Nadia, a marketing director, wants to know if a new video ad campaign will be effective. She discusses impact tracking and key performance indicators with the speaker before launching the campaign. The speaker recommends taking a Bayesian approach when evaluating marketing campaigns, even when A/B testing is not possible. This involves using tools like CausalImpact to build a model based on historical data, simulate what would have occurred without the campaign, and compare it to actual post-campaign data to estimate the campaign's impact. The speaker provides a workflow example using Bayesian time series modeling and CausalImpact to help Nadia evaluate her new ad spend policy for a website, without running a controlled experiment.