BigQuery previews analysis of changes after a campaign launch

What changed

The function reference describes a univariate ARIMA_PLUS baseline without a required control series or external covariates. Outputs include effect estimates and statistical measures. Google Cloud supplies the original documentation; there is no named individual author. Preview availability should not be interpreted as a production test completed for a particular advertiser.

A campaign launch is a plausible use case, but its date rarely exists in isolation. Suppose a brand launches media while introducing a discount and restocking a popular product. Sales above the forecast could reflect several changes. Labelling the difference as advertising's contribution would require evidence beyond the timing of the campaign. This illustrative example explains an interpretation risk, not a limitation observed in an executed query.

Before using an estimate in a budget discussion, review the input series. Confirm that the outcome means the same thing before and after the intervention, and note any changes in data completeness. Consider whether unusual trading periods or breaks in the business make the historical forecast a reasonable comparison. Record these judgments alongside the result so that another analyst can assess them.

Why it matters for advertisers

The attraction is a more convenient analytical workflow. Convenience should help the team examine a hypothesis, rather than turn the function's name into proof of causation. Compare the finding with other available evidence and be explicit about uncertainty. A stronger experiment may be needed when a large investment depends on separating the campaign from other simultaneous events.