The mistakes
- Broad context hints. "CRM software" reaches everyone. "CRM for a 40-person field service company" reaches your buyers.
- One ad group for everything. Mixing problem, comparison and switching questions makes every ad vague.
- Sending clicks to the homepage. Send buyers to the page that continues their question.
- No conversion event. Without one, you can't move to conversion bidding or prove value.
- No company identification. Most visitors never fill in a form; without identification they stay anonymous.
- Including customers. Upload a hashed customer list as an exclusion. It has no minimum size.
- Wrong country setup. Delivery is restricted only to countries you include; exclusions alone don't restrict it.
- Changing the objective mid-test. You can't. Plan click and conversion campaigns from the start.
- Judging on clicks after 30 days. B2B needs 60–90 days and should be judged on companies, meetings and pipeline.
The pilot that shows why it matters
A published B2B pilot spent $1,470 for 259 clicks and recorded zero conversions. Nothing connected the clicks to pipeline. The lesson isn't that the channel failed; it's that the test couldn't tell. How to judge a test.
A pre-launch check that catches most of them
| Check | Done? |
|---|---|
| Each ad group is one buyer question | |
| Hints include company size, industry or stack | |
| Each ad links to the page that continues the question | |
| lead_created fires and is deduplicated | |
| Company identification is running | |
| Customer list excluded | |
| Countries included explicitly | |
| Click and conversion campaigns planned separately | |
| Thresholds for day 90 agreed in writing |
Two less obvious mistakes
- Copy that repeats the answer. The answer above already explained the category. Your ad should add a reason to choose you.
- No sales follow-up plan. Identified companies that wait a week go cold. Agree who calls whom, and how fast, before launch.
Questions
What's the single most common mistake?
Broad context hints. They spend budget on conversations your buyers aren't in.
How quickly should we change a failing ad group?
Give it two to four weeks of data, then change one thing: the hint, the ad or the page.
Is it a mistake to start small?
No, as long as the test is big enough to reach your thresholds. Two campaigns at the minimum are a sensible floor.
Start here: ChatGPT ads for B2B. More articles in ChatGPT ads.