Data Management Platform

CPG goes DMP

 

a95d8358e93c679917a7785d9cce4a35
I often wonder how Wayne Gretzky feels when he looks at this photograph. Maybe he’s like, “man, that blazer was totally boss.”

If you think about the companies with perhaps least amount of consumer data, you would automatically think about consumer packaged goods (CPG) manufacturers. Hardly anybody registers for their website or joins their loyalty clubs; moms don’t flock to their branded diaper sites; and they are at arms-length from any valuable transaction data (store sales) until well after the fact. So, with little registration, website, or offline sales data, why are so many large CPG firms licensing an expensive first-party data management platform?

 

While CPG companies will never have the vast amounts of point-of-sale, loyalty-card, app, and website data that a big box retailer might have, they do spend a ton of dough on media. And, as we all know, with large media expenditures come tons of waste. Combine this with the increasingly large investment and influence that activist investors and private equity companies have in CPG, and you can see where this leads. PE companies have installed zero-based budgeting that forces CPG concerns to rationalize every penny of the marketing budget—which, until lately, has been subject to the Wannamaker Rule (“I know half of my budget is working, but not which half”). Enter the DMP for measurement and global frequency control, cutting off and reallocating potentially millions of dollars in “long tail” spending. Now, the data that the CPG marketer actually has in abundance (media exposure data), can be leveraged to the hilt.

This first and most obvious CPG use case has been discussed extensively in past articles. But there is much more to data management for CPG companies. Here are just a few tactics big consumer marketers have written into their data-driven playbooks:

 

The Move to Purchase-Based Targeting (PBT)

Marketers have come a long way from demographic targeting. Yes, gender, age and income are all reliable proxies for finding those “household CEOs,” but we live in complicated times and “woman, aged 25-54, with 2 children in household” is still a fairly broad way to target media in 2016. Today, men are increasingly as likely to go grocery shopping on a Thursday night. Marketers saw this and shifted more budget to behavioral, psychographic, and contextual targeting—but finding cereal buyers using proxies such as site visitation sharpened the targeting arrow only slightly more than demography.

Packaged goods marketers have long understood the value of past purchases (loyalty cards and coupons), but until the emergence of data management technologies, have struggled to activate audiences based on such data. Now, big marketers can look at online coupon redemption or build special store purchase segments (Datalogix, Nieslen Catalina, News America Marketing) and create high value purchase-based segments. The problem? Such seed segments are small, and must be modeled to achieve scale. Also, by the time the store sales data comes in, it’s often far too late to optimize a media plan. That said, CPG marketers are finding that product purchasers share key data attributes that reveal much about their household composition, behavior, and—most interestingly—affinity for a company’s other products. It may not seem obvious that a shopping basket contains diapers and beer—until you understand that Mom sent Dad out to the store to pick up some Huggies, and he took the opportunity to grab a cold six-pack of Bud Light. These insights are shaping modern digital audience segmentation strategy, and those tactics are becoming more and more automated through the use of algorithmic modeling and machine-learning. CPG has seen the future, and it is using PBT to increase relevant reach.

Optimizing Category Reach

CPG marketers are constantly thinking about how to grow the amount of product they sell, and those thoughts typically vary between focusing on folks who are immensely loyal (“heavy” category buyers) versus those who infrequently purchase (“light” or “medium” category buyers). Who to target? It’s an interesting question, and one answered more decisively with purchase-based sales data.

Take the large global soda company as an example. Their average amount of colas their customer consumes is 15 a year, but that is an immensely deceptive number. The truth is that the company has a good amount of “power users” who drink 900 colas a year (two and a half per day), and a lot of people who may only drink 2-3 colas during the entire year. Using the age-old “80/20 Rule” as a guideline, you would perhaps be inclined to focus most of the marketing budget on the 20% of users who supposedly make up 80% of sales volume. However, closer examination reveals that heavy category buyers may only be driving as little as 50% of total purchase volume. So, the marketer’s quandary is, “Do I try and sell the heavy buyer his 901st cola, or do I try and get the light buyer to double his purchase from two to four colas a year?”

Leveraging data helps CPG companies not have to decide. Increasingly, companies are adopting frequency approaches that identify the right amount of messaging to nurture the heavy users (maybe 2-3 messages per user, per month) and bring light buyers to higher levels of purchase consideration (up to 20 messages per month). Moreover, by using DMP technology to segment these buyers based on their category membership, creatives can be adjusted based on the audience. Heavy buyers get messages that reinforce why the love the brand (“share the love”), and light buyers can receive more convincing messages (“tastes better”).

Increasing Lift through Cross-Channel Messaging

CPG marketers have some highly evolved models that show just how much lift a working media dollar has on sales, and they use this guide to decision on media investment by both channel and partner. With the power of DMPs for cross-channel measurement, CPG companies are finally able to apply even small insights they can to tweak sales lift.

What if the data reveal that a 50% mixture of equity and direct response ad creatives lifts coupon downloads by 200%? In other words, instead of just showing “Corn Flakes are Yummy” ads, you mixed in a few “Buy Flakes now at Kroger and save!” creatives afterwards, and you saw a huge impact on your display performance? Sadly, this simple insight was not available before data management platforms corralled cross-channel spending and associated it with an individual, but now these small insights are adding up to appreciable sales lift.

