Data Management Platform · DMP · nPario

What is Data Science?

A Conversation with Ankur Teredesai, Data Scientist, nPario

These days, the term “bid data” is all the rage and over a dozen data management platforms are competing for the right to manage audience segmentation, targeting, analytics, and lookalike modeling for advertisers and publishers. I recently sat down with noted data scientist Ankur Teredesai to help understand data science, and how ad technology companies are using data science principles to help publishers and marketers understand audiences better. Ankur is the head of data science for nPario, a WPP portfolio company focused on data management, and he is also a professor at the University of Washington.

You hear lots of digital advertising people talk about “data science.” Does their perception differ from the broader, academic understanding of what data science is? Is the notion of utilizing data science in digital marketing applications new?

Ankur Teredesai (AT): It’s very interesting to see that the digital advertising folks have started deep conversations with data scientists. Data science is a very interesting space these days where data mining and database management technology is now helping variety of disciplines in establishing a scientific approach to decision making. Data science dealing with the problems of finding patterns in large amounts of data is not a new concept for digital marketing. What is new is the advent of technologies that now support finding useful patterns in large variety and velocity of data in addition to volume; thereby advancing the state of the art in marketing analytics.

Do ad technology companies really rely on data science? What does being a data-driven organization really mean?

AT: No ad-tech company can afford to NOT rely on data science in some shape or form. The power of predictive modeling is quickly differentiating the players who are making quick inroads by using the low-hanging fruits of data science for these domains from the ones that are treating data science as a passing buzzword. My advice to all ad-tech companies is to get at least one data scientist in their ranks; even if they don’t like the term data science for some or the other reason. Machine Learning, data mining and big data analytics are all equally acceptable today.

Describe the concept of data modeling for the non-academic user. What kind of models are being built for digital marketing applications?

AT: A variety of problems in digital marketing are being addressed using predictive modeling. Some examples of the work we are doing at nPario are (a) look-alike modeling that helps find targetable audiences to “rightsize” or expand a particular segment, (b) recommend cost-aware segments that are similar to desired audience segments for targeting, (c) provide comparative analytics for exploring the unique properties of a given segment, (d) enabling real-time audience classification to reduce the time to target in an efficient and effective manner.

Is the concept of lookalike modeling legitimate? How does this work? Is LAM a scalable targeting practice?

AT: Given customer behavior data, the lookalike model estimates and exploits the variations and similarities in behavior across various segments. Once the model figures out which similarities or differences are robust enough in the audience to be useful for predicting future behavior, it exploits these attributes in the data to expand a particular segment’s size by including those customers that are similar to the base segment but were not included because they did not meet the segment definition criteria. This allows fairly restrictive criteria to be relaxed using data science methods such as association mining and regression analysis to expand the segment size accurately, confidently, and in a scalable manner.

The entire digital advertising ecosystem is driven by data. What are the most valuable types of data for targeting? How do you see the future for the ecosystem? Will those with the most data win?

AT: This is central to success of the entire digital advertising world and the benefit of end consumers in finding the right and useful advertising. If we have to understand the user and focus on making digital advertising useful without making it adversarial, we have to focus carefully on the types and granularity of the data being collected. At both nPario and the Institute of Technology, University of Washington Tacoma where I hold an academic appointment, we stress the need for developing an ecosystem of data collection, management and mining that is customer centric with highest regards for security through robust multitenancy, cryptography and privacy aware practices.  The entire technology stack at nPario is for example, data agnostic. We decided very early on in the company to not be supplies of data but to be data pool neutral to allow our clients to bring their own first, second and third party data. Our platforms help clients derive value from the variety of big datasets while at the same time ensuring that customer and end-user privacy is preserved and not compromised through our actions in any manner whatsoever. To address your question if those with most data win I would like to quote that : Everybody has some data and some just have lots of data. It is the ones that have the right tools at the right time that will monetize their data the best.

This interview, among many others, appears in EConsultancy’s recently published Best Practices in Data Management by Chris O’Hara. Chris is an ad technology executive, the author of Best Practices in Digital Display Media, a frequent contributor to a number of trade publications, and a blogger.

Advertising Agencies · Data Management Platform · DMP

The Role of the Agency in Data Management

A Conversation with David Spitz of WPP Digital 

When it comes to the role of the agency in data-driven digital media, few holding companies have put their money where their mouth is to the extent that WPP Digital has. After setting the tone with a bold acquisition of 24/7 Real Media, the holding company has gone on to place strategic bets on a variety of sectors within the Kawaja map. The question for marketers is whether or not they should be relying upon their agencies when it comes to technology and data. Many argue that the agency model cannot support the type of deep domain expertise needed for the complicated integrations, data science, and modeling that has become an everyday issue in modern marketing. So, should data management be the sole province of the Adobes and IBMs of the world, or is there room for agencies to play? I recently reached out to EVP of Strategy and Corporate Development David Spitz to ask about how he sees agencies working with large brands to define their data strategy.

WPP is working with some of the world’s largest brands. I suspect that many have siloed pockets of valuable data across their enterprises. What are the data challenges and, more importantly, opportunities for global brands?

David Spitz (DS): You are right; there are many data challenges across large enterprises. They range from organizational issues (what group or department should even be running these programs?), to legal and commercial issues (do we have the right to the data we want to use?), to skill set gaps, to challenges posed by legacy technologies, to lack of data standards across channels, brands, regions, or even campaigns. In my experience, though, one thing is clear: it is rarely lack of data that is the problem.

