You launched two campaigns a fortnight ago, each with its own target. The first has one positive reply, the second has two. It's tempting to cut the first and move everything to the second.
That's too early, by a long way. This piece shows why with actual numbers, then works through the calculation that tells you how many sends you really need, one step at a time, without assuming you've studied statistics. Throughout, a send means one prospect contacted, follow-ups included.
Why 20 sends prove nothing
Take two targets of 20 prospects each:
- target A: 1 positive reply out of 20, or 5%;
- target B: 2 positive replies out of 20, or 10%.
B looks twice as good. Now imagine both targets actually respond exactly the same way: every prospect has a 5 in 100 chance of replying positively, in A as in B. Over 20 sends, here is what chance alone produces:
- 0 replies in 35.8% of cases (that's 0.95 multiplied by itself 20 times);
- 1 reply in 37.7% of cases;
- 2 replies in 18.9% of cases;
- 3 or more replies in 7.5% of cases.
So a target at 5% gives you 1 reply out of 20 a bit more than one time in three, and 2 replies nearly one time in five. Getting 1 on one side and 2 on the other from two identical targets happens 14% of the time. More broadly, two identical targets show different scores 69% of the time: on 20 sends there's nearly always a “winner”, even when there isn't one.
The underlying problem fits in one sentence: at 20 sends, a single reply moves the rate by 5 points. Chance weighs more than the gap you're looking for.
Two possible mistakes, and the risk you accept for each
Comparing two targets exposes you to two opposite mistakes.
Seeing a winner that doesn't exist. The two targets are equal, chance puts one ahead, and you declare it better. For this mistake, the usual convention is a 5% risk: across twenty comparisons of identical targets, on average one would name a false winner. This is the error risk, written α.
Missing a real gap. One target really is better, but the numbers don't show it yet and you conclude they're equal. Power is the probability of spotting the gap when it exists. The usual aim is 80%: a real gap is caught 8 times out of 10 and missed 2 times out of 10.
These two values are conventions, and you can change them. Tightening them (1% risk, 90% power) takes more sends.
The calculation, step by step
Starting point: target A replies at 2% (positive replies), target B at 5%. These are assumptions for the calculation, not market averages. Nobody knows a target's rate in advance; that's exactly what you're trying to measure.
The standard formula for comparing two proportions gives n, the number of sends needed per target:
The ingredients:
- p₁ = 0.02 and p₂ = 0.05: the two assumed rates;
- p̄ = (0.02 + 0.05) ÷ 2 = 0.035: their average;
- zα = 1.96: the value matching a 5% risk split across both directions (A better than B, or the other way round);
- zβ = 0.8416: the value matching 80% power.
Textbooks write these as z₁₋α/₂ and z₁₋β. They're the same numbers, read from a normal distribution table.
The gap to detect
0.05 − 0.02 = 0.03. Squared: 0.03 × 0.03 = 0.0009. That's the denominator.
The risk term
2 × 0.035 × 0.965 = 0.06755. Square root: 0.2599. Times 1.96: 0.5094.
The power term
0.02 × 0.98 = 0.0196 and 0.05 × 0.95 = 0.0475. Sum: 0.0671. Square root: 0.2590. Times 0.8416: 0.2180.
Add them, then square
0.5094 + 0.2180 = 0.7274. Squared: 0.7274 × 0.7274 = 0.5291.
Divide by the squared gap
0.5291 ÷ 0.0009 = 587.9. Always round up to the next whole number: 588 sends per target.
588
sends per target to tell a 2% positive reply rate from 5%
Two-sided test on two proportions, 5% risk, 80% power. That's 1,176 sends across both targets.
Why so many? Because the gap is squared in the denominator. As it shrinks, the number of sends climbs much faster than the gap falls.
What other gaps cost
Sends needed per target, by assumed rates
| If target A replies at | and target B at | Sends per target | Weeks at 23 sends per week per target |
|---|---|---|---|
| 2% | 3% | 3,826 | 166 |
| 2% | 4% | 1,141 | 50 |
| 2% | 5% | 588 | 26 |
| 3% | 6% | 749 | 33 |
| 5% | 10% | 435 | 19 |
| 10% | 15% | 686 | 30 |
| 10% | 20% | 199 | 9 |
Three things stand out:
- A small gap is close to invisible at low volume. If one target replies at 2% and the other at 3%, you need 3,826 sends per target, which is more than three years at 23 a week.
- For the same ratio, higher rates are cheaper to separate. In 2% against 4%, as in 5% against 10%, one target replies twice as often as the other. Yet the first needs 1,141 sends per target and the second 435.
- The gap in points isn't enough on its own. 10% against 15% (5 points) needs more sends than 2% against 5% (3 points). At 10%, the number of replies swings more from one batch of prospects to the next, and what counts is the gap measured against that noise.
What that means in weeks
The pace of 23 sends per week per target is the 200 credits a month tier (200 prospects contacted) split between two targets: 200 × 12 ÷ 52 ≈ 46 prospects a week, so 23 per target. The same sum gives 69 per target for 600 credits a month, and 173 for 1,500.
To reach 588 sends per target:
- 200 credits a month (23 per week per target): 588 ÷ 23 ≈ 25.6, so about 26 weeks;
- 600 credits a month (69 per week per target): 588 ÷ 69 ≈ 8.5 weeks;
- 1,500 credits a month (173 per week per target): 588 ÷ 173 ≈ 3.4 weeks.
The credit tiers are on the pricing page. On the first tier, six months to find out which of two targets responds better is a long time. Fortunately it's not the only route.
Launch two campaigns and follow how their targets rank, labelled early signal or reliable.
See 5 prospects for my offerFree, no card needed
Three ways to avoid waiting six months
Measure a more frequent signal: the click
Clicking the hook link (the audit, the guide, the case study, the calendar, the trial) takes less effort than replying. Clicks therefore come more often and sooner, and gaps between targets show up with fewer sends. If one target clicks at 10% and the other at 20%, 199 sends per target are enough: about 9 weeks at 23 a week per target, 3 weeks at 69, and just over a week at 173.
The limit: a click isn't a meeting. A target can click a lot and never buy. Use clicks to sort quickly, then confirm on positive replies once the volume allows. The simple question has no link, so with it you only have replies, which means the long road.
Compare targets that really differ
You don't choose the real gap between two targets, but you do choose what you compare. Two variants separated by a detail may well respond almost identically, and the table shows what a small gap costs: 3,826 sends per target for 2% against 3%. Two variants that differ on something that changes the prospect's situation (a signal present or absent, a founder rather than a head of marketing, 1 to 9 staff rather than 50 to 249) stand a better chance of producing a gap that shows quickly.
Still keep to one difference per test. The method is in the piece on checking your prospecting target.
Don't declare a winner while it's an “early signal”
588 is for planning. If the real gap is wider than 2% against 5%, it will become clear sooner; if it's narrower, later, or never at your volume. Real results decide.
In MonSDR, every target or hook ranking is labelled early signal (the gap could still be chance, and the tool estimates how many more sends it will take to decide) or reliable (the gap passes a two-proportion test at a 5% risk, or the two groups' Wilson intervals no longer overlap, with at least 30 sends per group).
While it's an early signal:
- keep both targets at the same volume;
- don't change the hook or the tone on one side only, or you're comparing something else and the count starts again;
- decide before launch how many sends you're prepared to wait for, and don't stop on the day the numbers lean the way you hoped.







