i got the actual spike in sales during a campaign (event period - 1 week). what i am trying to do is forecast daily sales during this 7-day period using historical daily sales in the past 1 or 2 years. then i want to compare the actual sales and the forecast, and make statements like - "because of the campaign, sales are 15% higher than what would have been the actual sales if there was no campaign"
any suggestions on how to go about it? any particular technique that would be more appropriate here?
Have a look at the UCM capabilities of SAS (available both in SAS/ETS and SAS Forecast Server). Unobserved component models "decompose" the model into components, so the impact of an event can be understood in a straight forward manner.