You Don't Have a List Problem. You Have an ICP Problem.
Most early-stage founders open Apollo looking for more prospects. That's backwards. Pre-PMF, the bottleneck isn't list size — it's that you genuinely don't know which of three or four plausible customer profiles will actually buy. Apollo's real value at this stage isn't its 200M+ contacts; it's the filters. Headcount, funding stage, tech stack, active hiring signals — the exact dimensions along which ICP hypotheses differ.
That makes Apollo less a list vendor and more a segmentation lab: define three competing ICP hypotheses as saved personas, pull a small sample of each, and run three parallel sequences to see which segment actually replies. What Apollo doesn't supply is the sending infrastructure those sequences run on — the secondary domains, mailboxes, and dedicated IPs that determine whether your experiment ever reaches an inbox. That's the half ColdRelay covers, and this guide shows how to wire the two together for disciplined ICP discovery.
Why Run Apollo on ColdRelay Infrastructure
Apollo bundles the database, sequences with email, call, and LinkedIn steps, and the ability to link your own mailboxes under Settings → Mailboxes. But it sends from whatever mailboxes you connect — it doesn't provision domains, assign IPs, or control the deliverability of the accounts themselves. For an ICP experiment, that gap is fatal in a quiet way: if your emails silently land in spam, every segment looks equally dead, and you'll conclude the wrong thing about your market.
ColdRelay closes that gap. You provision mailboxes on secondary domains, hosted on isolated Azure tenants with dedicated IPs, with SPF, DKIM, and DMARC pre-configured — live in about an hour, with no warmup waiting period because warmup runs continuously inside each mailbox's daily budget of 4 sends (2 outbound + 2 warmup). With 95%+ inbox placement, a zero-reply segment means the message or the segment failed — not the plumbing. That's what makes the experiment readable.
The pairing is additive, not competitive: Apollo is the data and sequencing layer, ColdRelay is the infrastructure layer underneath it. Apollo finds and sequences the prospects; ColdRelay makes sure the test actually reaches them.
Visit Apollo →Connecting ColdRelay Mailboxes to Apollo
Provision a small experiment pool on ColdRelay
ICP discovery doesn't need volume — it needs clean signal. Most pre-PMF founders provision 15-30 mailboxes on one secondary domain (ColdRelay supports 100-150 mailboxes per domain, so there's plenty of headroom for the scale-up later). Everything lands on isolated Azure tenants with dedicated IPs in about an hour, DNS (SPF, DKIM, DMARC) already configured.
Link the mailboxes under Settings → Mailboxes in Apollo
In Apollo, go to Settings → Mailboxes and connect each ColdRelay mailbox so Apollo can send sequences through them. Once linked, set each mailbox's daily send limit in Apollo to 2 outbound emails per day — mirroring ColdRelay's per-mailbox budget of 4 sends/day total, split 2 outbound + 2 warmup. ColdRelay's continuous warmup handles the other 2; don't layer additional warmup on top.
Define three ICP hypotheses as saved personas
This is the step that separates discovery from spraying. Build three saved personas in Apollo that differ on a real dimension — say, seed-stage SaaS ops leads vs. mid-market RevOps at 200-500 headcount vs. agencies showing active sales hiring. Use the filters that encode your actual hypotheses: headcount, funding stage, tech stack installed, and hiring signals. Save each as a persona and a saved search so the segment definition is explicit and repeatable.
Pull a small sample per persona — not the whole list
From each saved search, add roughly 100-150 contacts to a dedicated list. Resist pulling thousands; the goal is a sample large enough to read reply patterns, small enough that a wrong hypothesis costs you two weeks instead of a quarter. Bigger pulls come after a segment proves out.
Run three parallel sequences and read the replies
Build one Apollo sequence per persona with the same core problem framing, so the variable under test is the segment, not the copy. Email steps carry the experiment; add Apollo's call or LinkedIn task steps for the segment you suspect is strongest if you want a second touch channel. At 2 outbound sends/day per mailbox, a 20-mailbox pool moves 40 emails a day — each 150-contact segment completes its first touch within a couple of weeks, and the replies tell you where to double down.
The Pre-PMF Apollo Playbook
Hold the message constant, vary the segment
Most A/B advice tells you to test copy. At the ICP-discovery stage, invert it: write one honest articulation of the problem you solve and send it to three different Apollo personas. If the framing lands with one segment and dies with the others, you've learned something about your market that no copy test reveals.