The most common single question we hear from the world’s largest brands is “Where do I start?” It helps to have a clear understanding of the opportunities and choose one or two to focus on to build confidence and momentum while keeping in mind what could come next. “Think big, start small” is one of my favorite phrases when it comes to data programs. In terms of what those opportunities are, it really boils down to what I’ll call the 4 R’s – Reach, Relevance, Resonance and ROI.

Most companies that label themselves as “DMPs” are focused on Reach (e.g., targeting) or sometimes ROI (e.g., campaign evaluation, attribution), and mostly only in a digital sense. That might be a good place to start. However, I have also seen relevance (personalization) and resonance (social amplification) as the jumping off point for some brands. Either way, because these tools exist and can be deployed at relatively low cost, it is often best to start with digital-only applications before expanding the data program into multichannel territory.

Whether you are thinking digital or not, these four areas–Reach, Relevance, Resonance and ROI–probably represent 80% of the data opportunity for big brands, and between them you can usually identify at least one solid quick win.

When it comes to marketing, are these brands looking to their agencies for answers, or are they looking to the IBMs of the world? It seems like the agency’s ability to make an impact ends with the marketing team. Can you extend the agency’s value through to IT teams, and get everyone working together?

DS: When it comes to marketing, brands are absolutely looking to their agencies for answers. It is one thing to come up with an “enterprise architecture” and quite another to have it implemented. In many marketing functions, agencies are on the front lines of where the dollars get expressed, customer engagement happens, and [you are able to] understand what it takes to get data into a place where value can be realized.

Still, do agencies need to do a better job of partnering with CIO’s? Without a doubt. Various WPP companies have in place major partnerships with IBM, Adobe and Infosys to do just that, and at WPP Digital we recently invested in a company called Fabric and acquired a company called Acceleration, both of which specialize in marketing technology systems and, essentially, gap bridging between the CIO and the CMO.

You are working on putting many of WPP’s global data resources together (the “Data Alliance”). Tell us about the project. Is this a global data exchange? Are there unique types of data within the Alliance that are unavailable elsewhere?

DS: Data is at the heart of a lot of what WPP does. You have to realize, WPP is not only the world’s largest communications services group, but if you looked at some of its operating companies as standalone you’d find inside WPP the world’s largest media buying company (GroupM), the second largest market research company (Kantar), and, with $4.7b in revenues coming from digital, including the likes of 24/7 Media, OgilvyOne, Wunderman, AKQA, VML, and Possible, WPP is the seventh largest digital company in the world – behind Google and Apple, but ahead of Facebook right now. So you can imagine, WPP as a whole is dealing with a lot of data.

What we are trying to do with The Data Alliance is analogous to the airline industry, where independently operated carriers have come together to create these inter-company frequent flyer programs (as in the Star Alliance) and coordinated route maps. The whole idea is to provide a more seamless customer experience while at the same time providing efficiencies for the member organizations. Without going into too many details, The Data Alliance is focused on three things: Creating greater interoperability across its members’ platforms and data sets, streamlining how we as a group engage with third-party partners (to make it easier on an Acxiom or Exelate, for example, to work with us broadly), and creating a more seamless experience for clients who are working with us more than one discipline (e.g., media, market research, CRM, and digital).

How we do this will involve many different tactics over time, for example, pooling of certain technology development efforts and greater standardization around certain things like policies, data structures, commercial terms, and API’s. You can speculate about some of the new products and business models that would result out of a program like this, but right now the primary focus is simply on creating the best solutions we possibly can for the top 30 clients who are our “frequent fliers” if you will.

Unlike some other holding companies, WPP has taken an active role in investing in, and acquiring, digital media technology. The “stack” that you are assembling at 24/7 Media, and some of the social media technology investments you have made suggest a commitment to being more than just a typical agency. The Data Alliance initiative is also instructive. Tell us what you look for in differentiated technologies.

DS: WPP comes at it very much from a client-side perspective and has partnered with technologies like Omniture and Buddy Media that share that view. In the cases of those two businesses in particular, both of which WPP invested in, there was beyond the obvious criteria also a strong cultural fit with the management team and a good amount of overlap between WPP’s customer base and theirs, so it just made a lot of sense.

In the case of 24/7, while they were known as a publisher-side technology before WPP acquired them in 2007, the intent was always to leverage their audience reach and technical know-how to build what people would now call a DMP/DSP – the tools that now power Xaxis. There were not any established players doing this at the time, so the 24/7 acquisition enabled GroupM to build these capabilities much faster than they could have otherwise. The acquisition of iBehavior, which operates a DMP of a different sort (mostly offline transactions), is also consistent with this strategy and is similar in that it’s accelerating Wunderman’s route to market with several new initiatives.

To your broader point about being not just a typical agency, I don’t believe agencies need to control all of the underlying technologies, but I do think that the techniques involved in connecting and analyzing diverse data streams – and doing so in a scalable, efficient and privacy-safe way – are too important a skill set for a company like WPP to outsource entirely. When digital is the direction most marketing channels are headed, and the ability to measure everything and act on data is a large part of what makes digital so exciting, not having a data integration and data sciences function (granted, it may be called something else) inside an agency holding company in ten years will seem as unusual as not having a media group would today.

This interview, among many others, appears in EConsultancy’s recently published Best Practices in Data Management by Chris O’Hara. Chris is an ad technology executive, the author of Best Practices in Digital Display Media, a frequent contributor to a number of trade publications, and a blogger.

This post also appeared on the iMediaConnection blog on 12/18/12.