In another example, a large CPG company sees massive lift in in-store coupon redemptions by running branded display ads on desktop all throughout the week—but giving a “mobile nudge” on the smartphone on Friday night when it’s time to fill the pantry. This cross-channel call-to-action has seen real results, and only involves grabbing a brand-favorable consumer’s attention on another device to create a big impact. Again, a simple tactic—but also impossible without the power of a DMP.

CPG marketers have been able to achieve a ton of progress by working with relatively sparse amounts of data. What can you do with yours?

 

 

 

Data Management Platform

Data Triangulation: How Second-Party Data Will Eat The Digital World

 

bearshark
This is the most bad ass illustration I have ever seen. Who painted this? I mean, whoa. 

Marketers are getting frustrated with spending up to 60% of their working media dollars to fund intermediaries between themselves and their publishing partners. By the time a marketer pays his agency, trading desk, exchange, third-party data provider, and subsidizes the publisher’s ad serving stack, dollars turn into dimes. Marketers want less fraud, more people, less ad tech, and to put more media dollars to work to drive performance. Quality publishers, who for so long sacrificed control for access to an always-on stream of programmatic cash, are now seeing balance return, as shady sources of inventory leave the ecosystem and start to create scarcity for premium supply.

 

Publishers with desired audiences are starting to leverage hacks like “header bidding” and private marketplaces to get more control and capture more revenue from transactions. But they are also starting to look at data-only transactions among trusted demand-side partners. Now that marketers are catching up with DMP technology, securely sharing audiences becomes possible, opening up a new era where “second party” data is poised to reign supreme. Before we talk about how that happens, let’s first define some data terms:

A Primer

First-party data is proprietary data that marketers and publishers have collected – with permission, of course – and, therefore, own. It can be cookies collected from a site visit, offline data onboarded into addressable IDs and even data from marketing campaigns. Second-party data is simply someone else’s first-party data. Second-party data gets created any time two companies strike up a deal for data that is not publicly available. The most common use case is that of a marketer – say a big airline –getting access to data for a publisher’s frequent travelers. Big Airline might say to Huge News Site with business travelers, “Let’s user match, so every time I see one of my frequent flyers on your site, I can serve him an ad.” Huge News Site may decide to allow Big Airline to target its users wherever they are found (a “bring your own data” deal) or make such a deal incumbent upon buying media. Either way, Big Airline now has tons of really valuable Huge News Site reader data available in its data-management platform (DMP) for modeling, analysis and targeting.

Despite the much heralded death or merely diminution of third-party data, it is still a staple of addressable media buying. This is data that is syndicated and made available for anyone to buy. This data could describe user behavior (Polk “auto intenders” of various stripes) or bucket people into interesting addressable segments based on their life circumstances (Nieslen “Suburban Strivers”), describe a user’s income level (Acxiom or Experian) or tell you where a user likes to go via location data (PlaceIQ or Foursquare). Most demand-side platforms (DSPs) make a wide variety of this data available within their platforms for targeting, and DMPs enable users to leverage third-party data for segment creation – usually allowing free usage for analytics and modeling purposes and getting paid upon successful activation. Data Quality And Scale So, which kind of data is the best? When asked that question by a marketer, the right question is inevitably, “all of it.” But, since that’s an annoying answer, let’s talk about the relative scale and value of each type of data. It’s easily visualized by this wonderfully over-simplified triangle:

TRIANGLE OF DATA

First-party data is the most limited in scope, yet the most powerful. For marketers –especially big CPG marketers who don’t get a lot of site traffic – first-party data is incredibly sparse but is still the absolute most valuable signal to use for modeling. Marketers can analyze first-party data attributes to understand what traits and behaviors consumers have in common and expand their reach using second- or third-party data. Retail and ecommerce players are more fortunate. A Big Box Store has first-party data out the wazoo: loyalty card data, point-of-sale system data, app data, website registration data, site visit data and maybe even credit card data if it owns and operates a finance arm. It can leverage a DMP to understand how media exposure drove a store visit, where customers were in the store (beacons!), what was purchased, how many coupons were remitted and whether or not they researched their purchase on the site. Talk about getting “closed loop” sales attribution. The power of first-party data is truly amazing.

The biggest problem with third-party data is that all of my competitors have it. In programmatic marketing, that means both Ford and Chevy are likely bidding on the same “auto-intender” and driving prices up. The other problem is that I don’t know how the data was created. What attributes went into deciding whether or not this “auto intender” is truly in-market for a car? There are no real rules about this stuff. A guy who read the word “car” in an article might be an “auto-intender” just as someone who looked a four-door sedans three times in the last 30 days on reputable auto sites. Quality varies. That being said, there is huge value in having third-party data at your disposal. Ginormous Music App, for example, has built a service that is essentially a DMP for music; it knows how to break down every song, assign very granular attributes to it and delivers highly customized listening experiences for free and paid users. Those users are highly engaged, have demonstrated a willingness to buy premium services and are, by virtue of their mobile device, easily found at precise geolocations. Yet, for all of that, the value to a marketer of a Maroon Five segment is rather small. Everyone likes Maroon Five, from grandmothers to tweens to Dads. A Maroon Five segment provides little value to an advertiser. Yet, if Ginormous Music App could push its app-based user data (IDFAs) into the cookie space and find a user match, it could effectively use third-party data to understand the income, behavior and general profile of many Maroon Five fans. And that’s what their advertisers like to buy. That’s pretty damn valuable.

