Start with a hypothesis
Describe who may buy and why. For example, first-time customers might need to see an installation process rather than a brand introduction. Match the hypothesis to one principal change in creative, audience or landing page.
Record enough to reproduce the test
Save the start time, time zone, creative files, landing-page version, audience, bidding method and budget limit. “Second video” is not an adequate identifier. Use the same observation window for orders and refunds.
Eliminate basic faults first
When clicks do not become visits, check links, loading and metric definitions. When carts do not become purchases, examine shipping, payment and delivery expectations. When orders lack contribution, examine costs instead of judging the creative only by click-through rate.
Allow an inconclusive result
A small sample may not support a winner. Set a loss boundary and a review date before continuing. Feed what you learn about customer questions into the next product-page version, so a stopped campaign still leaves useful knowledge.
