The 50-90 word framework, opener types that work, subject line rules, and follow-up sequences that turn cold prospects into conversations.
Rees Bayba
Founder, Astra GTM
TL;DR
Most cold emails fail because they are too long, too vague, and too self-centered. The prospect does not care about your company, your features, or your founding story. They care about their problems. A good cold email is 50-90 words that make the prospect think 'this person understands my situation' -- and then asks one simple question. This guide covers exactly how to write that email.
The 50-90 word constraint is not arbitrary. It forces three things that make cold email work: one idea (not three), one ask (not a menu of options), and no filler (every word earns its place). Most cold emails are 150-250 words. The sender tries to explain their entire product, address every possible objection, and close for a meeting -- all in one email. The result reads like a brochure, not a conversation.
This is the benchmark. Below 2% means your targeting is off, your copy is weak, or both. Above 5% means your signal-based targeting and personalization are working. Above 10% is exceptional and usually indicates strong timing signals.
A 50-90 word email takes 15-20 seconds to read on a phone screen. That is all the time you get. The prospect decides in those seconds whether to reply, ignore, or mark as spam. Respect their time by being concise.
Every cold email that gets replies follows the same structure, whether the sender knows it or not. Four parts, four sentences, one purpose: start a conversation.
The opener determines whether the prospect reads the rest of your email. Three opener types consistently produce replies. Pick the one that fits your data -- if you have specific signals about the prospect, use a signal-based opener. If you do not, use a situation observation.
State something true about their industry or role that leads naturally to your product. This works when you do not have specific data about the company.
"Most [industry] companies grow through referrals until they hit a ceiling." "Hiring 3 SDRs costs the same as building an automated outbound system -- but one scales." "B2B teams that rely on inbound eventually run out of companies that already know about them."
Reference something you observed about their company -- a job posting, a funding round, a leadership change. This shows you did your homework and gives the email immediate relevance.
"Noticed you're hiring 3 SDRs -- usually means outbound is becoming a priority." "Saw the Series B announcement. Teams at your stage typically need pipeline faster than inbound can deliver." "Your VP Sales started 6 weeks ago -- new leaders usually audit the pipeline engine in their first 90 days."
Say something unexpected that breaks the pattern of sales emails they receive daily. Use sparingly -- it works well but can feel gimmicky if overdone.
"This isn't a pitch." "I'm not going to pretend we have a mutual connection." "You probably get 10 of these a day. Here's why this one is different."
The difference between a cold email that gets replies and one that gets deleted is usually 10 specific mistakes. Here is what a bad email looks like next to a good one.
Don't do this
Hi John, I hope this email finds you well! My name is Sarah and I'm the VP of Sales at DataFlow. We're an innovative AI-powered data integration platform that helps companies like yours streamline their data pipelines, reduce manual ETL work, and accelerate time-to-insight by up to 3x. We work with leading companies like Stripe, Notion, and Datadog to solve complex data challenges. Our platform features include: - Real-time data sync - No-code pipeline builder - SOC 2 compliance built in - 200+ native integrations I'd love to schedule a 30-minute call to walk you through how we've helped similar companies reduce their data engineering overhead by 60%. Would Tuesday or Thursday work for you? Best regards, Sarah
Do this instead
John -- noticed you posted a Senior Data Engineer role last week. Teams hiring for that role are usually buried in pipeline maintenance that pulls engineers off product work. We automate the ETL layer so data teams spend time on analysis instead of plumbing. Cut DataCorp's pipeline maintenance by 60%. Worth a conversation? Sarah
The bad email is 152 words of self-promotion. It starts with the sender's name and company, lists features nobody asked about, name-drops clients, and asks for 30 minutes. The good email is 62 words. It opens with something specific to John, names his problem, states the solution in one sentence, provides one proof point, and asks a question that takes two seconds to answer.
Your subject line has one job: get the email opened. It is not a headline, a value prop, or a summary. It should look like something a colleague or business contact would send -- not something a marketing team wrote.
Good: "quick question" | "re: pipeline" | "thought on this" | "{first_name} -- one idea" | "data team" Bad: "Revolutionize Your Data Pipeline with DataFlow AI!" | "Exclusive Offer Inside" | "Meeting Request: 30-Min Demo" | "Quick Question About Your Company's Data Strategy and How We Can Help"
Most replies come from follow-ups, not the first email. A 4-step sequence over 14 days is the standard structure. Each follow-up should add new information or a new angle -- never just 'bumping this to the top of your inbox.'
| Step | Timing | Format | What it does |
|---|---|---|---|
| Step 1 | Day 1 | New thread | The core email: opener + problem + bridge + CTA |
| Step 2 | Day 3-4 | Thread reply | New angle -- a different way to frame the problem, or a specific result you drove for a similar company |
| Step 3 | Day 7 | Thread reply | Social proof -- a specific case study, metric, or customer quote that builds credibility |
| Step 4 | Day 14 | Thread reply | Breakup -- let them know this is the last email. Give them a reason to reply now if they are interested. |
Certain patterns kill reply rates immediately. If you see any of these in your emails, rewrite.
Guessing what works is expensive. Testing tells you. But testing requires discipline -- most teams test wrong and draw false conclusions from tiny samples.
| What to test | Impact level | Minimum sample size | What to measure |
|---|---|---|---|
| Subject lines | High | 200 sends per variant | Open rate |
| Opener type | High | 200 sends per variant | Reply rate |
| CTA phrasing | Medium | 200 sends per variant | Reply rate |
| Email length | Medium | 300 sends per variant | Reply rate |
| Send time | Low | 500 sends per variant | Open rate |
| Sender name | Low | 300 sends per variant | Open rate |
How long should a cold email be?
50-90 words. This is enough for an opener, a one-sentence problem statement, a one-sentence bridge, and a CTA. Emails under 50 words feel abrupt. Emails over 100 words get skimmed. The sweet spot forces you to cut everything that does not directly contribute to getting a reply.
How many follow-ups should I send?
Four total emails (1 initial + 3 follow-ups) over 14 days is the standard sequence. Most replies come from Steps 2 and 3, not Step 1. After Step 4, stop. Sending 7-8 follow-ups does not increase replies -- it increases spam complaints and unsubscribes.
Should I personalize every email?
Yes, but personalization does not mean spending 15 minutes per email. The opener should reference something specific -- a job posting, a funding round, a company detail. That takes 30-60 seconds with the right data. The rest of the email can be templated. Signal-based personalization in the first sentence is what matters. Do not rewrite the entire email for every prospect.
What reply rate should I expect?
3-5% for well-written cold email to a well-targeted ICP. Below 2% means your copy, your targeting, or both need work. Above 5% means your signals and personalization are strong. Above 10% is exceptional and usually comes from strong timing signals (hiring, funding, leadership change) combined with tight targeting.
Should I use HTML formatting or plain text?
Plain text. Cold emails should look like a real email from a real person -- not a marketing newsletter. HTML formatting, images, logos, and fancy signatures increase spam scoring and signal 'mass email' to the recipient. Use plain text with a simple signature (name, title, company, phone number).
What time of day should I send cold emails?
Tuesday through Thursday, 8-10 AM in the recipient's local timezone is the most common recommendation. But send time has less impact than copy quality and targeting. If your email is relevant and well-written, it will get replies whether it arrives at 8 AM or 2 PM. Test send times only after you have optimized subject lines, openers, and CTAs.
We implement these systems end-to-end. First sends within 14 days.