Big Data · Data Management Platform · Digital Display · DMP · Sales

The Five Things to Expect in a DMP

Getting back control over their inventory is giving publishers a lot to think about.

“We want to make sure that we’re controlling what happens with data . . . we want to make sure we control pricing. Control’s a very important message. We don’t want there to be a cottage industry built on our backs” – Nick Johnson, SVP, NBC Universal

What do publishers really want? It’s simple, really: Power and control. In order to survive the ad technology era, publishers need the power to monetize their audiences without relying on third parties, and complete control over how they sell their inventory. In this era of “Big Data,” there is a fire hose stream of tremendously valuable information for publishers to take advantage of, such as keyword-based search data, attitudinal survey data, customer declared data from forms, page-level semantic data, and all the 3rd party audience data you can shake a stick at.

All of this data (cheap to produce, and ever-cheaper to store) has given rise to companies who can help publishers bring that data together, make sense of it, and use it to their advantage. Currently, ad technology companies have been using the era of data to their advantage, utilizing it to create vertical ad networks, ad exchanges, data exchanges, DSPs, and a variety of other smart-sounding acronyms that ultimately purport to help publishers monetize their audiences, but end up monetizing themselves.

Rather than power the ad tech ecosystem, what if data could actually help publishers take back their audiences? If “data is the new gold” as the pundits are saying, then smart publishers should mine it to increase margins, and take control of their audiences back from networks and exchanges. Here are the five things a good data management platform should enable them to do:

  • Unlock the Value of 1st Party Data: Publishers collect a ton of great data, but a lot of them (and a LOT of big publishers) don’t leverage it like they should. Consider this recent stat: according to a recent MediaPost article, news sites only use in-site audience targeting on 47% of their impressions, as opposed to almost 70% for Yahoo News.  By leveraging site-side behavioral data, combined with CRM data and other sources, it is possible to layer targeting on almost every impression a publisher has. Why serve a “blind” run-of-site (ROS) ad, when you can charge a premium CPM for audience-targeted inventory?
  • Decrease Reliance on 3rd Parties: The real reason to leverage a DMP is to get your organization off the 3rd party crack pipe. Yes, the networks and SSPs are a great “plug and play” solution (and can help monetize some “undiscoverable” impressions), but why are publishers selling raw inventory at $0.35 and letting the people with the data resell those impressions for $3.50? It’s time to turn away those monthly checks, and start writing some to data management companies that can help you layer your own data on top of your impressions, and charge (and keep) the $3.50 yourself. Today’s solutions don’t have to rely on pre-packaged 3rd party segments to work, either, meaning you can really reduce your data costs. With the right data infrastructure, and today’s smart algorithm-derived models, a small amount of seed data can be utilized to create discrete, marketable audience segments that publishers can own, rather than license.
  • Generate Unique Audience Insights: Every publisher reports on clicks and impressions, but what advertisers are hungry for (especially brand advertisers) are audience details. What segments are most likely to engage with certain ad content? Which segments convert after seeing the least amount of impressions? More importantly, how do people feel about an ad campaign, and who are they exactly? Data management technology is able to meld audience and campaign performance data to provide unique insights in near real-time, without having to write complicated database queries and wait long times for results. Additionally, with the cost of storing data getting lower all the time, “lookback windows” are increasing, enabling publishers to give credit for conversion path activity going back several months. Before publishers embraced data management, all the insights were in the hands of the agency, who leveraged the data to their own advantage. Now, publishers can start to leverage truly powerful data points to create differentiated insights for clients directly, and provide consultative services with them, or offer them as a value-added benefit.
  • Create New Sales Channels: Before publisher-side data management, when a publisher ran out of the Travel section impressions, he had to turn away the next airline or hotel advertiser, or offer them cheap ROS inventory. Now, data management technology can enable sales and ops personnel to mine their audience in real time and find “travel intenders” across their property—and extend that type of audience through lookalike modeling, ensuring additional audience reach. By enabling publishers to build custom audience segments for marketers on the fly, a DMP solution ensures that no RFP will go unanswered, and ROS inventory gets monetized at premium prices. 
  • Create Efficiency: How many account managers does it take to generate your weekly ad activity reports? How much highly paid account management time are publishers burning by manually putting together performance reports? Why not provide an application that advertisers can log into, set report parameters, and export reports into a friendly format? Or, better yet, a system that pre-populates frequent reports into a user interface, and pushes them out to clients via an e-mail link? You would think this technology was ubiquitous today, but you would be wrong. Ninety-nine percent of publishers still do this the hard (expensive) way, and they don’t have to anymore.

It’s time for publishers to dig into their data, and start mining it like the valuable commodity it is. Data used to be the handcuffs which kept publishers chained to the ad technology ecosystem, where they grew and hosted a cottage industry of ad tech remoras. The future that is being written now is one of publishers’ leveraging ad technologies to take back control, so they can understand and manage their own data and have the freedom to sell their inventory for what it is truly worth.

That’s a future worth fighting for.

[This post originally appeared in ClickZ on 2/29/12]

Advertising Agencies · Data Management Platform · Digital Display · Digital Media Ecosystem · DMP · Publishing · Real Time Bidding (RTB) · Uncategorized

Same Turkey, New Knife

The way the ad tech world looked pre-DSP...and pre-DMP

Technology may still capture the most advertising value, but what if publishers own it?

A few years ago, ad technology banker Terence Kawaja gave a groundbreaking IAB presentation entitled, “Parsing the Mayhem: Developments in the Advertising Technology Landscape.” Ever since then, his famed logo vomit slide featuring (then) 290 different tech companies has been passed around more than a Derek Jeter rookie card.

