The main difference between a media buyer and "someone who runs ads" is test discipline. A buyer knows upfront how much they will spend on a hypothesis, which metrics decide the result and what they will do if the numbers do not add up. Without that, budgets drain away on "let's wait a little longer". Here is how we build test plans in our media buying service.
In short: the test budget per hypothesis is based on the acceptable cost per conversion (usually 2–3× CPA), success and stop criteria are fixed before launch, and every source is tracked the same way so comparisons are fair.
What a test plan contains
| Element | What we fix |
|---|---|
| Hypothesis | Source, geo, audience, creative, offer |
| Target metric | CPA, CPL, ROMI or cost per qualified lead |
| Test budget | Spend before a decision is made |
| Success criterion | The metric threshold at which we scale |
| Stop rule | When we switch off with no regrets |
| Tracking | UTMs, sub-IDs, events — identical across sources |
How to calculate the test budget
- Derive the acceptable cost per conversion from offer economics: margin, close rate, repeat purchases.
- Allocate 2–3 acceptable CPAs per hypothesis — usually enough to see a signal.
- For expensive B2B conversions use intermediate events: qualified lead, booked call.
- Split the total test budget across 3–5 hypotheses instead of betting on one.
Stop and scale rules
- Stop: 2–3 CPAs spent with zero conversions, or quality metrics (bounces, time on site) point to fraud.
- Extend: conversions exist but CPA is no more than 20–30% above target — optimise creative and landing page.
- Scale: CPA within target at volume — raise the budget gradually while watching stability.
Traffic quality control
New sources need checking: abnormal CTR, click spikes from one subnet, zero time on site are all signs of low-quality traffic. Unified tracking and server-side analytics reveal this in days, not months. We set it up together with marketing analytics.
Frequently asked questions
How long does a test take?
Think in budget and number of events rather than days. Usually from a few days to two weeks depending on conversion frequency.
Can I test many sources at once?
Yes, if the budget covers a proper test for each. Otherwise go in waves of 2–3 hypotheses.
Should I start with the source or the creative?
First make sure the offer and landing page work on a proven source such as Meta Ads or Google Ads. Then expand to new sources with the same creative.
Want systematic media buying instead of chaotic launches? Send a request on the media buying page and we will model the economics, build a test plan with limits and rules, and run the launch.