Does Cold Email Work? How to Judge Replies, Meetings, and Revenue

Yes. Cold email generated replies and booked meetings in the provider datasets below. Whether it wins customers profitably for your offer depends on reaching the right buyer, making a credible case, getting the message delivered, and earning enough qualified business to justify the cost. An open or a generic reply cannot answer that question by itself.

In brief: Published cold email benchmarks show real replies, but their rates use different denominators and provider-selected populations. Measure your own campaign from unique prospects through positive replies, held meetings, qualified opportunities, and customers. Use a consistent cohort and time window. If delivery, buyer fit, or offer evidence is missing, fix that before scaling sends.

Does Cold Email Still Work for B2B Sales?

It works in the limited, observable sense that cold outreach generates replies and meetings in some B2B programs. Belkins' analysis of its 2025 client campaigns reports 34,393 unique replies from 7,530,489 cold emails. Gong's analysis of more than 28 million cold emails reports an average of 344 emails sent per booked meeting in its sample. Both are provider reports about their own data. Neither tells you the probability that a particular new offer will win a customer.

The useful decision is narrower: Can this team reach a defined buyer group with a credible offer, produce qualified conversations at an acceptable cost, and repeat the result? A campaign can collect replies and still fail that test if the replies are objections, the meetings never happen, or the opportunities have no fit.

Cold email also competes with other ways to reach the same buyers. If your audience is small and well known, direct introductions, partner referrals, events, or account-specific outreach may warrant a comparison. The channel should earn its place in your acquisition mix through actual outcomes, not a benchmark headline.

How Often Do Cold Emails Get Replies?

There is no single rate that honestly answers this for every B2B team. Before using a benchmark, ask who was counted, what counted as a reply, and what the denominator was.

Belkins: unique replies divided by emails sent. Its 2026 report publishes 0.45% overall and 0.51% for its US slice of more than 5.6 million sends in 2025 client campaigns. The numerator is unique replies; the denominator is emails sent. Its displayed totals of 34,393 and 7,530,489 calculate to about 0.46% under conventional rounding, a small discrepancy in the published report. The report does not establish a new campaign's positive-reply, held-meeting, or customer rate.

Instantly: all replies divided by total sends. Its 2026 platform report gives 3.43% overall across active workspaces for January 1–December 18, 2025, counting follow-up responses. This is also a per-send denominator, but it does not attribute every reply to one particular email, and its users, campaign mix, and numerator differ from Belkins'. The two numbers cannot be treated as competing estimates from one representative population.

Hunter: recipients replying in a sequence. Its 2026 report gives 4.5% overall and 3% for sales outreach, expressed as recipients who replied during a sequence divided by recipients in Hunter user campaigns. A sequence may contain multiple sends, so neither rate is a per-email rate.

Gong: emails per booked meeting. Its analysis reports 344 emails sent per booked meeting for the average rep in its sample. Its public article does not establish whether those meetings were held, qualified, or led to revenue.

Do not average these figures. The units, provider-selected users, campaign mixes and outcomes differ. Even Belkins and Instantly, which both use emails sent as the denominator, have materially different rates in different populations. Belkins also changed its denominator from replies per unique opener in earlier reports to replies per total send in its 2026 report, so a year-over-year comparison of those published rates would be misleading.

The honest conclusion is that cold email can earn replies, while a universal “good reply rate” remains unsupported by these sources. For planning, use your own segment and outcome data; treat outside figures as examples of measurement, not targets you are owed.

What Counts as a Successful Cold Email Campaign?

Start with the business outcome you can observe and work backward. Record each stage separately:

1. Unique prospects enrolled: Distinct people who entered the test. This is the denominator for per-person sequence results.
2. Emails sent and delivery problems: Individual messages attempted, bounces, and obvious blocks. A send is an activity count; it is not proof of inbox placement or reading.
3. Human replies: Deduplicated people who responded. Exclude automatic replies and bounce notices. Keep objections and opt-outs visible, but separate them from interest.
4. Positive replies: People who express relevant interest or agree to a useful next step under a definition set before the test. A “not now” may be valuable context, but it is not a positive reply by default. Qualification of an opportunity comes later under the team's sales rules.
5. Meetings booked and held: Count each once. A booked call that is canceled or never attended is not a held meeting.
6. Qualified opportunities and customers: Apply the team's normal qualification rules. Attribute revenue only after a documented customer outcome and a reasonable attribution window.

Report the numerator and denominator together: for example, “15 distinct responding prospects across 1,500 sends” and “15 responding prospects from 1,000 enrolled prospects.” Both can be true. They answer different questions.