While the eye chart continues to change, the really important slide in that deck essentially remains the same. The “Carving up the stack” slide (see above), which depicts how little revenue publishers see at the end of the ad technology chain, has changed little since May 2010. In fact you could argue that it has gotten worse. The original slide described the path of an advertiser’s $5 as it made it’s way past the agency, through ad networks and exchanges, and finally into the publisher’s pocket.

The agency took about $0.50 (10%), the ad networks grabbed the biggest portion at $2.00 (40%), the data provider took two bits (5%), the ad exchange sucked out $0.35 (7%), and the ad server grabbed a small sliver worth $0.10 (2%), for a grand total of 64%. The publisher was left with a measly $1.80. The story hasn’t changed, and neither have the players, but the amounts have altered slightly.

While Kawaja correctly argued that DSPs provided some value back to both advertisers and publishers through efficiency, let’s look ahead through the lens of the original slide. Here’s what has happened to the players over the last 2 years:

  • Advertiser: The advertiser continues to be in the cat bird seat, enjoying the fact that more and more technology is coming to his aid to make buying directly a fact of life. Yes, the agency is still a necessary (and welcomed) evil, but with Facebook, Google, Pandora, and all of the big publishers willing to provide robust customer service for the biggest spenders, he’s not giving up much. Plus, agency margins continue to shrink, meaning more of their $5.00 ends up as display, video, and rich media units on popular sites.
  • Agency: It’s been a tough ride for agencies lately. Let’s face it: more and more spending is going to social networks, and you don’t need to pay 10%-15% to find audiences with Facebook. You simply plug in audience attributes and buy. With average CPMs in the $0.50 range (as opposed to $2.50 for the Web as a whole), advertisers have more and more reason to find targeted reach by themselves, or with Facebook’s help. Google nascent search-keyword powered display network isn’t exactly helping matters. Agencies are trying to adapt and become technology enablers, but that’s a long putt for an industry that has long depended on underpaying 22 year olds to manage multi-million dollar ad budgets, rather than overpaying 22 year old engineers to build products.
  • Networks: Everyone’s talking about the demise of the ad network, but they really haven’t disappeared. Yesterday’s ad networks (Turn, Lotame) are today’s “data management platforms.” Instead of packaging the inventory, they are letting publishers do it themselves. This is the right instinct, but legacy networks may well be overestimating the extent to which the bulk of publishers are willing (and able) to do this work. Networks (and especially vertical networks) thrived because they were convenient—and they worked. Horizontal networks are dying, and the money is simply leaking into the data-powered exchange space…
  • Data Providers: There’s data, and then there’s data. With ubiquitous access to Experian, IXI, and other popular data types through 3rd party providers, the value of 3rd party segments has declined dramatically. Great exchanges like eXelate give marketers a one-stop shop for almost every off-the-shelf segment worth purchasing, so you don’t need to strike 20 different license deals. Yet, data is still the lifeblood of the ecosystem. Unfortunately for pure-play segment providers, the real value is in helping advertisers unlock the value of their first party data. The value of 3rd party data will continue to decline, especially as more and more marketers use less of it to create “seeds” from which lookalike models are created.
  • Exchanges: Exchanges have been the biggest beneficiary of the move away from ad networks. Data + Exchange = Ad Network. Now that there are so many plug and play technologies giving advertisers access to the world of exchanges, the money had flowed away from the networks and into the pockets of Google AdX, Microsoft, Rubicon. PubMatic, and RMX.
  • Ad Serving: Ad serving will always be a tax on digital advertising but, as providers in the video and rich media space provide more value, their chunk of the advertiser pie has increased. Yes, serving is a $0.03 commodity, but there is still money to be made in dynamic allocation technology, reporting, and tag management. As an industry, we like to solve the problems we create, and make our solutions expensive. As the technology moves away from standardized display, new “ad enablement” technologies will add value, and be able to capture more share.
  • Publisher: Agencies, networks, and technologists have bamboozled big publishers for years, but now smart publishers are starting to strike back. With smart data management, they are now able to realize the value of their own audiences—without the networks and exchanges getting the lion’s share of the budget. This has everything to do with leveraging today’s new data management technology to unlock the value of first party data—and more quickly aggregate all available data types to do rapid audience discovery and segmentation.

 The slide we are going to be seeing in 2012, 2013 and beyond will show publishers with a much larger share, as they take control of their own data. Data management technology is not just the sole province of the “Big Five” publishers anymore. Now, even mid-sized publishers can leverage data management technology to discover their audiences, segment them, and create reach extension through lookalike modeling. Instead of going to a network and getting $0.65 for “in-market auto intenders” they are creating their own—and getting $15.00.

Now, that’s a much bigger slice of the advertising pie.

[This post originally appeared in ClickZ on 2/1/12]

Advertising Agencies · Data Management Platform · DMP · Online Media · Platforms · Real Time Bidding (RTB) · Uncategorized

The Data Driven Agency

Three ways you can supercharge your digital media agency with data

Today’s digital media agency has access to enormous amounts of data, but using it effectively is what is going to make the difference between the shops of the future and the also-rans. Delivering data-driven insights is the key to being a 21st century agency. Here are three ways you should be working with data to secure your future:

Visualize it

How much time are you and your colleagues spending collating data, building reports, and formatting spreadsheets and PowerPoint decks for your clients? Most of the agencies I have worked with over the years admit to dedicating an embarrassingly large amount of (highly expensive) time towards these menial tasks. It’s not that getting your clients the data they need is not worth the time, it’s simply that there are now so many automated ways to deliver the data without burning salary.

