Learn / Lead lists
Lookalike lead lists beat filter guesswork
We sent over 5 million cold emails in the past 12 months, and the thing that decides whether a campaign books calls or dies quietly is almost always who ended up on the list. Not the subject line. Not the sequence length. Lookalike lead lists are the cleanest fix I know for that, because they start from companies you have already made money for instead of from a filter panel and a guess about who your buyer is.
This is the written version of the training below: the sourcing pipeline that turned 7 client domains into 5,400 prospects, the three-input research that validates the offer before you send anything, and the AI research brief we run instead of a pre-qualification form. Watch it, or read the whole build here.
The filter panel hands everyone the same list
Open a database, tick a few boxes, export whatever it returns. That is what every other agency, founder, and outbound operator in your market is doing this morning, with roughly the same boxes ticked, because the panel only offers so many. Which means the people on your list have already been hit three, four, five times this month by someone pitching the exact thing you are about to pitch. Their guard is up before you arrive.
Filters also encode a guess. You describe who you THINK your ideal customer is, then ask a database to go find more of that description. A category is a bucket, and you can't write to a bucket. If you are going to work a filter panel anyway, at least know which filters carry weight, which is what the Apollo walkthrough covers filter by filter.
Start from the customers you already got results for
Flip the starting point. Rather than describing the buyer, list the companies you have already produced results for and go find more companies that look like them. Odds are they are dealing with the same problem you already solved, which gives you a fresh list full of people predisposed to want what you sell.
Ocean.io was built off the back of its lookalike search. You hand it the domain URLs of your best-fit customers and it returns companies that match on dozens of features pulled off the websites themselves. Their data section says they re-crawl half of their domain index every single month, specifically to keep the features that produce good matches current. That is the part worth paying for: the match is made on what a company actually publishes about itself today, not on an industry code somebody typed in years ago.
The build: 7 domains to 5,400 prospects
- Pull the domains of your best case studies. In the build shown here, a cold storage campaign, that was 7 domains, and 7 was plenty. Seed it with customers you produced real results for, because the search copies whatever you feed it.
- Run the search on precise company matches. Relevance is graded A to C, so precise is the setting worth holding. Add two filters and stop there: company size and headquarters location. That returned 7,400 lookalike companies.
- Export the domains only, as a CSV. Ocean can find people too, but prospect-level data belongs in whichever database you actually store leads in.
- Upload those domains under company domain in your contact database. We use AI-Ark, which is Apollo-shaped but wins on data freshness and on keeping everything in one place. Those 7,400 companies held roughly 100,000 prospects.
- Cut by job title, and almost nothing else. The company-level work already happened in the lookalike search, so size, location, and match quality are baked into the 7,400 before you arrive. Titles took 100,000 down to 5,400 prospects.
- Fine-tune only if the campaign calls for it. Social follower count is the one I actually use. If email is going out alongside LinkedIn, filtering for high follower or connection counts means you are messaging people who still open the app, so connection rates hold up.
- Verify before it goes near a sender. Push the export through your verifier, or pull it into Clay, clean the data, and waterfall enrich the contacts whose business email is missing. Prospeo is the second layer we use for those.
Your best customers are the filter. Everything after that is fine-tuning.
Then validate the offer against three inputs
A better list only earns you the right to be ignored more precisely. The list decides who reads the email; the offer decides whether anyone replies. And in 2026 there are dozens of companies selling your exact service, so the free audit and the strategy call are both dead on arrival. What you need is something a prospect cannot glance at and say "we already have that covered." Three research inputs, run in this order:
| Input | Where you run it | What it proves |
|---|---|---|
| Prospect-level desire | Claude or Perplexity, with the exact company type and persona spelled out: job title, company size, all of it | What that person deals with day to day, personally and in the business. One or two pains sit lower than the rest, usually revenue-related |
| Competitor ads | The Meta ad library, for 5 to 10 competitors running paid | Which angles are already winning. 10, 20, 50 variations of one angle means somebody is spending money on it every day |
| Your own sales calls | Recordings of every deal you won and every deal you lost | The exact words your buyers use for the pain and for your solution. A great offer positioned in language they do not use will not land |
Then look for the commonalities. The research tells you what hurts. The ad library proves someone is already paying to solve it. The call recordings hand you the words to say it in, because your buyers are never quite the same as your competitors' buyers. Anything only one of the three supports is still a guess.
What that produced for a clinical ops staffing agency
The research said a VP of clinical ops does not have time to interview 30 candidates to find two good ones. The ad library showed every competitor pitching around finding talent fast, and none of them touching the interview load itself. The call recordings supplied the language. The offer came out as: three pre-vetted clinical ops candidates on your desk in 14 days, and if none are worth a final round, we replace all three for free.