So, how about “second-party” data? These are the “frequent business travelers” on Huge News Site and the “Mitsubishi intenders” on Large Auto Site. These are real users, with true demonstrated intent and behavior that has been validated on real properties. One of the most valuable things about audiences built on second-party data is that there is usually transparency regarding how those users found their way into a segment.

The ironic and kind of beautiful thing about the emergence of second-party data is that it is most often merely a connection to a premium publisher’s users. However, it can be uncoupled from a publisher’s media sales practice. Marketers, increasingly sick of all the fraud and junk in the programmatic ecosystem are turning toward second-party data to access the same audiences they bought heavily in print 30 years ago. This time, however, they are starting to get both the quality – and the quantitative results – they were looking for. On the flip side, quality publishers are starting to understand that, when offered in a strict, policy-controlled environment that protects their largest asset – audience data – they can make way more money with data deals than media deals.

Put simply, second-party data is heralding a return to the good old days when big marketers depended on relationships with big publishers as the stewards of audiences, and they created deep, direct relationships to ensure an ongoing value exchange. Today, that exchange increasingly happens through web-based software rather than martini lunches.

[This article originally appeared in AdExchanger on 1/25/16]

Big Data · Data Management Platform

Leapfrogging the Lumascape

 

Leapfrogging-baby-elephants-at-chester-zoo-by-Mike-Shaw
Trying to make money with programmatic advertising is as easy as….

 

Marketers have always craved access to quality audience at scale. That was once as easy as scheduling buys on the top three broadcast networks and buying full-page ads in national newspapers. Today, the world is more complicated, with attention shifting into a splintered digital universe of thousands of channels across multiple media types.

Ad tech companies have tried to corral a massively expanding world of inventory in ad exchanges, along with the means to bid inside them. This “programmatic” world of inventory procurement is deeply flawed, yet still the best thing we have at the moment.

It’s flawed because it mostly offers access to commoditized “display” ad units of dubious value and struggles to deliver real audiences, rather than robots. But it’s also good because we have taken the first steps past a ridiculous paradigm of buying media through relationships and fax machines, while starting to bring an analytical discipline to media investment that is based on measurement.

So, as we sled the downward slope of the programmatic buying Hype Cycle, we are starting to see some new trends in inventory procurement – namely, a strategy that involves replacing some or all of the licensed programmatic architecture, as well a growing reliance on one’s own data.

But first, before we get into the nuts and bolts of how that works, some history:

The Monster We Created

After convincing ourselves of the lack of scalability in the direct model, where we would call an ad rep, we have set up a lot of distance between a marketer and their desired audience.

chrisoharachart1-921

Imagine I am a cereal manufacturer and have discovered through media mix modeling that digital moms on Meredith sites drive a lot of offline purchases. They are the “household CEOs” that drive grocery store purchasing, try new things and are influential among their peer group, in terms of recommending new products. In today’s new media procurement paradigm, there are many “friends” standing between my target and me:

  • Media agency: This is a must-have, unless marketers want to add another 100 people to their headcount with an expertise in media, but this adds 5% to 10% in costs to media buys.
  • Trading desk: Although many marketers are starting to take this functionality in-house, whether you trade internally or leverage an agency trading desk, you can expect 10% to 15% of media costs to go to the personnel needed to run this type of operation.
  • Demand-side platform (DSP): Don’t forget about the technology. A 15% bid reduction fee is usually required to leverage the smart tools necessary to find your inventory at scale across exchanges.
  • Private marketplace: But wait! We use private marketplaces to make exclusive deals among a small pool of preferred vendors. Yes, but they operate inside DSPs and carry transactional fees that can add between 5% and 10% extra.
  • Third-party data: You can’t target effectively without adding a nice layer of audience data on your buy, but expect to pay at least $1 CPM for the most basic demographic targeting – a significant percentage of cost even on premium buys.
  • Exchanges: Maybe you pay for this via your DSP, but someone is paying for a seat on an ad exchange and that cost is passed through a provider, which can add another several percentage points.
  • Supply-side platform (SSP): It’s not just the demand side that needs to leverage expensive technology to navigate the new world of digital media. Publishers pay up to 15% in fees to deploy SSPs, a smart inventory management technology to help them manage their “daisy chain” of networks and channel sales providers to get the best yield. This is baked into the media cost and passed along to the advertiser.
  • Ad server: Finally, the publisher pays a fee to get the ad delivered to the site. It is a somewhat small price, but one that is passed along to the advertiser, usually baked in to the media cost.

This is essentially the middle of a crowded LUMAscape, a bunch of different disintermediating technologies that stand between an advertiser and the publisher. Marketers pay for everything I just described. They may not license the publisher’s SSP for them, but they are subsidizing it. After running this gauntlet, marketers with $10 to spend on “cereal moms” end up with much less than half in media value – the amount the publisher ends up with after the disintermediation takes place. This can be anywhere from 10% to 40% of the working media spend.

