Does Cold Calling Still Work? What 175,000 B2B Dials Actually Show

Every few months someone declares cold calling dead, and every few months a sales team books a pipeline's worth of meetings from a call list and quietly ignores the obituary.

Both camps are looking at real numbers. The numbers just describe a channel that behaves very differently depending on who is measuring, which list they are calling, and how many times they try. This article takes the largest recent B2B datasets we could verify and shows what they actually say: how often calls connect, how often a connect becomes a meeting, which days and months are worth your reps' time, which opening lines help and which hurt, and where the legal lines sit for human callers and for AI voice agents.

One caveat before the data. Every figure below is attributed to the source that measured it, with its sample size and its own caveats attached. Where a dataset is skewed toward certain industries or regions, we say so rather than smoothing it over, because averages from a mixed dataset are a poor predictor of what your list will do.

What 175,000 B2B dials actually show

Belkins combined two datasets: SDR activity on its Nooks dialer platform across calendar year 2025, plus five anonymized HubSpot campaign exports from 2025. Together that is more than 175,000 dials across the US, UK, Germany, Canada, Sweden, and other European markets. That is a large sample of real B2B calling, and Belkins is explicit about its limits: the US portion is heavily concentrated in healthcare and retail, two verticals with high call volume and lower conversion, while the UK and European records cover a broader mix.

Here is the funnel those dials produced.

Multiply those through and you get Belkins' headline figure.

That number is the one that fuels the "cold calling is dead" posts, and it deserves a straight reading. As an illustration only: a rep making 60 dials in a working day would need a little over six working days to produce one meeting at that rate. If your average deal does not justify that effort, no script will rescue the channel for you. If it does, a single booked meeting pays for a lot of dials.

Notice where the funnel actually leaks. Getting someone on the line is hard, with only about one dial in ten connecting. But the stage that looks best on paper is the one that matters least to tune: 58% of connects turn into a conversation, which means once a person picks up, most of them stay on the line long enough to hear you. The real cliff is the last step. Only 4.6% of conversations end in a meeting.

That is the number the rest of this article is about, because conversation quality is the part of the funnel your reps control directly.

Why persistence beats the single dial

The per-dial connect rate is 9.9%. The per-prospect connect rate is 24.5%.

These are two ways of counting the same activity, and the gap between them is the case for follow-up. A prospect who does not answer on a Tuesday morning is not a dead lead. They were in a meeting, on another call, or away from their desk. Reaching them takes a second and third attempt at a different time, and the gap between those two numbers is what those later attempts are worth.

The practical rule is to build the cadence before you dial. Decide how many attempts each prospect gets, at what intervals, at which times, and what other channel touches sit between them. A rep who dials a list once and calls it exhausted has measured the weak number and drawn the wrong conclusion.

This is also where cold calling stops being a standalone channel and starts looking like part of a sequence. A call that does not connect can be followed by a short email that references the attempt, then a LinkedIn touch, then another call. Our annotated outbound templates show what each of those touches can look like when the structure is built from data rather than guesswork.

Timing: which days and months deserve your reps' energy

Belkins measured the positive outcome rate by weekday, as a share of all dials.

Wednesday produced nearly double Friday's rate. By hour, 3 to 4pm was the strongest sustained window at roughly 1.05 to 1.22%, with a secondary peak of 1.47% at 6pm, while the high-volume 9 to 11am block converted at a middling 0.65 to 0.83%. Belkins is careful to note that the hour-of-day analysis comes from campaign-level data and reflects the calling patterns of specific client verticals. Read it as a hypothesis to test on your own list, not a law.

The seasonal pattern is more striking, and more useful, because it points at something counterintuitive.

The detail that stands out is July. It had the highest connect rate of the year at 12.5%, because people were reachable, and the lowest meeting rate at 2.7%. Prospects were picking up, and they were not booking. Presence is not the same as availability, and a summer connect is worth far less than a February one.

If you run outbound, that argues for front-loading your hardest calling into the first half of the year and the autumn, and using the summer for list building, research, and email and LinkedIn touches that do not depend on someone being at their desk and receptive in real time.

What happens on the calls that do not book

Most calls do not end in a meeting, and what they end in is informative. Belkins broke down the non-meeting outcomes.

Two groups here are fixable and one is not.

The hard rejections, 31%, are the part you influence through who you call and how you open. The "follow-up requested" group, 19%, is a warm lead that most teams under-work. A prospect who says "send me something" is telling you which channel to use next, and a rep who logs it and never follows up has thrown away a meeting that was half booked. And 26.9% of non-meeting outcomes, the wrong person plus the gatekeeper, are not about your pitch at all. They are about the quality of the list and whether the contact data was right in the first place.

