Data Management Platform (DMP) - What it means for the future of Analytics?

Today there is a tremendous amount of "buzz" about the Data Management Platform or DMP space.  Top venture capital firms like Kleiner Perkins are placing large investments in upstart companies that are focused on building these capabilities.

The majority of the new DMP companies’ principal focus is to create a marketing platform to manage the tremendous volume of click stream and ad serving data for the purpose of improving ad targeting.

These companies envision that the Data Management Platform will function as the backbone for all advertising operations.  The intended scale is to manage billions of events per day.  As an aspirational vision, they also hope to incorporate online consumer interaction and profile data to create a universal data collection and centralized storage environment.

Many DMP companies have built audience management and distribution capabilities including deep integration with online ad networks, automated agency trading desks, ad exchanges and content owners.

The centralized availability of this robust data set coupled with the unique ability to immediately action insight through the real time integration with ad inventory creates an exciting proposition for marketers.  Clearly, the ability that this will give marketers is very exciting.

Several data management platforms are now starting to include some offline interaction and profile data in addition to the rich anonymous online data streams they contain.  At the same time, traditional offline consumer data companies are venturing into the management of online behavioral streams.

Over the next twelve months, key battles will be won and lost for market share of the “right” to manage online and offline consumer interactions in a centralized management platform.

While today it is impossible to pick who the winners will be for the right to manage the centralized data platform for brands, one key thing is clear… the “action-ability” of this massive hadoop of data will be enabled by analytics and data science.

The ability to create consumer-centric insights from multi-channel interactions will be a key enabler for marketers.  The science of channel attribution will separate the pack of leaders and laggards.

In its current state, the practice and profession of marketing analytics is ill equipped to address this challenge.

To create this “action-ability”; analytic approaches, skills and culture from data base marketing, web analytics, and big data science (such as genetic mapping or climate modeling) will have to merge into a single team and methodology.

Traditional database marketing analytics has developed a unique capability over the last 25 years of combining campaign interaction data with metadata about populations (in a known capacity) to create target-marketing segments.  These rich metadata descriptions of populations enable the ability to derive segments or personas.  Database marketers have at their finger tips a nearly limitless availability of third party data sources that help them enhance and describe their segments or personas in tremendous detail.

Database marketing analytic professionals however are generally not as adept when dealing with transactional data, anonymous and unstructured data.  In addition, database marketers generally believe philosophically in a hard and direct attribution approach, which is fundamentally flawed.

This philosophy demonstrates an undue bias between the relationship of cause and effect.  Just because a brand sends you a snail mail piece or an email… and you call or click and buy, this does not imply 100% attribution as often the result is recorded.  Because analytics in this environment deal with a myopic data set, additional influences (such as above the line media spend; TV, out of home and print) are often ignored.

Similarly, web analytics professionals place their stock in anonymous behavioral data and context coupled with the outcomes of multi-variant tests they perform leveraging cheap discriminates.  The allure of this approach cannot be denied because of the massive volumes of data that is collected, the number of experiments that can be conducted and the relatively short time frames with which iterations can be implemented.

As voluminous as this data set is however, it too is myopic.  Any predictive applications such as In Market Timing, Next Logical Product or Brand Affinity will be woefully inaccurate if they omit the offline interactions that the consumer exhibits with the brand. 

The approach commonly utilized requires a number of interactions (typically in the form of clicks) from the anonymous visitor before a behavioral segment can be predicted and a treatment applied.  The leading marketing approaches over rely on the associative analysis of products in the ecommerce catalog and the shopping basket relationship between them, under emphasizing the attributes about the consumer.

Anonymous analytic approaches alone will always be limited to behaviors and context.  Fundamentally, these approaches will miss out on the important ability to “jump-start” personalization engines with known persona behaviors.

Professionals in the anonymous analytics practice lack the availability, choice and match-ability of third party data sources that their offline counterparts enjoy.

In addition to the coming mash-up of online and offline analytic methodologies, it is no secret that marketing data sets are growing exponentially in scale and volume.  Gigabytes have come and gone to be replaced by tera, peta, exa, and zettabytes.  What may be a surprise to some however is that these large datasets are not new. 

The scientific fields of climate modeling and prediction as well as the Human Genome and Proteome projects have been working with big data analytics for almost ten years.

Commerce once again has the perennial opportunity to borrow a play from Science and adopt key learning’s from the big data methodologies that have been developed. 

Sagacious marketing analytics professionals will begin to adopt the spatial modeling and advanced visualization techniques that data scientists have perfected as well as new tools like “Revolution R”.

In summary, database marketing practitioners are exceptionally skilled at working with the "Known" populations.  Web analytics practitioners are very good at working with "Anonymous" behaviors and context.  Data scientists have developed leading methods for spatial analysis and data visualization that will guide marketing professionals into the big data world.

Specifically, expect these five key trends in the future of Analytics as a result of the emergence of a true Data Management Platform:
  • A "softening", expansion, or evolution in the way that database marketers think about attribution
  • Emergence of online matching capabilities
  • An expanding set of third party "Enhancement" data sources for online behaviors, context and intent
  • A greater use of visualization technologies and software
  • The expansion of the "Known" and "Anonymous" population segments to include a third category of "Persona" which will be implied from the pattern analysis of web behaviors of Known populations, then inferred across a subset of the "Anonymous" dataset

Anonymous - lacking individuality, unique character, or distinction [Described principally by behavior or context]

Persona – a person’s perceived or evident personality [Inferred from a “Known” population that consistently exhibits a similar behavior and context pattern]

Known (or Logged In) – to be acquainted with (a thing, place, person, ect.), as by sight, experience, or report [A population that has previously registered, self identifying or a population that has been identified from a previous self identification and is technically matched such as in the co-registration process]