That’s probably the biggest problem in ad tech right now.

We’ve essentially created a layer of technology so gigantic in between marketers and audiences, that 60% to 70% of media investment dollars land up in venture-funded technology companies’ hands, rather than the media owner creating the perceived value. How do we change that paradigm?

chrisoharachart2-921

Leapfrogging the Middleware

Data management technology is increasingly replacing some of the middleware in this procurement equation, effectively writing the third chapter in the saga we know as programmatic direct.

Here is a bit of background.

What I call “Programmatic Direct 1.0” was the short-lived period in which companies leveraging the DoubleClick for Publishers (DFP) ad-serving API built static marketplaces of premium inventory.

For example, a premium publisher like Forbes might decide to place a chunk of 500,000 home page impressions in a marketplace at a $15 CPM. Buyers could go into an interface, transact directly with the publisher and secure the inventory. The problem that inventory owners had a hard time valuing their future inventory and buyers weren’t keen to log into yet another platform to buy media. This phase effectively ended with the Rubicon Project buying several leaders in the space, ShinyAds and iSocket, and AdSlot taking over workflow automation software provider Facilitate Media. Suddenly, “programmatic direct” platforms started to live inside systems where media planners actually bought things.

Programmatic direct’s second act (2.0) is prevalent today. Companies use deal IDs or build PMPs within real-time systems and exchanges to have more control over procurement than what is available in an auction environment. Sellers can set prices and buyers can secure rights to inventory at a set, transparent cost. This works pretty well, but comes with the same gigantic stack of providers as before and includes additional transaction fees. This is akin to making a deal to buy a house directly from the owner, but agreeing to pay the real estate broker fee anyway. The thing about programmatic direct transactions is that they are fundamentally different than RTB because they don’t have to take place in “real time,” nor do they involve bidding. A brand-new set of pipes is required.

“Programmatic direct 3.0” – or whatever we decide to call it – looks a bit different. Let’s say the big cereal company uses a data-management platform (DMP) to collect its first-party data and creates segments of users from both offline user attributes and page-level attributes from site visitation behavior. The marketers have created a universal ID (UID) for every user. Let’s imagine they discovered 200,000 were females, 24 to 40 years old, living in two-child households with income greater than $150,000 and interested in health and fitness. Great.

Now imagine that a huge women’s interest site deployed its own first-party DMP and collected similar attributes about their users, who were assigned UIDs. If the marketer and publisher have the same enterprise data architecture, they could match their users, make a deal and discover that there’s an overlap of 125,000 of users on the site. Maybe the marketer agrees to spend $7 CPM to target those users, along with users who are statistically similar, every time they are seen on the site for November.

The DMP can push that segment directly into the publisher’s DFP. No trading desk fees, DSP fees, third-party data costs or SSPs involved. The same is true for a variety of companies that have built header bidding solutions, although they see less data than first-party DMPs.

With this 3.0 approach, most of the marketer’s $7 is spent on media, rather than a basket of technologies, and the publisher gets to keep quite a bit of that revenue.

Sounds like a good deal.

Follow Chris O’Hara (@chrisohara) and AdExchanger (@adexchanger) on Twitter. 

Data Management Platform

Is 2015 the Year of Programmatic Branding?

BrandingPortfolio

With companies like Kraft and Kellogg’s starting to leverage the programmatic pipes for equity advertising, we are starting to hear a lot of buzz about the potential for “programmatic branding,” or leveraging ad tech pipes to drive upper-funnel consumer engagement. It makes sense. Combine 20 years in online infrastructure investment with rapidly shifting consumer attention from linear to digital channels, and you have the perfect environment to test whether or not digital advertising can create “awareness” and “interest,” the first two pieces of the age old “AIDA” funnel.

The answer, put simply, is yes.

Online reach is considerably less expensive than linear reach, and we are starting to have the ability to reliably measure how that brand engagement is generated. Marketers want an “always-on” stream of equity advertising that comes with measurement—but they also need it. With attention rapidly shifting from traditional channels, investments in linear television are starting to return fewer sales. But most marketers are just starting to gain the digital competency to make programmatic branding a reality.

That competency is called data management—the ability to segment, activate, and analyze consumer audiences in a reliable way at scale. Why is that so?

The most fundamental problem with digital branding is that it is truly a one-to-one marketing exercise. If we dream of the “right message, right person, right time,” then matching a user with her devices is table stakes for programmatic branding. How do I know that Sally Smith on desktop is the same as Sally Smith on tablet? Cross-device identity management (CDIM or, alternatively, CDUI) is the key. Device IDs must be mapped to cookies, other mobile identifiers, and Safari browser signals in order to get a sense of who’s who. Once you unlock user identity, many amazing things become possible.

