You cannot improve what you do not measure, and you cannot trust what you measure badly. Cold email is full of numbers that look precise and mean very little. This chapter shows which numbers matter, which ones mislead, how to read small samples without fooling yourself, and how to run simple tests that actually teach you something.
Measure the outcome, not the activity
It is easy to measure activity: emails sent, sequences started, contacts added. Those numbers always go up, and they feel like progress. But nobody buys because you sent more email. The point of outreach is conversations with the right people, and then meetings and customers. Activity numbers only matter as inputs to those outcomes.
So start from the end. Decide what a good result looks like for this campaign: a booked call, a reply asking for the guide, a referral to the right person. Then work backwards to the few numbers that tell you whether you are getting there, and where things go wrong when you are not.
Delivered. Emails that reached a mailbox. Broken by bad data, missing authentication or a damaged sender reputation.
The numbers that matter
| Metric | What it tells you | How to read it |
|---|---|---|
| Bounce rate | Whether your list data is clean | Should stay very low. A jump means bad data or a broken source. Stop and clean before sending more. |
| Spam complaint rate | Whether readers feel your email is unwanted | Watch it in Google Postmaster Tools. Keep it well under Google's limit, covered below. |
| Reply rate | Whether the email earns a response at all | Count human replies only. Leave out auto-replies and bounces. |
| Positive reply rate | Whether the right people are interested | The best single measure of targeting plus message. |
| Unsubscribe rate | Whether you are irritating people | A rise at one step means that step is the problem. |
| Meetings booked | Whether replies turn into conversations | Track per campaign, not just in total. |
| Time to first answer | How fast you respond to interested replies | Same working day is the goal. Slow answers lose warm leads. |
Why open rates mislead
For years, open rate was the headline number in email. It is now one of the least reliable. Opens are counted by loading a tiny invisible image. Some mail apps now load images automatically for privacy, whether or not a person reads the email, which makes opens look higher than they are. Others block images, which makes opens look lower. Many security filters also open links and images to scan them. The number you see is a mix of people, machines and privacy features.
Tracking pixels and rewritten tracking links also add weight to cold email that can work against delivery. Many careful senders turn open tracking off for cold outreach and judge campaigns by replies instead. Replies are slower to collect, but a reply is a person, every time.
Classify every reply
A raw reply count hides the story. Ten replies that all say "please remove me" are worse than none. Sort every reply into a small set of labels, the same ones every time, so you can compare campaigns fairly.
- Interested. Wants to know more, asks a question, or agrees to a call.
- Not now. Fit is right, timing is wrong. Note the date they gave.
- Referral. Not the right person, but pointed you to who is.
- Not interested. A clear, polite no.
- Unsubscribe or negative. Wants no contact, or is annoyed.
- Automatic. Out of office, changed jobs, mailbox full. Not a human reply.
Referrals and "not now" replies are worth more than they look. A referral comes with a warm first line for the next email. A "not now" comes with a date. Count them, follow up on them, and do not lump them in with plain rejections.
Small numbers, big mistakes
Most cold campaigns are small. A few hundred emails might produce a handful of replies. At those sizes, luck plays a big part. One campaign can look twice as good as another just because a couple of extra people happened to be at their desks. Before you declare a winner, ask whether the difference would survive another week of sending.
A rule of thumb for small samples
Do not change course based on a difference of one or two replies. Wait until each version has been sent to enough similar people that the replies number in the dozens, not the handful. If you cannot reach that size, treat the result as a hint to test again, not a conclusion.
- Compare like with like. Same kind of contacts, same weeks, same mailboxes. A campaign to warm referrals will always beat one to a cold list.
- Change one thing at a time. If you change the subject line and the trigger and the ask together, you will not know which change mattered.
- Watch for timing effects. Holidays, month ends and big news days change reply patterns for everyone.
- Look at the replies themselves. Reading twenty replies teaches you more than any chart.
Run a simple test
Write down the question
One question per test. Example: does opening with the job ad trigger earn more interested replies than opening with the phone-booking problem?
Split the list fairly
Divide similar contacts at random into two groups. Do not put the best accounts in one group.
Change only the thing you are testing
Same subject, same ask, same sending times and mailboxes. Only the opening changes.
Decide the finish line first
Pick how many sends and how many days before you look. Peeking early tempts you to stop on a lucky streak.
Read and record the result
Note the replies by label, what you learned, and what you will test next. Keep this log; it becomes your playbook.
