A 0.45% response rate and a 5% response rate can describe the same campaign. Belkins shows this in its 2026 study of 2025 sends: the first is calculated on every message sent, the second only on recipients who opened. Before rewriting a cold email sequence, check which denominator you’re looking at.
The diagnosis follows a fixed order. First measure the gap against benchmarks, then check that messages arrive and that the list holds up. Targeting and copy come next, follow-ups last: copy can do nothing for a message that landed in the spam folder.
A 0.45% or 3.43% response rate: which measure should you use?
Both figures coexist because the populations differ. Belkins measures its clients’ campaigns, relative to all messages sent; Instantly, which makes a sending tool, publishes in its 2026 benchmark the average across its users’ workspaces.
Belkins analyzed 7.5 million cold emails sent in 2025 and found a 0.45% response rate on total sends. The average rate fell from 0.50% in the first half to 0.40% in the second. The size of the targeted company weighs heavily: 0.72% for companies with 0 to 10 employees, 0.22% from 10,000 upward. By role, founders reply at 0.57%, versus 0.42% for C-level and 0.32% for VPs (Belkins study on response rates).
Instantly’s average reaches 3.43%, with top performers above 10%. A gap this size probably comes partly from who is measured and how replies are counted. Putting the two figures side by side as they stand misleads.
For your own sequence, measuring on total sends is the hardest to game. The open rate depends on a tracking pixel that some email clients block or trigger automatically, so the number of openers is unreliable and a rate calculated on opens is hard to compare. Also separate positive replies from the rest (polite refusals, auto-replies), then count booked meetings on their own.
None of these benchmarks accounts for your offer or the maturity of your market. They flag a rough drop-off and nothing finer: a target set to the tenth of a point asks too much of them.
Deliverability: the first check before touching the copy
A sequence with zero replies may not have been read by anyone. Start with the bounce rate and the spam rate, then move to authentication of the sending domain.
Instantly puts the ideal bounce rate below 2%. Above that, examine the list and your bounce rate management before touching anything else: repeated bounces look like those of a sender who doesn’t know their recipients. Soft bounces (full mailbox, server unavailable) call for a different response than hard bounces, which point to a nonexistent address.
For spam, Google sets the benchmark in its sender guidelines: a complaint rate below 0.10% in Postmaster Tools, never reaching 0.30%. In its FAQ, Google specifies that bulk senders lose access to mitigation measures as long as this rate exceeds 0.3% (Google’s sender guidelines). A lone salesperson’s volumes often can’t feed Postmaster Tools; in that case, the absence of data proves nothing.
Authentication comes next. Since February 1, 2024, all senders to Gmail must at minimum configure SPF or DKIM and keep their spam rate below 0.3%. Above 5,000 messages per day, Google also requires DMARC and domain alignment, plus one-click unsubscribe. A smaller cold email program doesn’t reach this second threshold, but authentication is still the first technical check. The protocol details are covered in this guide to SPF, DKIM, DMARC and BIMI.
That leaves the sending domain decision. If you prospect from your main domain, the company’s mailbox absorbs the reputation the sequence builds; a secondary domain limits the exposure of the main one. It needs a warm-up with a gradual ramp-up and a sending cap per mailbox, without which the domain’s reputation can degrade fast. You monitor it through the sender score; this article on deliverability lists the classic mistakes.
A placement test with seed mailboxes (the seed testing offered by tools like GlockApps) finally shows where the message lands: the inbox or the spam folder.
The contact list, the first link to break
An old or purchased list hurts results before the first sentence of the message is read. It inflates bounces and attracts complaints that damage the sending domain’s reputation.
Contacts change jobs regularly, and their work mailbox disappears with them. Databases pulled from trade shows or directories age the same way. Purchased lists mix outdated addresses with people who have never heard of you. Then there are catch-all domains, whose server accepts any address on receipt: the send goes through with no visible error, and nothing guarantees a human reads it.
Verifying addresses before sending changes the outcome. Most bounces are recorded at send time or shortly after, so filtering afterward only records the damage on a reputation that’s already hit. Address verification means checking that each one exists and won’t bounce. Data hygiene practices extend this check to duplicates and typos.
Before the next sequence, pull a sample from the list and run it through verification; the bounce rate that comes back tells you whether to clean the rest before sending.
Target and message: what the recipient sees
An inventory management solution pitched to a marketing director has little chance of succeeding, however good the copy. Targeting comes before writing.