Global Frequency Capping

One of the reasons programmatic branding has yet to gain serious ground with marketers is because of waste. This is both real (lots of wasted impressions due to invisible ads or robotic traffic) and perceived (how many impressions are ineffective due to frequency issues). The former problem is getting solved by smart technology, and already somewhat mitigated by market pricing. But the latter problem is solvable with data management. Assuming the marketer understands the ideal effective frequency of impressions per channel, or on a global basis, a DMP can manage how many impressions an individual sees by controlling segment membership in various platforms. Let’s say the ideal frequency for cereal advertising aimed at Moms is 30 per day across channels. The advertiser knows less than 30 impressions lessens effectiveness—and over 30 impressions has negligible impact. Advertisers using multiple channels (direct-to-publisher, plus a mobile, video, and display DSPs) are likely over serving impressions in each channel, and maybe underserving in key channels like video. Connecting user identity helps control global frequency, and can save literally millions of dollars, while optimizing the effectiveness of cross-channel advertising.

Sequential Messaging

If “right person” technology is enabled as above, then it makes sense to try and get to “right place and right time.” Data management can enable this Holy Grail of branding, helping marketers create relevance for consumers as they embark on the customer journey. What brand marketers have dreamed of is now possible, and starting to happen. Dad, in the auto-intender bucket, gets exposed to a 15 second pre-roll ad before logging into his newspaper subscription on his tablet in the morning; gets the message reinforced by more equity display ads in the afternoon at work; and, while checking messages on his mobile phone on the way home, gets an offer for $500 off with a qualified test drive. After he hits the dealership and checks in via the CRM system, he receives an e-mail thanking him for his visit and reminding him of the $500 coupon he earned. These tactics are not possible without tying both user identity and systems together. Doing so not only enables sequential messaging, but also the ability to test and measure different approaches (A/B testing).

Cross Channel Attribution

How about attribution? It’s impossible to perform cross-channel attribution without knowing who saw what ad. At the end of the day, it’s really about the insights. Proctor and Gamble is famous for spending millions of dollars every year to understand the “moment of truth,” or why people choose Tide over Surf detergent. Although they know consumer segmentation and behavior better than anyone, even the biggest brand marketers struggle to gain quality insights from digital channels. Data management is starting to make a more reliable view possible. Brand advertising is just another form of investment. Money is the input, and the output is sales and—as important—data on what drove those sales. In the past, brand marketers were reliant upon panel-based measurement to judge campaign effectiveness. Now, data management helps brands understand which channels drove results—and how each contributed. It is early days for truly reliable cross-channel attribution modeling, but we are finally starting to see the death of the “last click” model. Smart marketers are using data to author their own flexible attribution models, making sure all channels involved receive variable credit for driving the final action. In the near future, machine learning will help drive dynamic models, which flex over time as new signals are acquired. We will then start to see just how effective (or not) tactics like standard display advertising are for driving upper funnel engagement.

So, is 2015 the year for programmatic branding? For a select group of marketers that are leveraging data management to enable the best practices outlines above, yes. The more accurately marketers can map online user identity and understand results, the more investment will flow from linear to addressable channels.

Data Management Platform

Remarks at ICOM 2015 in San Sebastien

I-COM Global Summit: Panel Discussion on Leveraging Big Data to take Programmatic to the Next Level – Chris O’Hara, Krux Digital, Eric Picard, Mediamath & Tom Simpson, MediaQuark

Leveraging big dataChris O’Hara, VP Strategic Accounts, Krux Digital, USA, Eric Picard, VP Strategic Partnerships, Mediamath, UK and Tom Simpson, CEO, MediaQuark, Singapore were speakers and David Smith, CEO & Founder, Mediasmith, USA was moderator in Leveraging Big Data to take Programmatic to the Next Level. This discussion had no presentation.

Data Management Platform

The Five (New) Things to Expect from a DMP

The-5-Pillars-of-Health

In early 2012, when data management technology was somewhat nascent, I wrote about “the five things to expect from a DMP.” They were: To unlock the power of one’s first party data; decrease reliance upon third party data; generate unique audience insights; use data to audience power new channels; and create efficiency. A little over three years later, those things still continue to drive interest in DMP technology—and great value for both publishers and marketers.

The “table-stakes” functionality of DMPs—segmentation, lookalike modeling, targeting, and analytics—continue to resonate. Even the least advanced DMPs have those abilities, and this is what people who buy DMP software should expect from any system. Unfortunately, there are now dozens of “platforms” that claim DMP technology. Some are legitimate players, born from the ground up to be “first-party” DMPs. Some have been created as “lightweight” DMPs to collect and distribute cookies for display advertising. And still others are legacy tag management or network platforms that have bolted on DMP functionality as they work towards a fuller “stack” solution that marketers say they want.

Writing this article again, three years later, I would still encourage software buyers to evaluate their DMP choice on the ability of their partner to meet the above-listed criteria. But, there has been so much nuance and development over the last several years. Therefore, additional selection criteria present themselves if one is expected to make a reasonably informed choice in DMP selection going forward.

