What Does a Data Management Platform Do, Anyway?

By Martin Kihn | January 07, 2015 | 0 Comments

Yes, it manages data. But what does that mean? Excel spreadsheets and the little calculator app in Windows manage data. So do Oracle databases, Hadoop clusters, and the circuits on the International Space Station.

Saying the DMP manages data doesn’t even get us out of the driveway.

One of the best definitions I’ve read of a DMP sits behind a firewall on the wiki of a company called [x+1], one of the pioneering DMPs now owned by (and called) Rocket Fuel. According to [x+1], a DMP:

“. . . provides the requisite, yet somewhat unnoticed, function of data collection, translation, classification, indexing and storage. It’s the ‘plumbing’ part of data-driven marketing online.”

Okay. Collection, translation, classification, indexing and storage. Also, intriguingly, plumbing. These are all activities that could be ascribed to enterprise databases, data warehouses and data marts, no? Doesn’t that Oracle box in your company’s basement do these things? Well, yes.

But it is impossible to understand – much less, learn to love – the mighty DMP outside of its native land. This land is most definitely marketing. If we are being honest, it is actually digital advertising. Or rather, digital banner advertising, what insiders call display. DMPs do more than target display ads, of course, but it remains job one.

So what do they do, already?

At the very highest orbit, DMPs do three things:


Let’s elaborate a bit on each of these three steps and hit pause in the middle for a beauty shot:

  1. IMPORT DATA – This just means the DMP is in the business of taking structured information from a number of different systems and organizing it at the customer level (if it’s about your customers or prospects), or at the cookie level (if the person is unknown). Typically, you would upload just important information about customers, such as their customer ID or email address (to identify them), what they have bought or looked at, their loyalty status – anything that could influence the message you show to them. Also, you would import demographic and other characteristics to help you do the stuff you do in step (2), which is . . .
  2. FIND SEGMENTS – These are known as “audiences” in the DMP world, not segments or clusters, but the idea is the same. You can define your own audiences (e.g., males in Florida over 50 who use an iPad, or whatever), or ask the DMP to help you find them.
  3. Pause: The Beauty Shot – So far, it’s hard to get excited about these things, but real magic happens at this point. The true visionquest of the DMP is not to manage your own customer data (which is just fine in a CRM system, isn’t it?), but to provide a way for you to use this data to find new customers online. It does this by giving you access to the sparkling world of data vendors, by synching up cookies from different places, and by doing some fancy math to help you select exactly the right group of anonymous cookies to buy ads against. Not surprisingly, these fancy groups are often based on the customer data you imported in step (1), and the audiences you discovered in step (2).
  4. SEND INSTRUCTIONS – Instructions are basically a combination of (a) who to target, (b) with what message, and sometimes (c) where, i.e., in what channel or on what device.

BTW, there is another class of marketing software that plugs in precisely here, right after step (4). This is called the Demand-Side Platform, or the DSP. In fact, DMPs and DSPs are neurotically codependent, and many of the best-known DSPs (such as Turn) also have a DMP. The DMP and DSP are separate organisms only in the technical sense.

Alright, so we have a drones’-eye view of the DMP. Subscribers to Gartner for Marketing Leaders research will be able to enjoy a very detailed description of the DMP and how to use it fairly soon. In the meantime, we can take it down one notch.

The three basic categories above can be broken down more usefully. What do DMPs do for marketers? They:

  • IMPORT DATA from your marketing systems, particularly multichannel campaign management, Email, e-commerce and customer loyalty platforms
  • MATCH CUSTOMER IDs where possible – that is, where you have two data sources with a common field like a customer ID or email address (or anonymized ID), these sources can be stored by the DMP as belonging to the same person
  • COLLECT NEW DATA as it becomes available, generally using tags (little pieces of Javascript code) that you physically put into your website, emails, advertisements, and mobile apps
  • HELP YOU DEFINE GROUPS, which marketers like to call segments or clusters, that have something in common
  • PROVIDE ACCESS TO DATA VENDORS – you probably know there are dozens, maybe hundreds, of companies that will happily sell marketers pre-defined or custom segments of (anonymous) people to target with your, ahem, advertising
  • ESTIMATE THE PRICE AND REACH each particular data vendor could get you, so you can manage your budget (“reach” is what media types call “number of people”)
  • SUGGEST NEW GROUPS OF PEOPLE TO TARGET with your advertising by finding people who look like your current customers (these are called, not very elaborately, “lookalikes”)
  • USE ALL THE ABOVE TO BUY ADVERTISING – also, personalize websites, and other things related to delivering messages to people
  • GENERATE REPORTS that tell you how wonderfully (or not) your new DMP is working

If you were paying attention last year, you will have noticed that there was a kind of DMP land grab, with a stampeding horde of Oracle, Neustar, Rocket Fuel and IgnitionOne harvesting the DMPs BlueKai, Aggregate Knowledge, [x+1] and Knotice, respectively. In a very practical sense, the DMP is the heart of the digital marketing hub.

And like a human heart, it keeps alive the hope for a prosperous 2015.

Categories: digital marketing