To paraphrase former agency head and Akamai leader David Kenny, if you are doing things with people that you can be doing with computers, you have already lost. Why spend time formatting Excel spreadsheets and populating PowerPoint report templates with data, when you can be spending salaried employee time selling more services, optimizing campaigns, and delivering great strategy and creative?  Today’s automated ad management solutions and DMPs offer powerful ways to port both audience and ad serving reporting data into a single interface, to get instant access to key metrics such as frequency to conversion, churn rate, and channel attribution.

Ask yourself if the cost of such a system is more than the cost of the time your employees you have been spending building reports—and, ultimately, more than the cost of your eventual demise, should you ignore the changes afoot in your business.

Aggregate and Activate it

Think of all the data you have access to from a digital media standpoint. If you are helping clients execute a digital media campaign, you have traditional serving data from your demand side server, such as DFA. You probably also have engagement data from your rich media ad server. If you have access to your clients’ website pages (or at least tags there), you have site-side data, including conversion event data. If you are using an audience measurement tool, or are doing audience-specific buying through a demand side platform, you also have audience measurement data. Great. What are you doing with all of it? Moreover, what kind of data does your client have that you can help them add to activate the common advertising data types I have just described?

Let’s take the example of an agency using an audience measurement reporting tool, alongside an ad server report. In this case, it is possible that the analyst knows that the highest frequency converters for his travel campaign belong to a popular PRIZM segment, and he may also know that visitors to a popular travel site are three times as likely to engage with his rich media ad creative. Now what? Obviously, the right move is to buy more of the audience segment and double up with guaranteed advertising on the travel site. But what about audience overlap?

How can the advertiser reduce ad waste by ensuring that members of his audience segment that he is securing for as little as $2.00 CPM on exchanges are not overrepresented on the premium site for which he is paying $18.00 CPM? Plus, how many members of that audience are also already registered as customers? If you are not deploying a DMP to aggregate your clients’ CRM (first-party) data alongside the site-side and ad serving (2nd party) data and the purchased (3rd party) data segments, then there is going to lots of duplicated uniques in your audience. Smart data aggregation creates ad activation through waste reduction, lifting conversion rates, while lowering cost per conversion. Getting an effective universal frequency cap across digital channels is very difficult, but every dollar not wasted on duplicate impressions is another dollar that may be spent finding a new audience member. Reducing waste adds reach—and performance, which every client likes.

Compare it

As a digital media agency, you’ve run hundreds, perhaps even thousands of campaigns, producing thousands of data-rich reports for your clients. How much of that knowledge are you leveraging? Although you might know the top travel sites and audience segments to reach “moms of school-age children in-market for a beach vacation,” how readily available is that knowledge? Is it sitting inside your Media Director’s head, or hidden in various documents that don’t talk to one another? How about access to normative campaign data? How quickly can you find out how certain sites performed against similar KPIs without doing hours of research?

Like or not, advertisers want to know how their campaigns are performing against known standards, and it’s gotten a lot more complicated than beating a 0.1% click-through rate lately. Knowing how your last 10 travel campaigns performed—from which guaranteed site buys succeeded, to which audience segments performed, to which creatives elicited the highest CTR—is just step one. Having that data available for quick reference means that every new campaign can start from an advanced performance level, and your media people don’t have to recreate the wheel every time you receive an RFP.

Today’s smart DMPs also feature the ability to leverage your data to an even greater extent, especially for audience buying. Why limit yourself to pre-packaged audience segments that do not include your client’s first-party data? Today’s more advanced DMPs give marketers the ability to create audience segments on the fly, building discrete segments from data that includes available third-party data—but also first-party data, such as registration details, transactional records, and signals from hosted social media listening solutions. It’s the difference between buying from an ad network and creating your own.

Summary

Buying into portals’ site sections was the first phase in the effort to bring contextual and audience relevance to ad buying. Networks followed, offering packaged audiences at scale. Then bidded exchange buying came, offering pre-packaged audience segments at the individual cookie level. Today’s best practices include marrying all available data types to give marketers the ability to create their own targeted buys, and modern data management platforms are helping the largest advertisers automate what they have been doing since the first direct mail piece went out: finding targeted audiences. Leveraging today’s DMP technology can not only help you find those audiences more easily, but help you understand who they are, why they respond, and help you find them again.

Chris O’Hara is head of strategic partnerships for nPario, a DMP with clients that include Yahoo! and Electronic Arts, among others. A frequent contributor to industry publications, this is his first column for The Agency Post. He can be reached through his blog on www.chrisohara.com

[This article originally appeared in The Agency Post on 1/25/12]

Data Management Platform · Demand Side Platform (DSP) · Digital Display · DMP · Media Planning · Platforms · Uncategorized

Know Your Audience

Using Audience Measurement Data to Optimize Digital Display Campaigns

These days, advertising and data platforms are giving marketers a wealth of information that can be used to validate their strategies, and optimize their digital campaigns for better performance. There is a lot of data to sort through—some more useful than others. Sometimes, good campaign optimization comes down to the basics: Understanding who your audience is, and why they are doing what they are doing.

Let’s look at a real life example of a digital display campaign, run through the digital ad agency of a popular mattress retailer. The agency wanted to test new inventory sources for the campaign by running broadly on general interest sites, evaluating the demography of audiences that showed purchase intent, and optimize over the course of the campaign to maximize impact.