Here’s what the modern DMP consumer should be looking out for:

  • Lookback: Three years ago I talked about “lookback windows” in the context of giving publishers the ability to attribute future conversion events to ads shown previously on their site. That is still a compelling publisher user case. What “lookback windows” really refer to is whether or not your DMP can capture 100% of the raw, log-level user event data—and store it. This necessitates an open taxonomy (because “you don’t know what you don’t know,”) and also the ability to store tons of data and make it accessible quickly. This is considered to be complete data architecture. Many DMPs operate with a rigid, defined taxonomy and only collect segment IDs—not the underlying data. That’s a problem for businesses that need to move fast and activate new segments opportunistically. Ask how—and for how long—your DMP stores data.
  • Onboarding: Lots of DMPs claim to have the ability to easily ingest CRM and other offline data and match it to cookies, but the truth is everyone depends on a limited set of “onboarding” vendors to provide the matches. That’s fine, but there are some nuances and subtleties involved in the process by which offline data enters the online identity space (hashing). DMPs should enable seamless connection to all three major onboarding providers, the ability to select the methodology by which offline identity is matched to online, and also be able to automatically choose which onboarding partner is right for each identity. Ask how each DMP you evaluate works with each vendor, what kind of match rates you can expect, and how each stores persistent user identity to insure better matches over time.
  • Measurement: Let’s face it, the ability to tweak programmatic audience delivery to online video viewability numbers up a few percentage points is great, but nothing moves the needle like linear television. Marketers spend a ton of money there, and will continue to do so for the foreseeable future—all the while moving incremental percentages of their budget into the digital channels where folks are spending an increasing amount of time. But, they are never really going to go full throttle with digital until they can reconcile reach and frequency across channels—and those channels must include linear! Your DMP should be able to handle overlap reporting, light attribution, and cross-channel media performance—but it should also start making some highly informed guesses about how linear audiences map to digital ones, in order to enable true attribution and media mix models. Ask how your DMP is positioned to tie the linear and digital strings together from a measurement perspective.
  • CDIM: Three years ago, we were still waiting for the “year of mobile” to occur, so “cross device identity management” was still largely pre-funded slideware on some entrepreneur’s computer. Jump to today, and “CDIM” and “CDUI” are at the tip of every ad tech tongue! As more and more people move from device to device—almost none of which support the traditional cookie as an identifier—marketers and publishers desperately need to map devices to people. It’s the only way to deliver the fabled “360 degree view” of the user. Ask your DMP vendor how they are prepared to deliver deterministic matches and, more importantly, how they reconcile identity without seeing a user logging in across devices. Doing great probabilistic matching necessitates not only strong algorithms but, more importantly, scale of users which breeds precision models. What is the size of their “truth set” of user data with which to probabilistically determine user identity? The quality and scale of that data will determine your choice.
  • Data Governance: I think the biggest question to ask a potential DMP vendor is their philosophy on data ownership. For both marketers and publishers, audience data is likely one of their top three assets. Trusting such data to a technology vendor is not something to be considered lightly. How is that data stored? What are the policy controls available to help you share that data with trusted partners? What about privacy and governance? How can my platform help me activate data in different places, where different rules about PII and data collection and storage apply? Knowing the answers to these before you buy can save lots of heartache (and legal fees) later. More importantly, how independent is your data? Is your partner also in the business of selling media or data? That can create some conflicts of interest—especially if your data might be valuable to a competitor. Finally, what if you want your data back? You have the right to get it out quickly, and in a useable format.

The bad news is that choosing a DMP isn’t any easier than it was three years ago. It’s a lot more complex, and you really need to dig in deeply to understand the very small nuances between platforms that appear, on the surface, to be very much the same. The good news is that there is a great deal of selection available, and some very high quality vendors to choose from. Take your time, put your vendors through a very rigorous process that includes asking the questions outlined above, and choose wisely!

[This post originally appeared in the EConsultancy blog on 5.11.15]

Data Management Platform

CDIM is Table Stakes in the Data Management Wars

IdentityCrisisA recent analyst report made an astute observation that all marketers should consider: It’s not about “digital marketing” anymore – it’s about marketing in a digital world. The nuance there is subtle, but the underlying truth is huge. The world has changed for marketers, and it’s more complicated than ever.

Most consumers spend more time on web-connected devices than television, creating a fragmented media landscape where attention is divided by multiple devices and thousands of addressable media outlets. For marketers, the old “AIDA” (attention, interest, desire and action) funnel persists, but fails in the face of the connected consumer.

When television, print and radio dominated, moving a consumer from product awareness to purchase had a fairly straightforward playbook. Today’s always-on, connected consumer is on a “customer journey,” interacting with a social media, review sites, pricing guides, blogs and chatting with friends to decide everything from small supermarket purchases to big investments like a new house or car.

Marketers want to be in the stream of the connected consumer and at key touch points on the customer journey. But, in order to understand the journey and be part of it, they must be able to map people across their devices. This is starting to be known as cross-device identity management (CDIM), and it is at the core of data-driven marketing.

In short, identity lies at the heart of successful people data activation.

Until very recently, managing online identity was largely about matching a customer’s online cookie with other cookies and CRM data, in order to ensure the desktop computer user was aligned with her digital footprint. Today, the identity landscape is highly varied, necessitating matching ID signals from several different browsers, device IDs from mobile phones and tablets, IDs from streaming devices and video game consoles and mobile app SDKs.

Matching a single user across their various connected devices is a challenge. Matching millions of users across multiple millions of devices is both a big data and data science challenge.

Real one-to-one marketing is only possible when the second party – the customer – is properly identified. This can be done using deterministic data, or information people volunteer about themselves, in a probabilistic manner, where the marketer guesses who the person is based on certain behavioral patterns and signals. Most digital marketing companies that offer identity management solutions take what data they have and use a proprietary algorithm to try and map device signals to users.

