Yes, with the exception of Objective data sets (data feeds). The modeling period is defined by the dates we have valid Objective data, however it is important to understand what to expect in the model results when there is missing data.
When data is missing (eg- there is no new file delivery for online_direct_marketing on the agreed upon normal delivery date), the model will interpret the missing data in one of two ways - either it assumes that nothing has been delivered because nothing has happened, or it “extrapolates” the missing data - meaning it assumes there should be data and so it makes an educated guess and applies that data until the actual data is delivered.
The interpretation is based on whether the missing data set is classified as either Type I, Type II, or Type III. See below for the definitions and classifications for each Type.
Type I
Definition: Data sets where no missing data is allowed, we must have valid data for every day included in the modeling period. E.g. if we have daily data until 2018-12-31, then the model can be updated no further than 2018-12-31.
Type I Classifications: Objective
Type II
Definition: Data sets where missing data is allowed, and the missing dates will be treated as if they have a zero value. This is because the model has to assume that nothing happened during these dates, and therefore there was zero effect against the objective.
Therefore, to ensure the model is providing the most accurate results, accurate model results, it is important to provide data for all dates where something has happened.
Type II Classifications: Paid Media (Paid Media Impressions, Paid Media Gross, Halo Media, & Other Media), Non Paid Online Visitors, Pricing, Competitor Pricing, Product changes, Internal Factors, External Factors, Online Direct Marketing, Offline Direct Marketing
Type III
Definition: Data sets where missing data is allowed, but data is necessary in order for the predictor going into the model does not allow any missing data. A typical example of this competitors spend, that typically has a lag of a month before we can acquire the data. Here, some extrapolation is needed in order to recover a complete set of predictors.
Type III Classifications: Competitor Spend, Branding, Distribution