Track 04
Measurement & ML
Measure whether a change helped, and use that evidence to decide what to try next.
Start here
Start with a decision you need to make
Name the decision. Write down what you would do differently, such as increasing next month’s ad budget.
Pick one outcome. Choose what you want to change and when you will measure it, such as completed orders over two weeks.
Plan the comparison. Decide what evidence would change your mind. A platform can credit an ad for an order. To estimate how many extra orders the ads caused, compare outcomes with and without the campaign.
Worked examples
Worked example · Preview
How to collect ChatGPT Ads reports
Separate what the report credits from the additional sales the campaign may have caused.
Interpret credited conversions
References
Research
Measuring Ad Effectiveness Using Geo Experiments
Jon Vaver and Jim Koehler show how to compare regions where a campaign runs with regions where it is held back.
Read the research on geographic experimentsPublication
Building to learn
Follow the decisions and trade-offs in a marketing optimisation prototype.
Read the optimisation prototype write-up
In development
- Build a demand forecast and compare it with a simple baseline.
- Test a budget allocation rule with past data before using it in a live campaign.