Open rates are a weak success measure. Mailchimp explains that Apple Mail may preload tracking pixels without a human reading the email; its open-tracking guide notes that blocked images and security software can also distort counts. Opens can help investigate a technical change in a consistent setup, but they cannot show buyer interest or revenue.

A Worked Example: One Campaign, Several Honest Rates

Illustrative data, not an Alore result or industry benchmark: A fictional team offering a scoped workflow audit enrolls 1,000 verified prospects in a defined US B2B segment. Its planned sequence sends 1,500 individual emails. At a stated checkpoint after the sequence and response window, 15 distinct people have responded, including six people with positive replies. Three meetings have been booked; two have been held, and the third remains pending. Of the two held meetings, one becomes a qualified opportunity. No customer has closed yet.

Before sending, this team defined a qualified opportunity as a held meeting with an in-scope workflow owner who confirms an active review within 90 days and agrees to examine a scoped audit proposal. In the first held meeting, the in-scope workflow owner confirms that project and agrees to the evaluation step. In the second, the buyer has no active review and declines an evaluation. The second meeting still happened; it does not meet this fictional team's opportunity rule. These criteria illustrate a reproducible decision, not a universal sales definition. The three other positive respondents have not booked a meeting by this checkpoint; their later outcomes remain open.

• Unique respondents per 100 sends: 15 responding people ÷ 1,500 emails × 100 = 1.0. This is an activity-normalized yield, not proof that exactly 15 particular emails received replies.
• Sequence reply rate per prospect: 15 responding people ÷ 1,000 enrolled people = 1.5%.
• Positive reply rate per prospect: 6 interested people ÷ 1,000 enrolled people = 0.6%.
• Held meeting rate per prospect: 2 held meetings ÷ 1,000 enrolled people = 0.2%, assuming one meeting per person.
• Qualified held meeting rate per prospect: 1 qualified held meeting ÷ 1,000 enrolled people = 0.1%.

These numbers describe one campaign, yet a chart that shows only “1.5% reply rate” hides send volume, reply quality, and meeting outcomes. The one qualified opportunity is encouraging, but it does not prove profitable acquisition.

Assumptions for this fictional cohort: The campaign has cost $2,000 so far, including list research, writing, tools, sending, reply handling, and sales time already spent through qualification. A won customer would contribute $10,000 after service delivery and all remaining sales costs.

Calculation at this checkpoint: One qualified held meeting has emerged, so cost per qualified held meeting to date is $2,000 ÷ 1 = $2,000. If that one opportunity remains the cohort's final qualified yield and no extra cost is incurred, it needs an eventual win probability of at least 20% to break even under these assumptions: $2,000 ÷ $10,000 = 0.20.

Interpretation: This is an expected break-even condition, not an observed win rate or ROI. The third booked meeting and the three other positive respondents may still produce outcomes. If another qualified opportunity appears or the team incurs more cost, recalculate using the final qualified count and total cost. One unclosed opportunity cannot establish the actual probability of winning. The diagram shows the current checkpoint, not a finished customer funnel.

For a real campaign, use your own full costs, qualification rules, customer contribution, attribution window and observed conversion history. Avoid double-counting sales time already included in campaign cost when calculating contribution after the meeting. A high-contribution offer may justify fewer meetings; a low-margin offer may not.

What should this fictional team do now? Limit the next cohort while the open meeting and opportunity mature. Then use the relevant evidence:

• Comparable win history exists: Compare its qualified-meeting-to-customer win rate with the recomputed break-even condition. A small controlled continuation may be justified if the economics hold.
• No comparable history exists: Set a spending cap and a review date tied to the sales cycle. Collect more qualified outcomes before expanding.
• Mature cohorts repeatedly fail: If they fall below the team's economic threshold and no specific buyer, offer, or delivery problem remains to test, stop this campaign.

The fictional numbers teach the decision; they are not market targets.

Why Results Vary So Much

The same copy can produce different results when the buyers, offer, evidence, or delivery conditions change. Separate these factors before rewriting a subject line:

• Buyer fit: Is this person a plausible owner of the problem and able to use the proposed next step? A large list of loosely matched titles inflates activity without improving relevance.
• Reason to contact: Is there a verified, current reason this offer matters to this buyer? A public hiring post can establish context, but it cannot prove a hidden pain or budget.
• Offer and proof: Can the sender explain a bounded result and support any claimed outcome? Unsupported percentages or invented customer stories erode the case.
• Delivery and list quality: Invalid addresses, missing authentication, or spam complaints can prevent evaluation of the offer at all. Google's sender guidelines describe requirements for mail to personal Gmail accounts; meeting those rules does not guarantee placement in a corporate inbox.
• Timing and competing priorities: A relevant buyer may have no current project. Silence is compatible with several explanations. It is not proof that the buyer read and rejected the offer.

