Standard Media Index is bringing clarity and accessibility to advertising. By obtaining raw invoice data from major media buying groups across the globe and then organizing that data, SMI provides the most accurate view into the advertising marketplace.
What Makes Us Unique
- Exclusiveness of our data
- Accuracy between 96% – 99% against networks internal numbers
- Granularity of unit cost by program/category, scatter/upfront/DR, ADUs
- Timeliness – about 2 weeks after month end
- Ingest in a data feed and combine with other data (e.g. ratings)
Since 2009, SMI has been a trusted source of accurate ad spend data and because we have insight into some of the largest national marketers, it is only natural for us to provide full market forecasts for national marketers across all media types – TV, Digital, Other traditional publishers (Print, OOH, Radio). One key component missing from our dataset is insight into small and medium businesses as well as local advertisers (hereon referred to as SMBs), which also make up a big portion of total advertising revenue, especially in digital. Hence, our objective is to not only forecast national marketers, but also provide forecasts for SMB marketers. Therefore, we will provide a full market overview of total advertising spend across all media types and media sub types
The final deliverable will be a full market 5-year quarterly forecast for the following media types / subtypes:
Ad Formats: Search, Display, and Video
Ad Platform: Social*
Shown by revenue from National marketers, SMB marketers, along with Total Revenue (National + SMB marketers)
- Cable TV
- Broadcast TV
- Syndication TV
- Local TV
Other Traditional Media
- Combined Radio
Shown by revenue from National marketers, SMB marketers, along with Total Revenue (National + SMB)
Note: Social has intersecting spend from both video and display ad revenue. Hence SMI considers it an ad platform instead of ad format unlike Search, Display, and Video. Please also note that television is naturally split by distribution, with Broadcast, Cable, and Syndication forming National TV and Local being a combination of Local Broadcast and Local Cable.
Time-Series Forecasting Methodology
Individual models were created for Cable, Broadcast, Syndication, Local TV, and then models for both total market and national advertisers for each of Radio, OOH, Newspapers, Magazines, Search, Display, Video, and Social.
Our forecasts for SMB advertisers are based on the difference between our total market and national advertiser forecasts, which we have true insights into.
For each model, we tested for autocorrelation within the subtype to find optimal lags for an autoregressive function with seasonality, tested a universe of economic indicators (which are listed below), and tested different types of models between linear regression with different data transformations.
In cases where a subtype is glaringly driven by or competing against another subtype (e.g. digital display being driven by growth in social), we also tested the subtype as a regressor in order to account for the unnatural growth or decline seen in the subtype we are modelling.
In cases where we know significant events are impacting our numbers (e.g. Olympics in TV), we use indicator variables to account for these outliers.
Backtesting each of these models, seeing how the models would have forecasted historical quarters using only data available on that day, we were able to select the best modelling technique for each subtype along with which economic indicators were most predictive according to how they performed in historical forecasts.
In order to account for pressures external to any individual subtype (competitive spending, overall advertising budgeting, etc.), we also created a model for how overall advertising spend between all subtypes (TV, Traditional Media, and Search, Display, and Video within Digital) trends over time.
Using this overall advertising spend model to normalize the rest of the forecasts, we are able to still capture relative growth rates of individual subtypes barring these external pressures, while obtaining advertising volumes that do account for them.
Economic Indicators used:
Industrial Production: FRED (Federal Reserve Economic Data)
Personal Consumption Expenditures: FRED (Federal Reserve Economic Data)
Gross Domestic Product: FRED (Federal Reserve Economic Data)
Corporate Profits (After Tax): FRED (Federal Reserve Economic Data)
Total Retail Sales: US Census Bureau
Retail Inventory to Sales Ratio: US Census Bureau
Consumer Confidence Index: OECD (Organization for Economic Co-operation and Development)