The effectiveness of device identity algorithms depends on two factors: the quality of the underlying deterministic data – the “truth set” – and its scale.

Data Quality Matters

There is data, and then there is data. The old software axiom of “garbage in, garbage out” certainly applies to cross-device user identity. Truly valuable deterministic data include things like age, gender and income data. In order to get such data, web publishers must offer their visitors a great deal of value and be trusted to hold such information securely. Therefore, large, trusted publishers – often with subscription paywalls – are able to collect highly valuable first-party user data.

Part of the quality equation also relates to the data’s ability to unlock cross-device signals. Does the site have users that are logged in across desktop, mobile phone and tablet? If so, those signals can be aggregated to determine that Sally Smith is the same person using several different devices. Publishers like The Wall Street Journal and The New York Times meet these criteria.

Scale Is Critical

In order to drive the best probabilistic user matches, algorithms need huge sets of data to learn from. In large data sets, even small statistical variances can yield surprising insights when tested repeatedly. The larger the set of deterministic data –the “truth” of identity – the better the machine is able to establish probability. A platform seeing several million unique users and their behavioral and technographic signatures may find similarities, but seeing billions of users will yield the minuscule differences that unlock the identity puzzle. Scale breeds precision, and precision counts when it comes to user identity.

As digital lives evolve beyond a few devices into more connected “things,” having a connected view of an individual is a top priority for marketers that want to enable the one-to-one relationship with consumers. Reliably mapping identity across devices opens up several possibilities.

Global Frequency Management: Marketers that leverage multiple execution platforms, including search, email, display, video and mobile, have the ability to limit frequency in each platform. That same user, however, looks like five different people without centralized identity management.

Many marketers don’t understand what ideal message frequency looks like at the start of a campaign, and most are serving ads far above the optimal effective frequency, resulting in large scale waste. Data management platforms can control segment membership across many different execution platforms and effectively cap user views at a “global” level, ensuring the user isn’t over-served in one channel and underserved in another.

Sequential Messaging: Another benefit of cross-device identity is that a user can be targeted with different ads based on where they are in the consumer journey. Knowing where a consumer is in an established conversion path or funnel is a critical part of creative decisioning. Optimizing the delivery of cross-channel messages at scale is what separates tactical digital marketers and enterprise-class digital companies that put people data at the heart of everything they do.

Customer Journey Modeling: Without connecting user identity in a centralized platform, understanding how disparate channels drive purchase intent is impossible. Today’s models bear the legacy of desktop performance metrics, such as last click, or have been engineered to favor display tactics, including first view. The true view of performance must involve all addressable channels, and even consider linear media investment that lacks deterministic data. This is challenging but all but impossible without cross-device identity management in place.

The ubiquity of personal technology has transformed today’s consumers into “digital natives” who seamlessly switch between devices, controlling the way they transmit and receive information. Marketers and publishers alike must adapt to a new reality that puts them in control of how editorial and advertising content is accessed. Delivering the right consumer experience is the new battleground for CMOs. Unlocking identity is the first step in winning the war.

[This post originally appeared in AdExchanger on 3.16.15]

Advertising Agencies · Data Management Platform

The Role of the Agency in Data Management

Data management may seem like a high hurdle to jump for some agencies.
Data management may seem like a high hurdle to jump for some agencies.

Twenty years after the first banner ad, the programmatic media era has firmly taken hold.

The holy grail for marketers is a map to the “consumer journey,” a circuitous route filled with multiple addressable customer touchpoints. With consumers spending more of their time on mobile devices – and interacting with brands like never before through social channels, review sites, pricing comparison sites and apps – how can marketers influence customers everywhere they encounter a brand?

It’s a tough nut to crack, but starting to become an achievable reality to companies dedicated to collecting, understanding and activating their data. Marketers are starting to turn toward data-management platforms (DMP), which help them connect people with their various devices, develop granular audience segments, gain valuable insights and integrate with various platforms where they can activate that data. In addition to technology, marketers also have to configure their entire enterprises to align with the new data-driven realities on the ground.

The question is: Where do marketers turn for help with this challenging, enterprise-level transition?

Many argue that agencies 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. But should data-management software selection and integration be the sole province of the Accentures and IBMs of the world, or is there room for agencies to play?

For lots of software companies, having an agency in between an advertiser and their marketing platform sounds like a problem to overcome, rather than a solution. Many ad tech sellers out there have lamented the process of the dreaded agency “lunch and learn” to develop a software capability “point of view” for a big client.

Yet there are highly compelling ways agencies add value to the software selection process. The best agencies insert themselves into the data conversation and use their media and creative expertise to influence what DMPs marketers choose, as well as their role within the managed stack.

From Digital To Enterprise

It makes perfect sense that agencies are involved with data management. The first intersection of data and media added the “targeting” column to the digital RFP. Agencies have started to evolve beyond the Excel-based media planning process to start their plans with an audience persona that is developed in conjunction with their clients. Today, plans begin with audience data applied to as many channels as are reachable. Audience data has moved beyond digital to become universal.