A theory being tested was that older audiences, who report more difficulty sleeping than younger demographic groups, would respond more favorably to the retailer’s online display ads. Campaigns were initially skewed to sites that over-indexed against audience composed of 50 and older.

Figure 1: Age of Ad Viewer, by Impressions.

As Figure 1 shows, a bulk of impressions during the discovery portion of the campaign were delivered to visitors aged 46-65 years of age, which was the desired demographic. After analysis of those who viewed or clicked on a display ad, and then went on to purchase, the audience composition was remarkably different. As shown in Figure 2, the bulk of conversions came from those aged 18-45.

Figure 2: Age of Mattress Purchaser (Conversions).

The agency adjusted the ad buy to heavy up on sites that over-indexed for a younger audience, and opted out of buys tailored to the older demographic. As wasted impressions were trimmed down in the overall plan, conversion rates increased dramatically. Testing and validating your instincts with data on an ongoing basis is the key to success in digital display advertising. The mattress retailer, who experienced better sales from older store visitors (offline), found a more responsive younger audience online. Although it seems obvious, having the initial data means being able to smartly allocate marketing capital, and having access to ongoing data means not having to rely on old insights in a changing marketplace.

Another offline theory the mattress retailer sought to validate was the mattress life cycle. After collecting brick and mortar sales data for years, the retailer knew that the average life of a mattress was approximately 7 years, and that the single greatest life event influencing the purchase of a new mattress was moving. Therefore, it made sense to target audiences based on length of residence (>7 years), and target content around buying or renting a new home.

Inventory was bought from a wide range of home-specific and moving sites, and measured using Aperture audience measurement populated with data sets from Experian, IXI financial, V12 demographic, and Nielsen PRIZM data.

 

Figure 3: Length of Residence, by Impressions.
Figure 4: Length of Residence, by Click.


As Figures 3 and 4 amply demonstrate, the mattress retailer was targeting the bulk of impressions towards individuals reporting over seven years residence in a single location, and clicks among that group indexed the highest in aggregate. That data validated the approach of buying into sites with a strong audience of self-reported homeowners. However, a deeper look into audience data revealed a strong distinction between renters and buyers.

Fig 5Comparing Impressions and Conversions by home ownership status.

As noted in Figure 5, although the bulk of impressions in the campaign were served to homeowners, renters were the ones buying the most mattresses. This learning did more than any other data point to drive campaign optimization.

Naturally, the next step in the campaign optimization process was to focus inventory delivery to sites that promised a concentrated audience of home renters. Sites such as ForRent.com, ApartmentGuide.com, and Renters.com were added to the optimization plan.

More insights came as the Aperture data was collected. Despite purporting to have a heavy concentration of renters, two of the more popular sites actually index much higher among homeowners, as shown in Figure 6. It looked as though homeowners that were looking into renting made up the majority audience—a fact that helped the retailer tailor specific messaging to them.

Figure 6: In this example, a media site aimed at renters, over-indexes against current homeowners.

Figure 6: In this example, a media site aimed at renters, over-indexes against current homeowners.

For this particular campaign, the ability for the retailer to validate certain audience assumptions using real demographic data was critical, as well as the ability to leverage the distinction between two types of potential customers: home owners, and renters. Additionally, getting real audience metrics beyond a publisher’s media kit or self-declared audience information enabled the retailer to craft its creative and messaging in a highly specific way that increased conversions.

When it comes to audience validation and campaign optimization, here are three keys:

  • Know Your Data: In today’s technology-driven marketing world, knowing how to leverage the data available to you is critical to both understanding and targeting your audience. Make sure your marketing investment decisions are driven through the analysis and usage of 1st party data, including registration data for demographic modeling; 2nd party data, such as ad server and search data for behavioral modeling; and 3rd party data, such as available audience segments from providers like Nielsen and Datalogix, for audience validation, matching, and lookalike modeling. Data is not just about buying audience segments for targeting; it’s about trying to get a 360-degree view of your ideal customer.
  • Choose the Right DMP: There are DMPs for every marketer, so be careful to choose the right one. Big Data needs call for pure play DMPs that can stitch together highly disparate data sets that include all data types, and make both insights, audience segments, and lookalike modeling available in real-time. Marketers looking to buy from a variety of 3rd party audience segment providers should choose a data marketplace such as Exelate, or be willing to access a more limited number of data sources inside a DSP such as AppNexus.
  • Leverage Audience Measurement: Finally, there is a lot that audience segments can bring to the table in terms of audience insights. Understanding the audience composition of who saw, clicked on, and converted after seeing your campaign gives you the ability to learn about your target customers, their online behaviors, and (most importantly) find more of them. Your DMP should have the ability to marry audience and campaign data to give you a highly granular level view of your best (and worst) performing audience types—down to the creative level.

Learnings from this case study, and other valuable information, can be found in my upcoming “Best Practices in Digital Display Media,” coming in January 2012 from eConsultancy.com.

[This article originally appeared in ClickZ on 1/4/2012]

 

Big Data · Data Management Platform · DMP · Real Time Bidding (RTB)

When Big Data Doesn’t Provide Big Insights

The right DMP solution can be golden for finding audiences.

What big marketers should look for in a next generation data management platform

“Big Data” is all the rage right now, and for a good reason. The other day, I was switching computers, and wanted to move about five gigabytes of photos and videos unto my new laptop, and my largest thumb drive was a measly 1 gig. I ended up getting an 8GB thumb drive for about $8 at the K-Mart in Penn Station. Think about how cheap that is. That’s less than half a cent per song, if you consider the typical 8GB MP3 device can hold about 2,000 high-quality recordings. Two terabyte drives are selling for about $130 from Western Digital. I don’t know about you, but I am not at the point where I need 2TB of data storage, and I hope I never get there. The point is that storing tons and tons of data has gotten very inexpensive, while the accessibility of that data has increased substantially in parallel.

