The minimum for a fair test
- ChatGPT's minimum is $25 a day per campaign in the US and £15 in the UK.
- Two campaigns at the minimum come to about $1,500 a month.
- At $4.50–6.50 a click, $3,000 a month buys roughly 460–670 clicks.
Below about $1,500 a month the test is too small to tell you much in 90 days.
Where to find the money
- The paid channel with the weakest link to pipeline. Often broad display or untargeted social.
- Branded search spend where you already rank first organically.
- Events with a poor record of meetings.
Moving money keeps the total flat while you test.
When to scale
- In-profile companies arrive every week from specific buyer questions.
- Sales books meetings from them.
- Pipeline from the test is on track to pay back the test's full cost.
Then raise budget on the questions that work, 10–20% a week.
Count the full cost
Media is one line; management, creative, tracking and company identification are the others. Full cost breakdown.
Three starting budgets
| Situation | Media per month | Why |
|---|---|---|
| Small company, one market | $1,500 | Two campaigns at the minimum, one question each |
| Mid-size, US and UK | $3,000–5,000 | Separate campaigns per country and stage |
| Larger, several products | $5,000–10,000 | Campaigns per product line and market |
How to present it to finance
- A fixed test budget with a stop date.
- The thresholds that trigger scaling.
- Which existing line funds it.
Questions
Is there a maximum that makes sense for a test?
Spend beyond what your buyers' questions can absorb buys clicks from the wrong companies. Increase only when fitting companies grow with spend.
Should media and management be budgeted separately?
Yes. Media scales with results; management and creative are fixed for the test.
What if we can't move money from another channel?
Then size the test so one won deal pays for it.
Start here: AI ads for B2B. More articles in AI advertising strategy.