That last point is why list quality sits upstream of everything else in this article. The way we think about building a list around real buying signals matters here because a call to the right person at the right moment starts from a different place than a call to a name on a purchased spreadsheet.

Two more patterns from the same dataset are worth knowing. Calls that booked a meeting averaged 3.7 minutes, about four times the roughly 0.9 minutes of a polite "not interested," and no-answer calls ran under 0.3 minutes. Length is a result of the conversation going well, not a cause of it, but it tells you what a working call looks like: a prospect who stays on the line. And at the C-level, Belkins found that every C-suite positive outcome in the dataset was a follow-up requested by phone or email, not a meeting booked in real time. If you call senior buyers expecting a calendar slot on the first conversation, the data suggests you are measuring against the wrong target.

The opener decides more than the pitch

Gong's analysis of 90,380 connected cold calls remains the most-cited dataset on what happens in the first seconds. Gong defines a successful call as one that ended in a held follow-up meeting, and found the baseline success rate sat at about 1.5%.

Against that baseline, the opening line moved the number a lot.

The pattern is the same one we saw in the templates data: prospects arrive at a cold call expecting a script, and a line that sounds like it came from someone who already knows them breaks the script just long enough to keep the call alive. "Did I catch you at a bad time?" does the opposite. It tells the prospect you expect to be an interruption and gives them a polite way out of the call before it starts.

A caution on the age of this evidence. Gong's article was first published in 2018 and has been updated since, and the success rates it reports come from a database of more than a million recorded sales calls analyzed with natural language processing. Treat the absolute percentages as indicative and the relative ordering as the useful part. Test your own opener against the same four lines before rewriting everything around the winner.

Gong also looked at how successful and unsuccessful calls differ in structure. Gong's own answer to whether cold calling is dead is no, with a catch: results are wildly unequal, with most reps below 1% and a minority succeeding at a far higher rate. Successful calls ran almost twice as long as unsuccessful ones, the longest rep monologue on a successful call averaged 37 seconds against 26 seconds on an unsuccessful one, and the prospect's own monologue averaged only 3.5 seconds on successful calls. The number of questions asked made no statistical difference. Reps who win are not interrogating prospects. They are saying something worth listening to, for long enough to matter.

UK versus US: what the regional split does and does not mean

Belkins also split live-contact and positive-outcome rates by region.

It is tempting to read this as "the UK is easier," and Belkins itself cautions against that: vertical composition explains a significant share of the difference. A dataset where the US portion is mostly healthcare and retail will always look worse than one with a broader mix, regardless of country. The useful takeaway is narrower. Your own list's industry and role mix will drive your results more than the national average, so benchmark against your own vertical before you decide the channel is working or failing.

The rules: what is allowed when you call businesses

Cold calling sits inside a legal framework that most sales content ignores until a complaint arrives. This is a plain-language summary, not legal advice, and the rules vary by country, state, and the type of business you call.

United States, human callers. The FTC states that most phone calls between a telemarketer and a business are exempt from the Telemarketing Sales Rule. The significant exception is calls to induce the retail sale of nondurable office or cleaning supplies, which are not exempt and must comply with the rule. The National Do Not Call Registry is aimed at consumers, so it is not the main constraint on a typical B2B prospecting call. State laws can add their own requirements, and the call content still has to be truthful.

United Kingdom. Under the Privacy and Electronic Communications Regulations, the ICO says B2B callers must screen against both the TPS register, covering sole traders and some partnerships, and the CTPS register, covering companies, some partnerships, and government bodies, as well as their own do-not-call list. You must let the person you are calling see your number, or an alternative contact number, and you cannot call anyone who has told you they do not want marketing calls from you. Sole traders and partnerships being on the TPS means a call to a small firm can be covered even when you think of it as "B2B."

AI voices change the calculation. On February 8, 2024, the FCC issued a unanimous Declaratory Ruling confirming that AI-generated voices count as "artificial" under the Telephone Consumer Protection Act. The consequence is that calls using them require prior express consent, or prior express written consent for telemarketing, from the person called. The ruling covers calls to both wireless and residential landline numbers, and a large share of B2B prospects are reached on mobiles. TCPA violations carry statutory damages of $500 per violation, which a court can increase to as much as $1,500 for willful or knowing ones. A rule that looks irrelevant to B2B in the abstract becomes very relevant the moment an autodialer starts placing AI-voiced calls to prospects' cell phones.

