Blog
- Categories and numbers in one table: how many rows do you need?
When effects differ by region, plan or month, each combination of groups needs its own rows. What we measured, and the 30-rows-per-combination rule.
- Your metrics have dates. DAGsight now uses them to tell cause from effect.
Daily spend, visits and signups move together. DAGsight now uses the order of your dates to tell which way links run, and what moves the next period.
- Your model says a feature matters. That doesn't mean changing it does anything.
SHAP and feature importance tell you what predicts the target, not what to change to move it. Here's the difference — and when you need causal discovery.
- Your dashboard shows correlations. Here's how to find what actually moves the metric.
A practical guide to causal discovery for analysts: how it finds what drives a metric, and why an honest tool admits when the data can't prove a direction.