# Social Content Experiment Designer

Design controlled social media experiments that isolate one meaningful variable, define success in advance, and avoid learning from noisy comparisons.

## Prompt

You are a social growth analyst who specializes in practical content experimentation and causal discipline.

Inputs:
1. Platform, account, audience, and business objective: {{context}}
2. Current baseline and available performance data: {{baseline}}
3. Hypothesis and variable to test: {{hypothesis}}
4. Content capacity, schedule, budget, and test duration: {{resources}}
5. Algorithm, brand, legal, and measurement constraints: {{constraints}}

Do the following:
1. Convert the proposed idea into a falsifiable hypothesis with one primary variable, expected mechanism, target audience behavior, and decision the result will inform.
2. Define control and treatment content that remain comparable in topic, quality, timing, distribution, and production effort while varying only the chosen factor as closely as practical.
3. Select one primary metric and supporting diagnostic metrics, set guardrails, define the analysis window, and explain confounders such as audience overlap, trend effects, paid distribution, and platform volatility.
4. Recommend a feasible sample and publishing sequence, plus rules for handling outliers, deleted posts, algorithmic anomalies, and inconclusive results without cherry-picking.
5. Produce an experiment brief, content matrix, tracking sheet structure, pre-registered decision rules, result-interpretation guide, and next-test recommendation for win, loss, or inconclusive outcomes.

## Best for

Social media teams testing formats, creative choices, or calls to action with more rigor than comparing whichever posts happened to perform best.

## Compatible tools

- Claude
- ChatGPT

## How to use

- Test one primary variable at a time.
- Use comparable topics and distribution conditions.
- Choose the primary metric before publishing.
- Document inconclusive results rather than forcing a winner.

## Customization tips

- Pair posts when account traffic varies by week.
- Track business outcomes as well as engagement.
- Set minimum guardrail performance.
- Plan the next test for every possible result.

## Example input

Context: LinkedIn page for a B2B energy consultancy with 18,400 followers. Baseline: document posts average 6.2% engagement and 34 website clicks. Hypothesis: showing one quantified insight on the first slide increases qualified website clicks compared with a question-led first slide. Resources: eight posts over four weeks, same designer, no paid spend. Constraints: topics must have similar commercial relevance; track UTM visits and consultation-form starts.

## Example output

The experiment pairs four topic-matched post sets, alternating publication order and weekday. Only the first-slide framing changes; body slides, caption length, call to action, and design system remain stable. Qualified website visits per 1,000 impressions is the primary metric, with swipe-through, saves, and form starts as diagnostics. The decision rule requires improvement in at least three pairs without a guardrail decline in negative feedback; otherwise the result is inconclusive and repeated.