meta-testing-framework

A structured approach to experiment design, A/B testing, and result interpretation for social media content and campaigns in Uganda and East Africa.

Use When

• The team wants to prove what works rather than guess: which hook, format, language (Swahili or English) or offer wins.
• A test needs a hypothesis, one changed variable such as caption style or format, guardrail metrics, a sample size and run length, and a rule for calling the winner.
• Results from a Meta Ads Manager A/B test or an organic post comparison need reading correctly, including negative results.
• Split a WhatsApp broadcast, SMS or email list between two versions of the message and compare replies or orders, checking the split is balanced before calling a winner.
• A monthly testing calendar is needed that avoids holiday and school-fees periods in Uganda and East Africa.
• Learning from finished tests must be logged and fed into the next campaign.

Do Not Use When

• ad-testing-and-scaling for deciding which live ads to scale, kill or refresh and how to raise budgets.
• meta-algorithm-guide for the platform ranking reference and posting-time schedule.
• advertising-attribution-and-measurement for incrementality, holdout or geo tests on ad spend.
• Stop before declaring a winner on a sample too small to call; report it as inconclusive.