For the modern marketer, that means having access to literally dozens of disparate data sources, each of which cranks out large volumes of data every day. Collecting, understanding, and taking action against those data sets is going to make or break companies from now on. Luckily, an almost endless variety of companies have sprung up to assist agencies and advertisers with the challenge. When it comes to the largest volumes of data, however, there are some highly specific attributes you should consider when selecting a data management platform (DMP).

Collection and Storage: It’s all About Scale, Cost, and Ownership

First of all, before you can do anything with large amounts of data, you need a place to keep it. That place is increasingly becoming “the cloud” (i.e., someone else’s servers), but it can also be your own servers. If you think you have a large of data now, you will be surprised at how much it will grow. As devices like the iPad proliferate, changing the way we find content, even more data will be generated. Companies that have data solutions with the proven ability to scale at low costs will be best able to extract real value out of this data. Make sure to understand how your your DMP scales and what kinds of hardware they use for storage and retrieval.

Speaking of hardware, be on the lookout for companies that formerly sold hardware (servers) getting into the data business so they can sell you more machines. When the data is the “razor,” the servers necessarily become the “blades.” You want a data solution whose architecture enables the easy ingestion of large, new data sets, and one that takes advantage of dynamic cloud provisioning to keep ongoing costs low. Not necessarily a hardware partner.

Additionally, your platform should be able to manage extremely high volumes of data quickly, have an architecture that enables other systems to plug in seamlessly, and whose core functionality enables multi-dimensional analysis of the stored data—at a highly granular level. Your data are going to grow exponentially, so the first rule of data management is making sure that, as your data grows, your ability to query them scales as well. Look for a partner that can deliver on those core attributes, and be wary of partners that have expertise in storing limited data sets. There are a lot of former ad networks out there with a great deal of experience managing common 3rd party data sets from vendors like Nielsen, IXI, and Datalogix. When it comes to basic audience segmentation, there is a need to manage access to those streams. But, if you are planning on capturing and analyzing data that includes CRM and transactional data, social signals, and other large data sets, you should look for a DMP that has experience working with 1st party data as well as 3rd party datasets.

The concept of ownership is also becoming increasingly important in the world of audience data. While the source of data will continue to be distributed, make sure that whether you choose a hosted or a self-hosted model, your data ultimately belongs to you. This allows you to control the policies around historical storage and enables you to use the data across multiple channels.

Consolidation and Insights: Welcome to the (Second) Party

Third party data (in this context, available audience segments for online targeting and measurement) is the stuff that the famous Kawaja logo vomit map was born from. Look at the map, and you are looking at over 250 companies dedicated to using 3rd party data to define and target audiences. A growing number of platforms help marketers analyze, purchase, and deploy that data for targeting (BlueKai, eXelate, Legolas being great examples). Other networks (Lotame, Collective, Turn) have leveraged their proprietary data along with their clients to offer audience management tools that combine their data and 3rd party data to optimize campaigns. Still others (PulsePoint’s Aperture tool being a great example) leverage all kinds of 3rd party data to measure online audiences, so they can be modeled and targeted against.

The key is not having the most 3rd party data, however. Your DMP should be about marrying highly validated 1st party data, and matching it against 3rd party data for the purposes of identifying, anonymizing, and matching third party users. DMPs must be able to consolidate and create as whole of a view of your audience as possible. Your DMP solution must be able to enrich the audience information using second and third party data. Second party data is the data associated with audience outside your network (for example, an ad viewed on a publisher site or search engine). While you must choose the right set of 3rd party providers that provide the best data set about your audience, your DMP must be able to increase reach by ensuring that you can collect information about as many relevant users as possible and through lookalike modeling.

For example, if I am selling cars and I find out that my on-site users who register for a test drive are most closely matched with PRIZM’s “Country Squires” segment,  it is not enough to buy the Nielsen segment. A good DMP enables you to create your own lookalike segment by leveraging that insight—and the tons of data you already have. In other words, the right DMP partner can help you leverage 3rd party data to activate your own (1st party) data.

Make sure your provider leads with management of 1st party data, has experience mining both types of data to produce the types of insights you need for your campaigns, and can get that data quickly. Data management platforms aren’t just about managing gigantic spreadsheets. They are about finding out who your customers are, and building an audience DNA that you can replicate.

Making it Work         

At the end of the day, it’s not just about getting all kind of nifty insights from the data. I mean, it’s big to know that your visitors that were exposed to search and display ads converted at a 16% higher rate, or that your customers have an average of two females in the household. It’s making those insights meaningful.

So, what to look for in a data management platform in terms of actionability? For the large agency or advertiser, the basic functionality has to be creating an audience segment. In other words, when the blend of data in the platform reveals that showing 5 display ads and two SEM ads to a household with 2 women in it creates sales, the platform should be able to seamlessly produce that segment and prepare it for ingestion into a DSP or advertising platform. That means a having an extensible architecture that enables the platform to integrate easily with other systems. Moreover, your DMP should enable you to do a wide range of experimentation with your insights. Marketers often wonder what levers they should pull to create specific results (i.e., if I change my display creative, and increase the frequency cap to X for a given audience segment, how much will conversions increase)? Great DMPs can help built those attribution scenarios, and help marketers visualize results. Deploying specific optimizations in a test environment first means less waste, and more performance. Optimizing in the cloud first is going to become the new standard in marketing.

