month, set up twice that much infrastructure. If everything gets burned at once, you have a backup while new domains warm up.
Which sending platform should you use?
For cold email at scale, Michael only recommends three platforms: SmartLead, Instantly and EmailBison. He uses SmartLead. He sees little difference between them, since they copy each other's features. All three include warm-up. On SmartLead, sending enough volume also gets you into a shared warm-up pool of verified senders.
How does spintax keep messages from looking identical?
You can't send 10,000 copies of the same message. Michael's spintax skill takes a message written in Claude and rewrites it into SmartLead's spintax format, producing 20 to 25 variations of the subject line, body and even the signature. He writes the message in Claude, asks it to spintax and upload into SmartLead, and it arrives ready to send.
The same flow runs a spam word checker that swaps risky words for safe ones. "Free," for example, becomes "at no cost." Doing all of this by hand, he estimates, would take about 50 virtual assistants.
How do you monitor domain health automatically?
Michael's domain health checks run weekly through skills:
- Reply rate check. For every domain, look at the last 1,000 emails sent and flag any domain below a 1% reply rate.
- Blacklist check. Test each domain against around 40 blacklists. You can do this by hand with a blacklist lookup tool, but a skill handles it across hundreds of domains.
- Automatic swap. Flagged domains get replaced. A skill connects to his domain registrar's MCP server, asks how many domains you want and at what price, and buys them. He can get standard .com domains for as little as $3 each, which keeps a high-volume setup cheap.
Why is the offer the other 50% of the game?
Everything above is the technical half. Michael tells every prospective client that the best infrastructure is useless if the offer doesn't convert on cold traffic. That's why every new client starts with a $100 test of their offer. "I'm not in the miracle business," he said.
The objection he hears is "but my offer works, it got us to $1M ARR." Often that was warm traffic: people who have followed you on LinkedIn for six months or came in as referrals. Cold traffic is people who don't know you or trust you at all, and that's most of your TAM. You have two options. Become so well known that cold traffic isn't cold anymore, which is unrealistic for almost everyone, or package your offer so it works on cold traffic.
What does an intro offer look like?
Michael's own business is the example. His retainer, $3,500 to $5,000 a month, sells easily to inbound leads, content followers and intros. Sold cold, it doesn't work, because strangers won't commit that much to someone they don't trust. So he sells a $100 test first, delivers results, then asks whether they want to continue. Upselling the retainer afterward turned out to be almost frictionless.
| Business | Main offer | Cold-traffic version |
|---|---|---|
| B2B SaaS or data subscription | Annual subscription | Jump on a call and get 500 leads |
| Agency | $3K to $5K monthly retainer | 10 creatives for a $1,000 deposit, another $1,000 if one beats your current creatives, then the retainer |
| Ecommerce ads agency | "Do you want me to manage your ads?" | "Can I create 15 creatives for you, and you only pay if they outperform your current ones?" |
When the offer is right, Michael sees about five times the reply rate, three times the conversion and a 10x ROI on marketing spend. He sees it improve every step of the funnel: replies, interested leads, booked meetings and revenue.
Some companies already have an offer that works cold. Some need to modify it, some need an intro offer in front of it, and some only work with warm traffic.
Can a lead magnet work instead of a piece of your product?
Yes, with a trade-off. An offer outside your product, like a website copy review or an invitation to a dinner, will attract more unqualified people, but it can still work if the targeting is good. Michael's example is enterprise software: invite your whole ICP to a Michelin-star dinner. You won't close most of them, but you build relationships, and at enterprise deal sizes the math can still work.
Which value propositions work on cold traffic?
Michael counts about five value propositions that matter in B2B:
| Value proposition | How well it works cold |
|---|---|
| More revenue | The best, roughly 10 times more likely to work |
| Save money | Only when the savings are huge |
| Save time | Lower on his list |
| Less risk | Very difficult |
| More status | Very difficult, with rare exceptions |
On revenue: an AI consulting firm could pitch saving money, but pitching hidden cross-sell and upsell opportunities will usually do better. On savings: "I'll save you 10% on infrastructure" won't get him on a call, while a credible system that cuts 90% would.
Status is the rare exception. For someone selling ghostwriting services, Michael tested revenue, savings and visibility angles. The one that won was status: become the most well-known person in your niche.
Is this playbook realistic for an in-house team?
Michael was frank about the cost of entry. The three main sending platforms are roughly equivalent, so picking the tool isn't the hard part. Running everything together is. It has gotten complicated, and the depth of expertise you need to make it work is now very high. That is why he publishes his skills: the process is meant to be copied, so a team can run it with him or on its own.
Key Takeaways for GTM Teams
- Treat cold email as an ads channel. Plan for 30,000+ emails a month against a 30,000+ contact TAM, and judge copy the way you judge ad creatives.
- Map the full TAM, then verify it. Find where your market is actually listed, and use a cheap true-or-false agent to strip out companies that only match your keyword.
- Spend on enrichment and verification. A cost-ordered waterfall plus catch-all verification reaches the contacts everyone else misses, and keeps bounces under 2%.
- Use clear thresholds. Scale copy at one interested reply per 300 sends, and retire any domain that drops below a 1% reply rate.
- Don't rely on intent signals. If you cover your whole TAM every 30 days, you reach the accounts with signals anyway.
- Fix the offer first. Test it cold with a small budget, lead with more revenue, and put a low-risk intro offer in front of a big commitment.
Want to Attend the Next Session?
GTM Clawllective is a community where go-to-market operators share how they're actually using AI in the field, live and unscripted. The live discussion stays off the record, and every recap includes the speaker's demo video so you can see the build for yourself. To catch the next one live, join the community.
About the Speaker
Michael Saruggia is a Fractional GTM Engineer and a returning Clawllective speaker, having previously presented a competitor takeover playbook in Clay. He has worked in GTM engineering since the field's early days, going back to his first Clay demo with the founder. The numbers he shares come from first-party data from his own client work, and his approach is tool-agnostic, picking the most cost-efficient tool at each step.
He runs cold email programs for clients, starting each engagement with a $100 offer test, and publishes the Claude skills behind his process in a public library.
Connect with Michael on LinkedIn or browse his GTM AI skills library.
Watch the session replay: The New Cold Email Playbook with Claude