A test is not «try it and see». It is a measurement with a volume and a stop rule set in advance. Without both you are not testing, you are spending.
How much volume
A hundred contacts per variant is the minimum at which numbers mean anything. At twenty you are looking at noise: two conversions instead of one looks like doubling and is not.
Practical consequence: do not run five variants when you can afford a hundred contacts total. Two variants of fifty beats five of twenty.
What to measure
- Delivered. What actually went out, not what you planned.
- Replies. Does the opening message work.
- Clicks. Does the transition to the offer work.
- Conversions. Does the offer work, and is it the right audience.
The stop rule
Set it before launch, in numbers: «close it if under five replies per hundred contacts». A rule invented afterwards is always lenient — you have already invested and do not want to admit it was wasted.
After a good test
Do not scale straight to double. Repeat the same volume once more: some good tests are luck, and a second run reveals that more cheaply than scaling does — the order for that is in scaling a setup.