Tom-James Lee is the CEO of Vagabond Digital a paid social media agency focused on creating outsized results leveraging data.

Why we love it

  1. Update copy/ creative/ targeting
  2. Test like crazy - one variable at a time
  3. Optimise based on what’s working

This is the framework to follow for testing things, specifically "one variable at a time".

Often times brands get carried away and look to test multiple variables at the same time which tends to lead to issues with tainted data, which requires people to go back through and retest elements.

Internally, we've always followed a 60-20-20 rule.

60% safe built upon baselines that we know to be successful

20% building upon what's safe and successful with the intention of improving it progressively

20% anything goes as long as there is a goal that can be tied to it

Optimization on what's working - this is becoming harder and harder to do with the platform only tools, what's working is something that I think most ad agencies don't really completely understand.

What we'd love to know more about

How do you factor in timing into optimizing?

What is the ideal testing period given changes in attribution modeling?

How do we factor in other elements that can tell us about intent for our modeling?