You can read why that works. The perceived likelihood of achievement is high, because you are hand-delivering the thing they want instead of describing a process. The mechanism is already believed, because everyone knows staffing agencies work. The front-end offer lets them experience the mechanism before committing to anything. And the guarantee moves the downside onto our side of the table. After that we score every offer on seven axes, packaging, economic scalability, market power, mechanism strength, proof, volume, and guarantee, then go fix whichever one came back weak. Building the deliverable itself is its own job, and the front-end offer breakdown covers it end to end.
Kill the pre-qualification form, brief yourself instead
A list and an offer get you booked calls, and then a second problem shows up: hours a day spent talking to people who were never going to buy, or walking into calls knowing nothing about the person on the other end.
The standard fix is a pre-qualification form, five quick questions before anyone reaches your calendar. The problem is where the friction sits. Someone was interested enough to click your link, hit a form, and left. Plenty of them would have been a great fit, and you will never know which ones.
So do the qualification after the booking rather than before it. Ours triggers off a record being created in the CRM. A web agent researches five things: what the company does, its approximate size, who its main competitors are, its recent news, and the pain points a company like that is likely facing. It posts back a summary of all five, two to three sentences each, sharp enough to read in under a minute.
One that landed while the training was being recorded: an M&A firm in Manhattan. The brief gave their ICP focus, the revenue bands they work in, their cost structure, how long they have been in business, how many deals they have done, their competitive landscape, and their recent priorities. That is enough to decide whether the call is worth taking and what to actually talk about on it, and the prospect never filled out a thing. Compare it to a form asking for a phone number, a website, average monthly revenue, and typical deal size, which tells you less and costs you bookings. The pre-call qualification breakdown goes deeper on the whole flow.
Where lookalike sourcing falls apart
Two places, and I would rather you know both before you build anything. The first is that it needs results to point at. No case studies means no seed domains, and the search has nothing to match on, so a brand new offer is back on filters and intent signals until the first handful of wins land.
The second is market size. 7,400 companies holding roughly 100,000 prospects is about the floor where outbound compounds. Below that, there is not enough volume to find the winning offer, and the honest answer is that some businesses should not run cold email at all. Either way, the exported list is a hypothesis, not a verdict. Segmenting it, testing the pieces, and cutting what stays quiet is how you find out which slice of those 5,400 was the real list.
tl;dr: build the list out of the customers you already made money for, validate the offer against real pain, competitor ad spend, and your own call recordings, then let AI brief you on the prospect instead of making the prospect fill out a form.
FAQ
What is a lookalike lead list?
A company list built by handing a tool the domains of your best existing customers and asking for companies that match them on dozens of website and firmographic features, instead of guessing at filters. The premise is simple: companies that resemble the ones you already got results for are dealing with the problem you already solved, so they reply at a higher rate.
Is Ocean.io better than Apollo for building lead lists?
They do different jobs. Ocean.io is a company-level lookalike engine: seed it with customer domains, get matching companies back. Apollo is a contact database you filter by hand. The build in this training uses both ends, a lookalike search for the company list and a contact database for prospect-level data and job titles. Filtering alone gets you the same list as everyone else.
How many customer domains do you need to build a lookalike list?
Seven was enough in the build shown here, and it returned 7,400 matching companies once a precise match setting, a company size filter, and a headquarters filter were applied. Quality beats count. Seed the search with customers you produced real results for, not every logo you have ever invoiced, because the search copies whatever you feed it.
How do you validate a cold email offer before sending it?
Three inputs, then the overlap. Deep research on what your exact persona deals with day to day, competitor ads in the Meta ad library to see which angles already have money behind them, and recordings of your own won and lost sales calls for the language buyers actually use. Anything only one of the three supports is still a guess.
Should I use a pre-qualification form before a sales call?
Usually not. The form sits between interested and booked, and plenty of great-fit prospects simply will not fill it out after clicking your link. Run the qualification after the booking instead: an AI brief covering what the company does, its size, its competitors, its recent news, and its likely pain points tells you more without adding any friction.
When does lookalike sourcing not work?
When you have no results to point at, and when the market is too small. No case studies means no seed domains, so a brand new offer is back on filters and intent signals until the first wins land. And under roughly 100,000 reachable buyers there is not enough volume for the system to compound, no matter how precise the matching is.
If you would rather have the whole system run for you (lists, offer, copy, infrastructure, and someone working the replies), that's what we do: 3,000+ sales calls booked for clients in the past year. Or keep learning free: every training, grouped by topic, lives on the trainings hub, and the full video for this one is on YouTube.
PS - the 7 seed domains were not the biggest logos available. They were the accounts with real results attached. Those two lists are different more often than anyone admits.