• by Anto Chittilappilly , Op-Ed Contributor, August 12, 2011

For some of us, summer is a time to take a relaxing ride down a long, scenic roadway and leave our worries behind for a few hours.  And for those who are lucky enough to drive a convertible, it's very clear how much "top-down" technology enhances this experience.

Marketers have their own "top-down" technology -- called top-down attribution -- that can be used to enhance the performance of their marketing portfolios.

Why Put the Top Down?

Understanding top-down attribution starts with understanding its name.  "Top-down" refers to utilizing a 50,000-foot level view of the marketing data to work toward specific insights and recommendations that enable marketers to perform cross channel marketing portfolio optimization. Bottom up attribution, on the other hand, utilizes more detailed 50-foot level data to work toward its own insights and recommendations.  Specifically, top-down attribution is utilized when any of the channels with which you are working:

·       Do not collect individual user-level (cookie) data - such as TV, radio, print, PR, out-of-home, etc.

·       Do not contain data with unique individual IDs to enable the linking of it to other touchpoints in the marketing touchpoint stack

·       Have missing or inconsistent (can't be trusted) cookie data

What Fuel Does It Use?

Top-down attribution can utilize "summary" level data that simply provides counts of individuals who were exposed to and/or took action upon various marketing initiatives, broken down to specific levels of detail - but without any unique identifying IDs or cookies.  For example, a slice of the total summary level data might state that 500K people were exposed to a given TV ad, 100K of whom viewed it on a given TV station, 50K of whom were located in a given DMA, 1K of whom viewed it on a particular date.

How Does the Engine Work?

Top-down attribution begins by collecting:

·       Summary-level data from the sources that can't provide more detailed user level data

·       User-level data from the sources where it's available

·       Customer-transaction data from CRM systems and other sources

·       Time- and location-specific econometric data that might have an impact on performance, such as:  gasoline price index, consumer confidence index, the weather, unemployment rate, etc.

·       Data pertaining to marketers' business policy changes over time, such as: changes in pricing, discounts, new products, special events, tightening or loosening of customer credit criteria, etc.

·       Marketers' business rules and data taxonomy

Once the datasets above are collected from all the internal and external sources where they are available, the data is cleaned and normalized to a common set of key performance metrics so that it can be compared and analyzed in an apples-to-apples manner across all channels based on your individual business goals.  After that process is complete, statistical algorithms such as regression and neural networks are applied to the data to calculate the relationships between every variable and dimension. 

The byproduct of that mathematical process is a statistical model that attributes monetary credit to every variable and dimension of all your marketing initiatives to which your target audience has been exposed, and applies that credit to each of your defined business goals.  But more than that, the model also enables you to plug numbers into a graphical user interface in a "what-if?" scenario manner to predict the outcome of potential future investments across your marketing portfolio. That's putting the keys to a powerful vehicle in the hands of the marketer.

Why Is It Better Than Your Father's Convertible?

Top-down attribution differs from these more commonly known forms of cross channel analysis such as:

·       Marketing Mix Modeling -- which typically provides a one-time, snapshot-in-time analysis using marketing campaign information, econometrics and marketers business policy changes.  Its output is usually in the form of PowerPoint slides or some other static document that details its findings, rather than an interactive model that provides ongoing insights.

·       Media Mix Modeling -- which provides a one-time, snapshot-in-time analysis whose output is in the form of a document.  But it focuses on media buys rather than the overall marketing portfolio. It analyzes media buys such as TV, Radio, OOH, search and display advertising, but typically not channels such as PR, social, events, etc.

In fact, top-down attribution is an iterative process that creates a virtuous cycle into which the latest data is continually fed and the statistical model is continually tested and refined so marketers always have up to date analysis and insights to drive their "what-if?" scenario planning, as well as steer their portfolio optimization process.

So roll back the ragtop, put on your shades, and consider taking top-down attribution out for a spin.