Every year someone declares cold email dead, and every year the teams running it properly keep filling pipelines. What has actually changed is not whether the channel works — it is the minimum standard for doing it competently. Cold email best practices moved, and a lot of teams did not move with them.
This is what the bar looks like now: the rule changes that raised it, the fundamentals that never moved, and the tactics that used to work and now actively cost you replies.
What Actually Changed
Bulk sender requirements became enforceable
The structural change was Google and Yahoo enforcing sender requirements from February 2024 and tightening since. Google’s email sender guidelines now require SPF, DKIM and DMARC from anyone sending more than 5,000 messages a day to Gmail, a working one-click unsubscribe on bulk sends, and a spam rate below 0.10% — with active filtering above 0.30%.
Teams without that infrastructure were filtered into spam and largely never recovered. If you have not audited your DNS recently, our MX, SPF, DKIM and DMARC setup guide covers every record.
The AI content flood trained everyone’s pattern recognition
AI-generated outreach became ubiquitous, and prospects got fast at spotting it. Generic openers, hollow personalisation tokens and feature-list value props now read as automated on sight. The scarce thing is specificity — evidence that you actually know something about this company’s situation.
Filtering became context-aware
Spam detection no longer scores content in isolation. Sender history, engagement patterns, content signals and sending behaviour are weighed together, and Microsoft documents the same layered approach on its side. Five hundred sends from a fresh domain on day one gets filtered regardless of how good the copy is. Infrastructure and warmup are preconditions, not optimisations.
The Cold Email Best Practices That Never Changed
Cold email best practices did not get more clever. They got stricter.
Surface tactics shift constantly. The cold email best practices underneath performance have not moved in a decade, and every one of them predates the rule changes above.
ICP precision beats volume
The highest reply rates do not come from sending more. They come from sending to better lists. A tightly defined ICP filtered on firmographic and technographic signals consistently beats a broad list at ten times the volume. If you cannot describe your ideal customer in three specific attributes — and “B2B SaaS” is not one — your targeting is too broad. Start from your ICP definition before every campaign.
The first two lines are still the whole audition
Email clients show a subject line and 60–80 characters of preview text. That is what you get. The emails that earn opens lead with something concrete — a trigger event, a number, a direct observation. Not “Hope this finds you well.” Not “I came across your profile.”
Short, single-ask emails
The first email should be under 100 words with one value proposition and one call to action. Everything else belongs later in the sequence. Long opening emails underperform in every test we run: they signal low confidence and loose targeting. Needing 300 words to explain your value is a positioning problem, not a copy problem.
Structured testing beats intuition
Teams running disciplined A/B tests on hooks, value propositions and calls to action consistently outperform teams shipping a single best guess. The discipline that matters: one variable at a time, and at least 100 sends per variant before you read anything into the result.
Where The Practices Moved
Segment specificity replaced token personalisation
The old approach was Hi {{first_name}}, I noticed {{company}} is hiring for {{role}}. The current one is segment-level: find a cluster of accounts sharing a real signal — recently funded, running a specific tool, hiring into a specific function — and write one email that speaks to that shared context. It scales better than per-contact research and converts better than token replacement.
Domain architecture became a performance variable
Running all outreach from one domain is a deliverability risk that caps your reply rate outright: if that domain is filtered, the whole pipeline stalls. Three to five sending domains, none above 200–500 sends a day, is now table stakes for anyone at scale.
Sequences got shorter
Three to four emails is the current standard; five was normal a couple of years ago. Past the fourth, returns fall off quickly and complaint rates climb. The one exception is a final resource-led email — a genuinely relevant guide or data point with no hard ask — which often pulls replies from people who were not ready earlier.
What to Stop Doing
| Outdated tactic | Why it costs you | Replace with |
|---|---|---|
| Sending from your primary domain | One complaint wave takes transactional email down with it | Dedicated sending domains only |
| First-name and company tokens as “personalisation” | Prospects recognise it instantly | Trigger-based or segment-level specificity |
| Long first emails, 300+ words | Lower open-to-reply conversion | Under 100 words, one ask |
| Buying unverified lists | Bounce rates destroy sender reputation in days | Build and verify before every campaign |
| Skipping warmup on a new domain | New domain plus high volume equals blacklist | Four to six weeks of warmup, always |
| Seven or more follow-ups | Diminishing returns and rising complaints | Three to four emails per sequence |
| Generic subject lines | “Quick question” is now a filter signal in itself | Specific, curiosity-led subject lines |
What Good Looks Like in Numbers
contacted 5,000 ├ delivered to inbox 4,750 95% placement, warmed domains ├ opened 2,140 unreliable — privacy inflates it ├ replied 330 6.6% total reply rate └ positive replies 160 3.2% ← the only one worth managing
Across active client campaigns running the system above:
- 95%+ inbox placement on properly warmed domains
- 5–15% total reply rate on well-targeted sequences
- 2–8% positive reply rate
- First qualified meetings inside 30–60 days of launch
None of that is exceptional. It is what the channel produces when it is built as a system rather than run as a series of one-off campaigns.
How to Apply These Cold Email Best Practices
In order, because each step assumes the one before it:
- Audit DNS on every sending domain — SPF, DKIM, DMARC, MX
- Move outreach off the primary domain onto three to five dedicated ones
- Warm every new domain for four to six weeks before real volume
- Tighten the ICP until you can state three specific qualifying attributes
- Rewrite email one to under 100 words with a single ask
- Cut sequences to three or four emails
- Run one A/B test at a time, 100 sends per variant minimum
Compliance sits underneath all of it — the FTC’s CAN-SPAM compliance guide sets the floor for commercial email in the US, and the one-click unsubscribe requirement now overlaps with it directly.
Further Reading
The eight-second rule for cold email hooks
How to personalise cold email at scale
Building a cold email follow-up sequence
How to double your cold email reply rate
The Bottom Line
Cold email best practices did not get more clever, they got stricter. The channel now rewards teams that treat it as an engineering problem — infrastructure, targeting, testing, iteration — and punishes teams treating it as a volume game. That is a higher floor, but it is also a higher ceiling, because the competitors who never updated are filtered before they reach the inbox you are competing for.
The practical read: fix infrastructure first, because nothing downstream can compensate for it, then tighten targeting, then shorten and test the copy. If you would rather have that system built and run for you, apply for the GTM Pilot.
FAQ
Is cold email still effective?
Yes, with proper infrastructure, tight ICP targeting and tested copy — teams sustaining 5–15% reply rates are not rare. The teams declaring the channel dead are almost always running it without warmup, without authentication records, or without any structured testing.
How many cold emails can I send per day?
A safe ceiling is 200–500 per domain per day on a fully warmed domain. With five sending domains that is 1,000–2,500 a day without concentration risk. Exceeding it on a single domain buys a reputation problem that takes weeks to unwind.
What open rate should I aim for?
Open rate has become a misleading metric because Apple Mail Privacy Protection and similar blockers inflate it. Judge on reply rate instead: 5–15% total and 2–8% positive are the benchmarks worth managing against.
Do I have to warm up every domain?
Every domain, every time, with no shortcuts. Four to six weeks before full volume. Running an automated warmup tool alongside a handful of manual sends to real contacts in the first fortnight is the fastest safe path.