What to test first
Not every test is worth the time. Test the things with the biggest effect first. In cold email, that is almost always who you write to and why, long before the wording of a subject line.
| Order | Test | Why it matters |
|---|---|---|
| 1 | Audience and segment | The wrong people will not reply to the best email. |
| 2 | Trigger | A strong reason to write now beats any wording trick. |
| 3 | Offer and ask | A small, easy ask earns more replies than a big one. |
| 4 | Opening line | Decides whether the email gets read past the preview. |
| 5 | Subject line | Matters, but less than people think once the rest is right. |
| 6 | Sequence timing | Gaps and number of steps. Fine-tune last. |
Watch deliverability every week
A sudden drop in replies is often a delivery problem, not a writing problem. Before you rewrite a campaign that stopped working, check that it is still reaching inboxes. Look at bounces, spam complaints in Postmaster Tools, and any rejection messages. Send a test to a few seed mailboxes you control and see where it lands. Chapter 9 covers the full check.
0 of 6 checked. Bounces and rejections: Any new spike? Which list or source did it come from?
Why borrowed benchmarks mislead
It is tempting to compare your numbers with a reply rate you read in a blog post. Resist it. Those figures come from other markets, other lists, other offers and often other ways of counting. Some count auto-replies. Some only report their best campaigns. A number from a warm list of past customers says nothing about a cold list of strangers. Your most useful benchmark is your own history: the same segment, the same kind of offer, last month versus this month. Beat yourself, not a stranger's headline.
Read the replies, not just the counts
Numbers tell you where to look. Replies tell you what is wrong. Once a week, read every reply from your main campaign in one sitting, including the rejections. Patterns appear quickly. If several people say "we already use something for this", your trigger finds the right companies but too late. If people ask what you actually do, the email is too vague. If replies come from the wrong role, your list needs a different title filter. Write each pattern down next to the change you will make.
Collect a week of replies
All of them, from one campaign, in one place.
Group them by what people said
Not by label alone. Two "not interested" replies can have very different reasons.
Name the most common reason
The single pattern that shows up most often.
Pick one change that answers it
A new trigger, a clearer offer, a different role. Test that change next week.
Attribution: be honest about what you know
When someone becomes a customer months after a cold email, after a webinar, a referral and a visit to your pricing page, it is hard to say which touch did it. Do not pretend to more precision than you have. Record the first touch and the touch that led to the meeting, and ask new customers how they first heard of you. Their answers, collected over time, are usually more useful than a complicated model.
Doing this in FirstSales
FirstSales shows each campaign's deliveries, bounces, replies and unsubscribes in one place, and sorts replies so interested ones reach you first. Sequences stop when someone replies, so your numbers reflect real conversations, not follow-ups sent after an answer. You can explore campaign results with sample data in the demo.
Key takeaways
- Measure outcomes like interested replies and meetings, not just activity.
- Open rates are unreliable. Judge cold campaigns by replies.
- Label every reply the same way so campaigns can be compared fairly.
- Small samples lie. Change one thing at a time and wait for enough replies.
- Test audience and trigger before wording.
- Check deliverability weekly, and before you rewrite a campaign that stopped working.
Quick answers about the masterclass
How many cold emails should I send per day?
Start with 5-10 per inbox per day for new accounts. After 2-3 weeks of warm-up, scale to 35-50 per inbox. If you need to send 500+ emails per day, use 10-15 warmed inboxes and rotate sends across all of them. Never push a single inbox beyond 50 cold emails per day.
What is a good cold email reply rate?
There is no honest single number. Reply rates depend on your market, your list, your offer and how replies are counted, so figures from other people's campaigns rarely transfer. Compare each campaign with your own past results for the same kind of audience, and judge by interested replies rather than total replies. Chapter 12 shows how.
How long should a cold email be?
Keep cold emails between 50-125 words. Some studies suggest up to 150 words for certain B2B audiences. Shorter emails consistently outperform longer ones. If you need more than 125 words to make your point, your message is not focused enough.
Should I use HTML or plain text for cold emails?
Plain text. Always. HTML emails with logos, images, and formatted layouts trigger spam filters and look like marketing emails. A plain text email from one person to another is what cold email should look like. The only exception is your email signature, which can include basic formatting.
How many follow-ups should I send?
Usually a few, spaced several days apart, each adding something new: a fresh angle, a useful resource or a simple question. Stop as soon as someone replies, and end with a short, polite close. Chapter 8 covers timing and what each step should say.