Filter on signals, not just titles
Two companies with identical titles can be opposite buyers. Use Apollo's signal filters to encode timing into the segment itself — companies actively hiring for the role you augment, companies that just raised, companies running a tech stack yours plugs into. A hypothesis with a signal baked in tests 'who has this problem right now,' which is the question pre-PMF outbound actually needs answered.
Score replies by quality, not count
A 4% reply rate of polite brush-offs loses to a 2% rate where prospects describe the problem back to you in their own words. Tag every reply by depth — brush-off, curious, problem-confirmed — and pick the winning segment on problem-confirmed replies. Those messages are also free discovery research: prospects telling you how to write the next campaign.
Don't buy the big list until a segment earns it
Apollo makes it one click to go from 150 contacts to 15,000, which is exactly why discipline matters. Only after one persona clearly wins do you expand its saved search and scale the pull — and because ColdRelay provisions additional mailboxes in about an hour at 100-150 per domain, capacity grows the same week conviction does. List spend follows evidence, never precedes it.
Typical Pre-PMF Startup Benchmarks (Apollo + ColdRelay)
| Metric | Benchmark | Notes |
|---|---|---|
| Inbox placement rate | 95%+ | Dedicated IPs and isolated tenants keep the experiment readable — zero replies means segment, not spam folder |
| Reply rate spread across 3 ICP tests | 1-8% | The spread is the result — a winning segment typically out-replies the losers 2-3x |
| Outbound capacity per mailbox | 2/day | 4 sends/day total per mailbox — 2 outbound + 2 warmup |
| Sample size per ICP hypothesis | 100-150 contacts | Large enough to read patterns, small enough to keep a wrong hypothesis cheap |
| Time to a validated segment | 2-4 weeks | ~60 minutes to provision on ColdRelay; three parallel Apollo sequences complete first touches within two weeks at 40 sends/day |
What It Costs: Apollo + ColdRelay
You pay per mailbox per month for the infrastructure, with volume tiers that drop as you scale (see the table below). DNS, dedicated IPs, and isolated Azure tenants are included — and a 15-30 mailbox experiment pool keeps the discovery phase comfortably inside a pre-seed budget.
Apollo is billed separately on its own subscription, covering database access, export credits, and sequencing. At the discovery stage you're pulling small samples, so credit consumption stays low until a segment earns a bigger list.
Both bills stay small while you're still guessing and grow only after the market answers. Infrastructure scales with mailbox count, Apollo spend scales with list size — and the whole point of this playbook is that neither should scale before a segment proves out.
| Mailboxes | ColdRelay price / mailbox / month |
|---|---|
| 1–199 | $1.00 |
| 200–999 | $0.85 |
| 1,000–4,999 | $0.70 |
| 5,000+ | $0.55 |
Each mailbox sends 4 emails per day — 2 outbound to prospects + 2 warmup. ColdRelay provisions mailboxes on isolated Azure tenants with dedicated IPs; Apollo handles the sending, sequencing, and inbox rotation on top.
Frequently Asked Questions
Is ColdRelay an alternative to Apollo?
No — they're complementary layers of one stack. Apollo provides the contact database, personas, saved searches, and sequences with email, call, and LinkedIn steps. ColdRelay provides the secondary domains, mailboxes, and dedicated IPs those sequences send from. You use both together: Apollo on top finding and sequencing prospects, ColdRelay underneath making sure the emails land.
Apollo already lets me link mailboxes — what does ColdRelay add?
Apollo's Settings → Mailboxes connects accounts; it doesn't create them or control their deliverability. ColdRelay supplies the accounts themselves: mailboxes on secondary domains, isolated Azure tenants with dedicated IPs, SPF, DKIM, and DMARC pre-configured, delivering 95%+ inbox placement. That matters most during ICP testing — if deliverability is shaky, every segment looks dead and the experiment teaches you nothing.
How many mailboxes do I need to test three ICP hypotheses?
Around 15-30. At 2 outbound sends/day per mailbox (4/day total including 2 warmup sends), 20 mailboxes gives you 40 outbound emails a day — enough to work through three 100-150 contact segments in parallel within a couple of weeks. Since ColdRelay supports 100-150 mailboxes per domain and provisions in about an hour, scaling up after a segment wins is a same-week move, not a re-architecture.
What if none of the three segments reply?
First, rule out the plumbing — which is exactly why running the test on ColdRelay infrastructure matters. With dedicated IPs, isolated Azure tenants, and 95%+ inbox placement, a silent campaign is a real market signal rather than a spam-folder artifact. From there, the fix is usually the problem framing, not the channel: rewrite the core message, keep the same three Apollo personas, and rerun. Three cheap failed hypotheses in a month is fast learning, not failure — that's the entire economics of testing small before buying big lists.