Causal Impact
Causal impact is the difference between an observed outcome (e.g., revenue) and a counterfactual prediction of what would have happened without a specific intervention (e.g., an ad campaign). It is the rigorous statistical definition of 'lift'. SpendSignal uses Bayesian structural time-series models to compute causal impact.
The Short Version
The mathematical proof that your ad actually did something.
Visual Explanation
GA4 Can't Measure Incrementality
Why the world's most popular analytics tool is blind to the world's most important metric.
Prerequisites
The Post-Hoc Fallacy
Marketing revenue went up after we launched the campaign. Great! But did the campaign cause it? or was it just payday?
Without calculating causal impact, you are prone to superstitious marketing—believing in tactics just because they coincided with a good sales week.
How it works
Predict the counterfactual baseline using pre-campaign data
Example: 'We expected to make $100k'
Compare to actuals: 'We made $120k'. The $20k gap is the Causal Impact.
Common Misconceptions
Confusing 'Correlation' with 'Impact'
Ignoring outside factors (Competitor outage, weather, holidays)
Manually drawing trend lines in Excel (Use a real statistical model)
Related Terms
Frequently Asked Questions
QCan causal impact be negative?
Yes. If you launch a bad campaign that alienates customers, your actual revenue can drop below the counterfactual baseline.
QHow is this different from A/B testing?
A/B testing uses a randomized control group. Causal Impact uses a synthetic control group constructed from historical data. It's faster and doesn't require holding back audiences.