SalsaCrazy / DanceCrazy · Businesses Evan operates

A better testing process.
With judgment built in.

Campaign performance provides the next question. AI helps develop variations. Human judgment decides what is worth running and what the evidence means.

From campaign feedback to the next test

I designed an AI-assisted Meta and Google advertising workflow that uses campaign performance to refine creative and audiences through repeated testing, expanding production pace, audience coverage, and marketing reach.

This work sits within SalsaCrazy and DanceCrazy, businesses I operate. It connects a practical use of AI with the ongoing decisions involved in marketing classes, events and educational offerings.

Workflow illustration
  1. Review campaign performance
  2. Choose an audience or creative hypothesis
  3. Develop variations with AI assistance
  4. Review the work and run the test
  5. Measure what happened
  6. Adapt the next test

Production speed is one part of the value

Being able to develop more variations opens room to explore different messages and audiences. The business question remains: does a variation help the right person understand the offer and take a useful next step?

My responsibility includes the choices around the offer, creative, budget and evaluation. AI supports the working process; performance feedback and human judgment guide its direction. The qualitative value of this work has been expanded production pace, audience coverage and marketing reach.

What I would inspect

These questions show how I would review a campaign. They are an illustrative analytical approach, not a report of results from a named client.

Audience intent

What was this person trying to do when the message reached them? Is the offer relevant to that need and to the local market?

Creative-message fit

Does the creative make the offer understandable? Does it set an expectation that the experience can actually meet?

Conversion behavior

Where do people take the next step, and where do they stop? I would connect that behavior with sales feedback and acquired customers.

Measurement reliability

Are the events being counted consistently? Do reporting periods, attribution assumptions and follow-up timing support a useful comparison?

The next test

Which unresolved question could change a decision? I would make the hypothesis explicit and choose a test that can produce interpretable evidence.

A useful conversation starts here

Let’s talk about
the business problem.

For consulting, collaboration or a conversation about bringing this experience to your team, get in touch.

Email Evan Connect on LinkedIn ↗