How to design an honest post-format test

How to design an honest post-format test

The question: which format wins for your audience — and how do you test it without fooling yourself?

Format studies (Socialinsider, Hootsuite, sample sizes in the tens of thousands of posts) consistently rank native documents and polls high for reach. But aggregate benchmarks rarely match your niche. You need your own test.

The playbook:
— Hold the variable: same topic, same week, same posting time, only format changes.
— Run each format at least 5 times — one post is an anecdote, not data.
— Fix your metric upfront: reach per impression, not raw likes.
— Randomize posting days to avoid day-of-week confounds.
— Wait 48 hours before reading results; LinkedIn distributes over days.
— Compare medians, not averages — one viral post distorts the mean.

Limitations: you can't truly randomize an audience that learns your patterns. Five posts per format is still a small sample — treat as directional, and re-test quarterly as the algorithm shifts.

What it means for B2B: the format that wins benchmarks may lose for a technical buyer audience that prefers dense text. Local data beats global benchmarks.

Bottom line: design the test before you crave the answer, or confirmation bias will design it for you.