Final Thoughts

There are a lot of great data management companies out there, some better suited than others when it comes to specific needs. If you are in the market for one, and you have a lot of first party data to manage, following these three rules will lead to success:

  • Go beyond 3rd party data by choosing a platform that enables you to develop deep audience profiles that leverage first and third party data insights. With ubiquitous access to 3rd party data, using your proprietary data stream for differentiation is key.
  • Choose a platform that makes acting on the data easy and effective. “Shiny, sexy” reports are great, but the right DMP should help you take the beautifully presented insights in your UI, and making them work for you.
  • Make sure your platform has an applications layer. DMPs must not only provide the ability to profile your segments, but also assist you with experimentation and attribution–and provide you with ability to easily perform complicated analyses (Churn, and Closed Loop being two great examples). If your platform can’t make the data dance, find another partner.

[This post was originally published in ClickZ on 11/9/11]

 

 

Advertising Agencies · AppNexus · Data Management Platform · demand side platform · Demand Side Platform (DSP) · Digital Display · Digital Media Ecosystem · DMP · Media Buying · Media Planning · Online Media · Real Time Bidding (RTB) · TRAFFIQ · Uncategorized

There’s No App for That

Building the Technology Stack for Next Generation Digital Media Buying and Selling

Last week’s IAB Network and Exchanges conference was full of the usual self-congratulatory “use cases” of byzantine, data-based strategies for squeezing conversions from web-based display banners for direct response campaigns—or, alternatively, helping to drive “branded performance,” based on the listener’s preference. I was sitting next to an attorney from a large media company, tasked with making sense of the ad technology business. “I have to be honest,” she said, “I have been looking at this business for 18 months, and I still don’t understand what you people are talking about half the time…and I’m a smart person.”*

Unfortunately, that is the exact sentiment of many media planners, account managers, and marketing managers confronting the vast array of choices in display advertising. Once they figure out the alphabet soup of DSPs, RTB, and (now) DMPs, they start to wonder if they actually want—or need—the technology in question. Agencies are trying to figure out how to be the gatekeepers, and advise their clients on the best technologies and practices to drive branding and performance, but the work required to string together all of the various options makes earning money difficult. Digital media margins are in the toilet right now, and will remain there until agencies can manage all of these disparate systems with efficiency.

In the ad technology business, there’s an “app” for almost any way one wants to find and buy an audience—and many more applications for getting and understanding performance. Unfortunately, there is no operating system that can host all of these and make them work together seamlessly. The ideal scenario would be a world in which marketers could bring the different media applications they want to use into a single, unified system. Call it a “media dashboard” that would enable an agency to create a campaign, plug in their 3rd party research data, ad server of record, segmentation data licenses, audience measurement/verification providers, and billing system and enjoy access and control from a single interface. Down the road, as more mature APIs become available, the OS would enable marketers to “plug in” their mobile ad providers, video DSPs, and bid management tools for search marketing.

Almost everyone agrees that this is the future of the business. A famous media investment banker recently remarked that “there are some very smart companies out there

Are you developing your ad technology for the wrong system?

building a technology stack” to address these very issues, but wondered whether SAP or Oracle will be first to the party. My opinion is that the IBMs and SAPs of the world will let a smaller company fight through the growing pains, and let the preferred standardization technology come to light, before swooping in. The big boys can afford to be patient—and nobody wants to be the guy who backed Betamax. The question now isn’t Betamax or VHS—or even PC vs. Mac. The question is, what will be the operating system of next generation digital media, who will support it, and can an active “ecosystem” be maintained that enables technology companies to develop smart applications for it?

I think the answer is yes—and that the next 12 months will be critical in determining what companies will fit into the increasingly complex landscape and those that fail to meet the task. Not long ago, it was extremely difficult to buy from a variety of networks and exchanges efficiently. In comes AppNexus, and suddenly every Tom, Dick, and Harry has access to over 800 inventory sources, and a great bid management tools to boot. Their OS for real time bidding creates real efficiency for marketers—especially when they go through the pain of integration on your behalf. I know quite a few AppNexus users—but very few who will work with data segments that are not natively available in the platform.  The next great media technologies are going to be built for integration into specific systems, offer APIs that enable “easy” data export and ingestion, and flexible so that others can customize them for specific needs.

Evolution is natural to the technology business. Networks become “platforms”…data providers become “DMPs.” Technology companies will forever try and stick their hand in the middle of the transaction between the demand and the supply side, and shave off a sliver of the pie. But, eventually, evolution becomes “revolution” and the game changes for everyone. We are about to find out who has the capital, talent, and vision to devise the next generation operating system for digital media. That system is going to be the one that every company has to develop an “app” for and support, and that system is going to shape the way digital media is bought and sold for a very long time.

As an ad technology company, it’s time to start figuring out how your technology will fit into the larger puzzle if such an OS becomes standard. Is your technology built for an open system, or does your technology (and, more importantly, business model) only thrive in a closed environment? There are a lot of “platforms” out there, but eventually there will only be one operating system. I think there are a lot of really awesome “apps” out there waiting to be plugged into this new operating system, which would benefit from standardization and an installed base of users.

There’s definitely an “app for that.” We are just waiting for the OS.

*That sentiment was also expressed wonderfully in Doug Weaver’s amazing keynote presentation which he was kind enough to make available this morning on iMedia.

[This post appearred on 5/23/11 in Business Insider]