Most teams try to improve cold call performance by rewriting the script. After analyzing 1,000,000 B2B cold calls, we would do it in the opposite order: conversations first, then data, then organization, and the script last. A great opener changes nothing when your rep only has 5 real conversations a day.

Why the script is the last thing to fix

Around 90% of cold calls end in voicemail. In our benchmark of 1,000,000 calls in France, the average connect rate is 11.3%, and it moves between 11% and 20% depending on the persona and the time slot.

That number sets the ceiling for everything else. A rep who dials one number at a time gets around 5 conversations a day. Even if a new opener doubles their meeting rate, that is a fraction of a meeting per day. The same improvement on 20 conversations a day is a different business.

Low conversation volume has a second cost. Reps who barely speak to anyone get very few chances to improve. They cannot practice a new objection answer if they only hear the objection twice a week.

So cold call performance is a stack of four layers. Each one only pays off if the layer below it is solid.

Layer 1: conversations, the number of real talks per day

Start with the number of actual conversations, not dials. Dials measure effort. Conversations measure opportunity.

• With single dialing, a rep gets around 5 conversations a day.
• With multi-dialing, that can go up to 15 conversations an hour.

Parallel dialing gives reps 5x more conversations, because the dialer calls several numbers at once and only hands the rep the one that picks up. Our north-star funnel at Sonik is 20 conversations per day per SDR. Below that, nothing else you do will show up in the pipeline.

The test is simple. Pull last week's call log and count the calls that lasted more than 30 seconds. If the answer is under 20 a day per rep, you have found your first problem, and it is not the script. We covered the math of this in how to make more sales calls without hiring more SDRs.

Layer 2: data, who you call and when

Once reps have volume, look at the list: ICP, persona, signal, timing. More dials on a weak list only produce more voicemails, faster.

Timing alone moves the connect rate by more than half. In our benchmark, calls at 4pm connect at 15.8%, calls at 10am at 14%, and calls at 8am at 9.8%. The best slot also depends on who you call: founders pick up more around 10am, marketing around 4pm, HR around 1pm. Sales and leadership personas connect above 15% overall.

Signals are the other half of the data layer. Reps use only about 20% of the signals already sitting in their CRM: website visits, job changes, closed-lost deals, dormant leads. At Sonik, 50% of our own pipeline comes from signals that were already in the CRM. A list built from those signals gives the rep a reason to call before the call even starts.

Layer 3: organization, the conditions to call properly

Cold calling gets pushed aside very quickly when it is just something reps are supposed to fit in around everything else. A demo runs late, a proposal is due, and the calling block disappears.

Three things protect it:

• Protected calling slots, ideally on the slots that connect best for your persona (10am and 4pm for most B2B teams).
• Regular training. At Sonik we train sales three times a week. Once a quarter is not training, it is a reminder.
• Recorded call reviews. You cannot coach what you have not heard. Reviewing real recordings is worth more than any role play.

Organization is the layer managers control most directly, and the one they most often skip because it looks like a calendar problem rather than a sales problem.

Layer 4: content, what reps actually say

Now work on the script: openers, objections, talk tracks, call preparation. This is where you refine what happens once someone picks up, and the gains per conversation are real.

• 65% of reps open by explaining what they sell. Trigger-based openers (a hire, a funding round, a website visit) reach a 7-10% meeting rate, versus 2-3% for pitch-first openers.
• Discovery questions that start with "how" lift conversion by 190%.
• Asking "What do you mean?" before answering an objection lifts conversion by 197%.

These are big multipliers. But they multiply the number of conversations you have. Multiply 5 conversations by a better opener and you still have a slow week.

Common mistakes when fixing cold call performance

Starting at layer 4. Script workshops are visible and fun. They rarely change the number of meetings when the volume and the list are the problem.

Measuring dials instead of conversations. A rep with 150 dials and 5 conversations looks busy. Report conversations and meetings per conversation instead.

Buying data before protecting time. A great list that nobody calls between 10am and 11am is a spreadsheet.

Training without recordings. Feedback based on what the rep remembers of the call is feedback on the rep's memory.

Fixing one layer and stopping. The four layers compound. A team that moves from 5 to 20 conversations, calls the right persona at the right hour, and opens with a trigger is not 3 times better. It is closer to 10 times better.

How Sonik approaches the four layers

Sonik was built around this order. The parallel dialer handles layer 1: it skips voicemails and gets reps 5x more conversations from the same calling block.

The signals module handles layer 2: dynamic lists of website visitors, job changes, closed-lost and dormant leads, so reps call people with a reason to pick up, at the slot where they do.

For layers 3 and 4, every connected call is recorded, and AI prepares each call with the context and the reason to call. Managers review real conversations instead of guessing, and reps get coached on the 20 conversations they had today rather than the 5 they had last week.

Conclusion

Cold call performance is a stack: conversations, data, organization, content. Work from the bottom up, because each layer multiplies the one beneath it. This week, open your call log and count conversations over 30 seconds per rep per day. If the number is under 20, you know where to start, and it is not the opener.