For the message itself, use Alore's guide to writing a cold email. For later touches, use the cold email follow-up sequence guide. This article's job is to judge the channel and the test, not to prescribe universal copy or send days.

How to Test Cold Email Without Fooling Yourself

Write the test plan before the first send. It can fit on one page:

1. Define the cohort and offer. Choose one buyer segment, one problem hypothesis, one offer, and one small next step. Save the inclusion rules and the date the contact evidence was checked.
2. Set the observation window. State when enrollment ends and when replies, held meetings, and opportunities will be counted. Allow the planned sequence and reasonable reply lag to finish before judging it. The right window depends on the sales cycle; a fixed number of days would be arbitrary.
3. Specify the outcome definitions. Decide what counts as a human reply, positive reply, held meeting, and qualified opportunity. Tag objections and opt-outs separately. Deduplicate people who reply twice.
4. Verify sending conditions. Check addresses and authentication, monitor bounces and complaints, and test the actual unsubscribe path. For US commercial email, the FTC says CAN-SPAM has no B2B exception. Use truthful headers and subjects, identify the message as an ad, provide a valid postal address and clear opt-out, keep the opt-out mechanism able to process requests for at least 30 days after sending, and honor opt-outs within 10 business days. Apply the full FTC checklist; provider rules and other applicable laws may add obligations.
5. Run a controlled comparison where practical. Keep the buyer segment and offer stable while changing one consequential element, such as the reason to contact or the offered next step. If several things change at once, describe the result as a new campaign, not proof that one sentence caused it.
6. Review the whole funnel and cost. Examine the counts, denominators, objections, held meetings and sales outcomes together. Preserve the original cohort, including nonresponders, when reporting rates.

Do not set a universal pass/fail percentage from a vendor report. Decide in advance what minimum commercial result would make the test worth repeating for your economics. A small first cohort may identify a broken assumption, but it may be too small to estimate a stable conversion rate. Treat an early zero or a lucky win as a reason to investigate, not a forecast.

When to Improve, Pause, or Stop

Use the earliest failing stage to choose the next action:

Many invalid addresses, bounces, or clear sending errors. The test may not have reached its intended buyers. Verify source quality, domain authentication, and sending logs before judging copy.

Sending logs show accepted messages, but almost no people reply. Buyer fit, relevance, offer, inbox placement, or timing may be weak. Acceptance by a mail server does not prove inbox placement. Check a sample of recipient roles and the evidence behind the offer; investigate deliverability separately. Change one major assumption in the next cohort.

Replies arrive, mostly objections or opt-outs. The audience or promise may be wrong, even if the reply rate looks good. Code the objections; remove poor-fit segments and narrow the claim or offer.

Positive replies arrive, but meetings do not happen. The proposed next step, handoff, calendar friction, or follow-through may be weak. Review response time, the actual reply conversation and cancellations before rewriting the original email.

Meetings happen, but few qualify or close. Cold email is creating conversations that may not match the economic buyer or offer. Inspect qualification, pricing fit, sales process, and margin before increasing send volume.

These are diagnostic possibilities, not automatic causal conclusions. Pause a cohort when list provenance, applicable consent rules, compliance, buyer fit, or sending setup is unresolved. Stop the test if repeated, well-measured cohorts fail the team's predeclared economic threshold and no credible, bounded change remains to test. If it succeeds, expand cautiously while watching whether quality holds as the audience broadens.

Common Questions About Cold Email Effectiveness

Is cold email dead?

No. The cited provider datasets document replies and meetings, although their samples do not predict your result. The more useful question is whether your particular buyers and offer produce enough qualified business at an acceptable cost.

What is a good cold email reply rate?

There is no universal number across per-send and per-recipient definitions, industries, lists, and offers. Compare campaigns only after aligning their units and outcome definitions. Positive replies and held, qualified meetings carry more business meaning than a blended human-reply rate.

How long should I test before deciding?

Long enough to complete the planned sends and response window and observe the downstream outcome you care about. Record the cohort, counts and timing before sending. A small cohort can expose a broken list or implausible offer; it usually cannot prove a precise future conversion rate.

Can cold email be legal in the US?

Yes, if applicable requirements are met. The FTC's CAN-SPAM guide says its rules apply to B2B commercial email as well as consumer email. A compliant message can still be ineffective or blocked by a provider, and other laws may apply to particular recipients or sending contexts.