A well-defined ICP shortens the list instead of lengthening it: industry, company size, decision-maker role, trigger event. The Belkins figures cited above (from 0.22% to 0.72% depending on the size of the targeted company) show how much this matters. You can then vary the angle by group rather than writing one message for everyone.
Personalization works when it rests on a verifiable fact: a recent hire or the tool the company already uses. A first name inserted in the opening line proves nothing. Frame the value proposition from the prospect’s side: say which specific problem gets simpler, and back it with proof.
The subject line remains the first barrier; the rules for an eye-catching subject line mostly help you avoid vague titles like “Follow-up” or “Proposal”. On length, according to Instantly’s benchmark, the best-performing campaigns in its sample run under 80 words. A message that mixes a company introduction with a meeting request no longer says what you expect from the reader; favor a single closed question, with one single call-to-action.
Sequence and follow-ups: what the numbers say
Instantly attributes 58% of replies to step 1 and 42% to follow-ups combined.
If you remove follow-ups to avoid “harassing” people, the sequence loses a large share of its replies. Instantly puts the ideal sequence length at 4 to 7 touchpoints: below that you give up too early, above it returns diminish unless each message adds something new. None of the cited studies sets the delay between two sends; you test that setting on your own target, one variable at a time. An A/B test that changes several variables at once teaches nothing.
A follow-up that repeats the first message word for word irritates without informing; it works better when it adds something new, such as a case study or a narrower question. One last mechanical point: the sequence must stop as soon as a contact replies or their address bounces. A follow-up sent to someone who has already said no generates spam complaints, which weigh directly on the rate monitored in Postmaster Tools.
B2B prospecting and the CNIL: what compliance changes
B2B email prospecting remains possible in France. The CNIL does not systematically require prior consent for prospecting between professionals, provided the message relates to the contacted person’s professional activity.
Recipients must be told where their data comes from and what the message is for. They must also be able to object easily to further solicitations, so each message includes a simple way to do so. The information provided at collection also covers the identity of the company (CNIL rules on electronic communication). An easy unsubscribe therefore protects two things: your GDPR position and your complaint rate. A contact who can’t find the unsubscribe link may click “report as spam”.
Which symptom points to which cause?
A single indicator rarely points to one culprit: each symptom below has its most likely cause and its check.
| Observed symptom | Likely cause | Check to run |
|---|---|---|
| Bounce rate above the 2% benchmark | Invalid addresses or old list | Verify a sample of the list and read the bounce error codes |
| Spam rate at 0.10% or more in Postmaster Tools | Complaints from poorly targeted recipients or volume ramped up too fast | Review the target, reduce volume per mailbox, check the unsubscribe |
| Sudden drop with no copy change | Failing authentication or damaged domain reputation | Check SPF, DKIM and DMARC, then test placement |
| Few replies from step 1 with a low bounce rate | Mismatched target or value proposition | Tighten the ICP and test another angle |
| Decent opens but no replies | Vague call-to-action or opens skewed by the tracking pixel | Count replies on total sends and cut down to a single question |
| Follow-ups bring nothing | Follow-ups copied from step 1 or no stop on reply | Add something new to each follow-up and wire in the stop rule |
A six-step diagnostic plan
Each step rules out one cause before moving to the next, in the order of the previous sections.
- Calculate the response rate on total sends of the last sequence; separate positive replies from the others.
- Record the bounce rate and, if volumes allow, the spam rate in Postmaster Tools.
- Check SPF, DKIM and DMARC, then the sending domain, the warm-up and the per-mailbox cap.
- Verify a sample of the list and measure the bounce gap before and after cleaning.
- Test only one variable at a time: the target first, then the message, then the follow-ups.
- Check the unsubscribe link and the sender identification before relaunching the sequence.
Frequently asked questions
What is a good cold email response rate?
Belkins measures 0.45% on all sends; Instantly reports a 3.43% average and more than 10% for top performers. Always compare your rate to a measure calculated the same way.
How many follow-ups should you plan?
According to Instantly, a sequence of 4 to 7 touchpoints is the right balance: below that you give up too early, above it returns diminish. Follow-ups account for 42% of replies. You find the right delay between sends by testing, with a stop rule as soon as the first reply comes in.
Is B2B cold email legal in France?
Yes, under conditions: a message related to the recipient’s professional activity, information on the source of their data and the purpose of the message, then a simple way to decline further ones. The B2B cold email guide details how to put it into practice.
Check the bounce rate of your last sequence today: if it clearly exceeds Instantly’s 2% benchmark, start by cleaning the list.