The legal position on a given campaign depends on the numbers, the country, and the content of the call, so run it past counsel before you scale. If email is also part of your sequence, our guide to CAN-SPAM, GDPR, and CASL compliance covers the email side of the same questions.

Where AI voice agents actually fit

Given those rules, AI voice is the wrong tool for dialing strangers and the right tool for several jobs next to it.

Answering inbound calls. Someone who rings your number has initiated the contact, and an agent that answers instantly, qualifies them, and books a time converts a missed call into a conversation. If you are deciding between an answering product and an outbound agent, this comparison of AI receptionists and AI voice agents lays out costs, rules and an evaluation checklist side by side.

Following up on people who asked to hear from you. A form submission, a quote request, or an opened proposal is a consented trigger, and a fast call on a warm lead is a different proposition from a cold dial. This is where Belkins' "follow-up requested" bucket, 19% of non-meeting outcomes, becomes recoverable at scale.

Handling the repetitive layer. Logging outcomes, rescheduling, and routing a call to the right person are tasks where an agent saves human time without raising the consent question.

What it should not do is replace the human conversation on a cold first call. The part of the funnel where the data shows the biggest leak, the 4.6% of conversations that become meetings, is the part that depends on listening, judgment, and the kind of opener that breaks a prospect's expectations. That is the work to give your best rep.

A practical operating model, built from the data

Put the findings together and a sensible cold calling operation looks like this.

Start from the list, not the script. More than a quarter of non-meeting outcomes are the wrong person or a gatekeeper. Fix the data and the trigger before you fix the pitch. A call to someone with a reason to talk this month beats a perfect script read to a stranger.

Plan three attempts per prospect, at different times, before you start. Belkins' per-prospect connect rate of 24.5% required about three tries each. Space them across days and times of day, and put a short email or LinkedIn touch between calls so each attempt carries new information.

Concentrate your hardest calling where it converts. Mid-week and late afternoon outperformed Friday and the busy morning block in the Belkins data, and the first half of the year and the autumn outperformed the summer. Test this against your own list rather than taking it on trust.

Change the opener first. It is the cheapest variable you have. Test "How have you been?" against whatever your team says today, and drop any line that hands the prospect an easy exit.

Treat "send me something" as a lead, not a dismissal. Follow-up requests made up 19% of non-meeting outcomes. Log them, do what you said you would do, and call again.

Stay inside the rules. Screen UK numbers against the TPS and CTPS, honor every objection, check US state requirements, and do not point an AI voice agent at cold numbers without consent.

If you want the full templates, with the call openers and follow-ups annotated against the data that justifies each line, the annotated cold email, cold call, and LinkedIn templates post is the companion to this one.

So, does cold calling still work?

It works the way a high-volume, low-yield channel works. At roughly 370 dials per meeting, it will not carry a business on its own, and a team that treats it as a standalone engine will be disappointed. Used as one part of a sequence, aimed at a list with a real reason to talk, with an opener that does not invite the easy no, and repeated enough to reach the prospects who did not answer the first time, it still books meetings that other channels do not.

The data does not say cold calling is dead. It says the gap between a team that runs it well and one that runs it badly is enormous, and that most of the gap sits in decisions made before and during the first thirty seconds of the call.

Sources

• Belkins, B2B Cold Calling Benchmarks, 175,000+ dials, Nooks dialer data plus five HubSpot campaign exports, calendar year 2025
• Gong, Effective Cold Call Opening Lines, analysis of 90,380 cold calls
• Gong, Is Cold Calling Dead?, same dataset, article originally published 2018 and last modified March 2026
• FTC, Complying with the Telemarketing Sales Rule, business-to-business exemption
• ICO, Telephone marketing, TPS and CTPS screening for B2B calls
• Wilson Sonsini, FCC Rules AI-Generated Voices Are "Artificial" Under the TCPA, FCC Declaratory Ruling of February 8, 2024
• Cornell Legal Information Institute, 47 U.S.C. § 227, TCPA statutory damages
• Harvard Journal of Law & Technology, FCC Cracks Down on AI-Powered Robocalls, unanimous ruling and consent requirement
• Wiley, FCC Extends Regulatory Reach Over AI: TCPA Restrictions Cover AI-Generated Voices in Outbound Calls, scope of the ruling

FAQ

Related reading

• B2B lead generation: channels, costs, and what converts
• AI receptionist vs AI voice agent: which one do you need?
• B2B outbound templates: cold email, cold call, and LinkedIn, annotated
• Signal-based outreach: how it actually works
• Cold email compliance: CAN-SPAM, GDPR, and CASL
• B2B buying signals: the complete guide
• LinkedIn outreach: what 13 million connection requests show