
To find your own best cold calling time, test four hourly slots for two weeks, with the same list type, the same reps and at least 200 dials per slot, and compare connect rate per dial. Start from the benchmark hypothesis: in 1,000,000 French B2B cold calls, 10am (14.0%) and 4pm (15.8%) beat 8am (9.8%). Then check it on your market instead of trusting ours.
A benchmark is an average of other people's lists. Ours covers B2B calls placed in France, across many personas. Your list has its own mix of personas, countries and number types, and each of those shifts the best window.
The benchmark itself says so: its gaps may be correlations without demonstrated causality, to be tested in your own context. A two-week test is how you do that without turning it into a research project.
The most common mistake is to test "timing" in general. Pick one question.
Start with the hour. It moves a frequent event, so it gives a readable answer in two weeks. Leave the day question for a quarter of data. What we already know about it is in the best day to cold call.
Every slot you add divides your dials. Four is enough to learn something:
Define each slot as a full hour in the prospect's local time. If you call several countries, test one country at a time.
We exclude every segment under 200 calls from the benchmark's connect and time-slot rates. Use the same floor: 200 dials per slot, 800 for the test.
Be honest about what 200 dials can show. At a connect rate around 11%, the margin of error on 200 dials is roughly plus or minus 4 points. That is just enough to see a gap as large as 9.8% against 15.8%. It is not enough to separate 11.6% from 11.9%.
If you want to rank close slots, aim for 500 dials per slot, where the margin falls to under 3 points. If you cannot reach 200, the test will tell you nothing: keep the benchmark windows and collect longer.
How long does that take? In the benchmark, a rep dials 22 numbers per hour by hand, 52 on a power dialer and 157 on a parallel dialer. For 200 dials in one slot:
A timing test only works if timing is the only thing that changes.
At the end, build one line per slot: dials, pickups, connect rate.
Then act on it for a month: move the main calling block to the winning hour and measure conversations per day. If the count does not move, the test was noise.
The same logic applies to scripts. We describe it in how to A/B test cold call openers without fooling yourself.
It tells you when people answer. It does not tell you whether the conversation is good. The hour is worth about 1.6x on connect rate in our data. The opener is worth about 4x on meeting rate. Run the timing test once, bank the result, and spend the next test on the first 30 seconds. We rank the levers in does cold call timing really matter, and explain why published studies differ in why best-time-to-cold-call studies disagree.
Sonik is an AI dialer for B2B outbound teams. Two things in it make this test short. Sessions are built per list, so each of the four piles becomes a session scheduled in its slot.

And the parallel dialer calls several numbers at once, connecting the rep only on a pickup. At the benchmark pace of 157 calls per hour, one rep collects the 200 dials of a slot in under two sessions instead of nine.
Four slots, 200 dials each, one persona, random split, two weeks. That is the whole method. Expect to confirm one or both benchmark peaks, and occasionally to find a window of your own. This week, export last month's dials by hour and count how many slots already pass 200: you may have half the answer before you start. The benchmark windows are in best time to cold call: what 1M B2B calls reveal.