Deliverability
Deliverability, diagnosed not guessed.
When reply rates move, most teams rebuild everything and hope. We isolate the one variable that broke, domain, copy, tenant, infrastructure, or inbox type, and change only that. This is the framework we operate.
001/ECONOMICS
Judge by economics, not reply rate alone
Reply rate is a lever, not the verdict. What decides whether cold email works is the chain from spend to profit. A half percent reply rate on a large market can out-earn a two percent reply rate that costs three times as much. Judge cold email like paid ads. Money in, calls and customers out.
- 01
Spend
Infrastructure, data, sending, and tools.
- 02
Replies
Cost per reply and positive reply rate.
- 03
Calls
How many positives turn into booked calls.
- 04
Closes
How many calls turn into customers.
- 05
Profit
Customer value minus cost to acquire and fulfill.
002/METHOD
Match the method to your market
A market of twenty thousand and a market of ten million should not be run the same way. Pick the playbook based on how many real prospects you can reach and how hard you need to protect the list.
Conservative
Small market or client work, around 20,000 targets
1 to 2 cold per inbox per day
Protect the list. High warm to cold ratio, at least 3 structural variants, stop at step 2, and a new lead magnet every month. Consider Group A and B rotation so domains get cooldown weeks.
Mid sized
Normal agency market, around 150,000 to 200,000 targets
2 to 3 cold per inbox per day
Build enough capacity to loop the market monthly, roughly 4,000 to 6,000 a day. New variants every week. Still stop at step 2. Test rotation only if burn has been high.
Volume
Huge market, hundreds of thousands to millions of targets
2 to 5 cold per inbox per day
Judge by cost per call. Often hit once and loop every 1 to 3 months. Switch a domain only when it falls below half a percent reply rate over a month. One master campaign fed new data daily.
003/OPERATING RULES
The operating rules
These lower the chance of creating a deliverability problem in the first place. Do them before you ever need to diagnose one.
Real copy variants, not tiny edits
Change the shape, the angle, and the feel so variants read like different people wrote them. A reworded email is the same fingerprint. Keep at least 3 structural variants and add new ones every week.
Heavy spintax, not a substitute for variants
Spin every repeated line. An unspun line is part of the fingerprint. Deep spintax across 3 structural variants gives you billions of line level combinations, but it does not replace fresh structural copy.
Start low on volume
One cold email per inbox per day is the sensible default. Test two when the data supports it. The ceiling of 5 per inbox is earned after the offer is proven, not a starting point.
Keep warm traffic well above cold
Run 5 to 6 warm emails per inbox for every 1 to 2 cold. More positive activity than cold activity lowers the effective complaint rate across the inbox.
Engineer positive engagement
Engagement is the biggest signal. Lead with real value, a relevant resource or audit offered with no pressure, instead of asking for a call up front. The offer that gets replies protects the domains.
Remove spam words and risky phrasing
Read the email aloud. If it sounds like a salesy cold email, it flags faster. Run copy through a checker and pull risky words before you scale a variant.
Redirects off by default, then tested
Split five domains with redirects and five without. If the redirect group underperforms, you have found a real infrastructure pattern. Mix registrars, sender names, and cheap TLDs like .com, .co, and .org.
Group A and B rotation when domains burn fast
Warm every domain for two weeks, then alternate weekly. One group sends cold while the other stays warm only, then swap. Each group gets cooldown. Not required for everyone.
004/WARMUP
Warm before you send
New domains are warmed for at least two weeks before any meaningful cold volume, and warmup keeps running while a campaign is live.
- Warm for longer if engagement is weak or domains are burning faster than usual.
- Match the warm pool to the provider mix of your target list. If most of your cold list is on Google but most of your warmup lands on other providers, your effective Google warmup is a fraction of what it looks like.
- Use a quality warm pool. A poor free pool with weak coverage means lower effective warmup, which means you must send fewer cold emails.
005/DIAGNOSIS
Isolate the cause with a cross section
A reply rate drop is only useful once you know which variable it follows. You get that signal only when the same domains run across several campaigns and each campaign draws from a mixed pool of domains. Build the test as a grid, then read the shape of the drop.
- A row drops
- One domain is bad across every campaign it is in. The domain is burned.
- A column drops
- One campaign is bad across many domains. The copy or offer is the problem.
- A block drops
- The drop follows a tenant, TLD, registrar, or redirect. The cause is infrastructure.
- An inbox type dips
- One provider is worse for a period while the rest hold. It usually recovers.
- Holdouts stay clean
- Untouched inboxes show whether damage stayed local or spread.
The rule is overlap without total dependency. A domain appears in more than one campaign, but not every campaign, and some inboxes stay held out. Without that, you cannot tell whether the domain burned, the copy burned, or a provider had a bad week.
006/REMEDY MATRIX
Different break, different fix
We isolate the cause first because the fix changes completely depending on what broke. Change only the thing the pattern points at.
| Pattern | Likely cause | Fix |
|---|---|---|
| Row goes bad | Domain burned | Switch domains. Move volume off the burned domain and leave healthy ones unchanged. |
| Column goes bad | Copy or offer | Make a big change to angle, structure, opening, and CTA. Not a few words. |
| A tenant drops together | Tenant pattern | Switch the tenant. Compare against other tenants before touching unrelated campaigns. |
| A shared setup drops | Infrastructure | Redo those domains and change the pattern. Rarer, and it usually shows at scale. |
| One inbox type dips | Provider period | Weather it and rebalance volume across providers. It normally recovers within a couple of weeks. |
007/QUESTIONS
Is a low reply rate always a problem?
No. On a large market, a low reply rate can still produce cheap calls and strong margin. We optimize reply rate, but we judge the campaign on cost per call, cost to acquire, and customer value.
How long before new domains can cold send?
At least two weeks of warmup before meaningful cold volume, longer if engagement is weak. Warmup keeps running while the campaign is live.
Why do you cap sends so low per inbox?
Because the future of cold email is more inbox footprint at lower volume per inbox, not squeezing more cold sends through each one. Lower per inbox volume with more inboxes burns less and gives you room to test.
How do you know what actually caused a drop?
We structure campaigns as a cross section of domains and campaigns before anything breaks, so a drop forms a visible shape. The shape tells us whether it was the domain, the copy, the tenant, the infrastructure, or the inbox type.
Want this run for you?
Sending infrastructure, warmup, rotation, and deliverability operated to this standard is our Infrastructure service. The audit shows where your current setup is leaking first.
Request the infrastructure audit