Your forecast ignores your ad spend. That's why it's wrong.
A model that only looks at the past misses the real driver of demand: marketing, launches, holidays. Exogenous regressors fix naive forecasting by feeding in what actually moves the needle.
Most forecasting tools do one thing: they look at your history and draw a line forward. It’s the statistical equivalent of driving by looking only in the rear-view mirror. You’ll stay on the road — until the road curves because of something ahead.
That “something ahead” is your ad spend. Your launch. The holiday. The thing you’re about to do that changes demand.
Why past-only models miss the turn
A time series model trained on history alone assumes tomorrow looks like yesterday, adjusted for seasonality. Fine, until you triple your marketing budget next month. The model has never seen that, so it predicts “normal” and you miss the spike by a mile — either short on stock or stuck with a plan built for a quieter world.
The fix is to tell the model what’s coming.
Exogenous regressors, plainly
An exogenous regressor (xreg, if you like the jargon) is just an outside variable you feed alongside history: marketing spend by week, campaign start dates, holiday flags, even web traffic. The model learns “when spend goes up, demand follows,” and then uses your planned spend to predict the future demand.
So instead of forecasting blind, you forecast with the levers in view. Combine the impression history with the spend plan and you get a March peak that actually accounts for the campaign you’re running.
The practical win
You stop asking “what will demand be?” and start asking “what demand do we want, and what spend gets us there?” That’s not forecasting anymore — that’s planning with a model instead of a hope.
If your forecast doesn’t know about your ad budget, it’s not forecasting your business. It’s forecasting last quarter.
KIRA.id supports exogenous regressors so your forecasts use the drivers, not just the history. See the feature or email us.