Does cold email still work in 2026?

Yes, under conditions. Published reply rate benchmarks run from 0.45% to 4.5% because they count different things over different senders. Cold email works when the list fits the offer, the message asks one small question, the sender sticks to authentication and sensible volume, and follow ups add something new. It fails on bought lists and blasts.

The honest answer starts with an awkward fact: the benchmark reports disagree by a factor of ten. This post says what each one measured, so you can see why, and then lists the conditions our own email studies point to.

Why do cold email benchmarks disagree so much?

Because they measure different senders, different periods and, above all, different denominators. We opened each report on 19 October 2026.

Three things explain most of the gap.

• The denominator. Replies divided by sends gives a different number from replies divided by delivered mail, which differs again from replies divided by openers. Open tracking is itself unreliable because mail clients preload tracking pixels, a point the lemlist article makes.
• What counts as a reply. Some reports count any response, including an auto reply or a no thanks. Belkins excludes auto replies and bounce notices.
• Who is in the sample. A report built on one sending platform reflects the people who choose that platform and often send high volume with automation. A report built on agency campaigns reflects the agency's clients and lists.

Treat every benchmark as a rough range for that sender's mix. A cold reply rate of 3% by one definition and 0.4% by another can describe the same campaign.

Is the trend down?

Probably, but the data is not clean. Woodpecker says the platform average fell from 5.1% in 2024 to 3.43% in 2026. Belkins' older study of 16.5 million emails (2024 data) reported 5.8%, and its 2025 report reports 0.45%, but it changed the denominator between the two, so the drop is mostly a change of ruler. Be careful with any chart that puts the two on one axis.

What the reports do agree on is the shape. Smaller lists reply better than larger ones (5.8% under 50 contacts against 2.1% above 1,000 in the Woodpecker roundup, drawing on Belkins). Founders and owners reply more than vice presidents (0.57% against 0.32% in Belkins). Companies with 0 to 10 employees reply more than companies with 10,000 or more (0.72% against 0.22%). The pattern says that audience and fit move the number more than any copy trick.

When does cold email work?

Our reading, from the reports above and from the email studies we ran:

When these conditions are missing, cold email fails in a predictable way: a bought list, a pitch in the first line, three more bumps and a rising complaint rate.

What do our own email studies show?

We have three email studies on the site. None is a controlled test, and their bases are different, so read each on its own terms.

Shelly installer programme. A Sellinger agent emailed 3,940 installers of a smart home device maker in 111 countries, asking four questions in the installer's language. This was a listening programme to people who already hold a portal account with the manufacturer, not a cold sales campaign. In the June to August waves, 13.6% replied. With the first email alone the rate would have been 3.3%. 76% of converted installers replied only after a follow up. Touches 4 and 5 produced the only two stop complaints, and Shelly capped the sequence at the first email plus two follow ups. See the Shelly case study.

Green Shield Logistics. A carrier for medical and pharmaceutical waste in Bulgaria wrote as the founder, in formal Bulgarian, to 546 organizations from 21 September to 9 October 2026. Named heads of quality, ESG, procurement and logistics replied at 8.2%, almost four times the 2.2% of generic clinic and lab inboxes, and produced 17 of the 18 positive replies. Three companies reached a concrete next step. This is a short window and a small base. See the Green Shield case study.

Hotel Pirin. The hotel is a related party. Its agent reached 5,499 B2B contacts by email from 22 July to 11 October 2026 with short Bulgarian letters that ask for dates and headcount. The study says the agent's letters opened four of the five booking conversations it traces, and reports that reply rate and reservations did not move together: event agencies replied at 17.1% and produced no reservation. See the Hotel Pirin case study.

Two things follow from these. First, the reply rate in a warm, relevant audience can sit well above the benchmarks. Second, a reply rate is not a result. The hotel's event agencies replied most and booked nothing.

What are the risks?

• Complaints. Google tells senders to keep the spam rate below 0.30% and to aim for 0.10%. Cold mail is most at risk of crossing it.
• Refusals pile up. In the studies behind our follow up post, the share of negative first replies rose with each extra touch.
• The law. Cold email is lawful in some places on some terms and needs consent in others. See is cold email legal.
• Over reading benchmarks. If you hold your reply rate against a benchmark with a different denominator, you will conclude the wrong thing.

How should you decide?

Run a small test before a large one. Pick 100 to 300 people who fit the offer and have a named role. Send a short first email with one question and at most two follow ups that add something. Count replies the way a benchmark you trust counts them, and also count the outcome you care about: a meeting, a referral, a usable answer. If you cannot get a reply from the best 100, the offer or the list is the problem, and more volume will not fix it.

If your buyers are easier to reach on LinkedIn, compare the channels first. We wrote that up in cold email vs LinkedIn outreach.

Sellinger's email agent writes and sends the emails, follows your sending limits and stops when someone opts out. It does not choose your list for you, and it does not promise a rate. We publish the studies above with their bases so you can judge for yourself.