<aside> 💡

Stop guessing which content will perform. Use the right regression technique for your data - and predict views, revenue, and growth before you hit publish.

</aside>

1. Which Regression Matches Your Data?

Answer two questions about what you're predicting to find the right technique.

<aside> ⚠️

The #1 Mistake

Most creators default to linear regression for everything - but linear regression can predict negative values, which is impossible for revenue, views, or engagement rates. Matching your technique to your data is what separates guessing from accurate forecasting.

</aside>

What are you predicting?	Data characteristics	Use this technique
Revenue / earnings	Non-negative, right-skewed (most months modest, occasional spikes)	Gamma Regression
View counts / reach	Non-negative, right-skewed, cannot be zero or below	Gamma Regression
Engagement rate	Bounded 0-100%, right-skewed, most values are low	Gamma Regression
Viral vs. not viral	Binary outcome - yes/no, 1/0	Logistic Regression
Subscriber / follower growth	Continuous, can follow seasonal curves or accelerating patterns	Polynomial Regression
Seasonal content performance	Non-linear patterns, repeating cycles, growth curves	Polynomial Regression
Blog traffic from ad spend	Continuous, normally distributed, can go up or down	Linear Regression
Multi-factor analysis	Many variables (time, hashtags, topic, length) - risk of overfitting	Regularized - Lasso / Ridge
Platform algorithm reach	Dozens of potential factors - need to identify which truly matter	Regularized - Lasso / Ridge

<aside> 💡

Quick Rule

If what you're predicting cannot be negative (revenue, views, engagement), use Gamma regression. If you're predicting a yes/no outcome, use Logistic regression. If you have many variables at once, use Regularized regression to prevent overfitting.

</aside>

2. The 5-Step Prediction Framework

Follow this process for any metric you want to predict. Fill in the fields as you go.

<aside> 📝

Your Prediction Goal

Metric I want to predict: [fill in]

Why this metric matters to me: [fill in]

</aside>

Step 1 - Identify Your Prediction Goal

[ ] Choose one specific metric to start with (subscriber growth, revenue, engagement rate, or viral potential)
[ ] Write it as a specific statement: "I want to predict monthly revenue from sponsorships"
[ ] Define your time horizon: weekly / monthly / quarterly

My specific prediction goal: [fill in] | Time horizon: [fill in]

Step 2 - Characterise Your Data

[ ] Can this value be negative? (if no, eliminate linear regression)
[ ] Is it a binary yes/no outcome? (if yes, use logistic regression)
[ ] Does it follow seasonal or curved patterns? (if yes, consider polynomial)