Agencies have also been at the tip of the spear, both from an audience research standpoint (understanding where the most relevant audiences can be found across channels) and an activation standpoint (applying huge media budgets to supply partners). Since they are on the front lines of where media dollars are expressed, they often get the first practical look at where data impacts consumer engagement. During and after campaigns conclude, the agency also owns the analytics piece. How did this channel, partner and creative perform? Why?

Having formerly limited agencies to campaign development and execution, marketers are now turning to the collected expertise of their agency media and analytics teams and asking them to embed the culture of audience data into their larger organization. When it’s time to select the DMP, the internal machine that drives the people-based marketing enterprise, the agency is naturally called upon.

Data Management Is About Ownership

Although a small portion of innovative marketers have begun leveraging DMP technology and taken media execution “in-house,” the vast majority stills relies on agencies and ad tech platform partners to operate their stacks through a managed services approach. Whether a marketer should own the capability to manage its own ad technology stack is a matter of choice, but data ownership shouldn’t be. Brands may not want to own the process of applying audience data to cross-channel media, but they absolutely must own their data.

Where Agencies Play in Data Management 

The Initial Approach: Most agencies have experience leveraging marketers’ first-party data through retargeting on display advertising. In an initial DMP engagement, marketers will rely on their agencies to build effective audience personas, map those to available attributes that exist within the marketer’s taxonomy and apply the segments to existing addressable channels. Marketers can and should rely on past campaign insights, attribution reports and other data insights from their agencies when test-driving DMPs.

Connect the Dots: For most marketers, agencies have been the de facto connector between their diverse platforms. Media teams operate display, video and mobile DSPs, ad-serving platforms and attribution systems. Helping a marketer and their DMP partner tie these execution platforms together and understand audience data and the performance data generated is a critical part of a successful DMP implementation.

Operator: Last, but not least, is the agency as operator of the DMP. Marketers want their data safely protected in their own DMP, with strong governance rules around how first-party data is shared. They also need a hub for using third-party data and integrating it with various execution and analytics platforms. Marketers may not want to operate the DMP themselves, though. Agencies can win by helping marketers wring the most value from their platforms.

Marketers have strong expertise in their products, markets and customer base – and should focus on their core strengths to grow. Agencies are great at finding audiences, building compelling creative and applying marketing investment dollars across channels, but are not necessarily the right stewards of others’ data.

Future success for agencies will come from helping marketers implement their data-management strategy, align their data with their existing technology stack and return insights that drive ongoing results.

[This post originally appeared in AdExchanger on 2.2.15]

Data Management Platform

2015 is the Year of Programmatic Branding

MuchWinWith companies like Kraft and Kellogg’s starting to leverage the programmatic pipes for equity advertising, we are starting to hear a lot of buzz about the potential for “programmatic branding,” or the use of ad tech pipes to drive upper-funnel consumer engagement.

It makes sense. Combine 20 years in online infrastructure investment with rapidly shifting consumer attention from linear to digital channels, and you have the perfect environment to test whether or not digital advertising can create “awareness” and “interest,” the first two pieces of the age old “AIDA” funnel.

The answer, put simply, is yes.

Online reach is considerably less expensive than linear reach, and we are starting to have the ability to reliably measure how that brand engagement is generated. Marketers don’t just want an “always-on” stream of brand advertising that comes with measurement – they also need it. With attention rapidly shifting from traditional channels, investments in linear television are starting to return fewer sales.

But most marketers are just starting to gain the digital competency to make programmatic branding a reality. That competency is called data management – the ability to segment, activate and analyze consumer audiences in a reliable way at scale.

The most fundamental problem with digital branding is that it is truly a one-to-one marketing exercise. If we dream of the “right message, right person, right time,” then matching a user with her devices is table stakes for programmatic branding. How do I know that Sally Smith on desktop is the same as Sally Smith on tablet?

Cross-device identity management is the key. Device IDs must be mapped to cookies, other mobile identifiers and Safari browser signals to get a sense of who’s who. Once you unlock user identity, many amazing things become possible.

Global Frequency Capping

One of the reasons programmatic branding has yet to gain serious ground with marketers is because of waste. This is both real, including all those wasted impressions due to invisible ads or robotic traffic, and perceived, such as impressions that are ineffective due to frequency issues.

Smart technology and market pricing solves the first problem, while data management solves the second. Assuming the marketer understands the ideal effective frequency of impressions per channel, or on a global basis, a DMP can manage how many impressions an individual sees by controlling segment membership in various platforms. Let’s say, for example, the ideal frequency for cereal advertising aimed at moms is 30 per day across channels. The advertiser knows showing fewer than 30 impressions reduces effectiveness, while more than 30 impressions has a negligible impact. Advertisers using multiple channels, such as direct-to-publisher, plus mobile, video and display DSPs, are likely overserving impressions in each channel and possibly underserving in key channels like video. Connecting user identity helps control global frequency and can save literally millions of dollars, while optimizing the effectiveness of cross-channel advertising.

Sequential Messaging

If “right person” technology is enabled as above, the next logical step is to try and get to “right place and right time.” Data management can enable this holy grail of branding, helping marketers create relevance for consumers as they embark on the customer journey. What brand marketers have dreamed of is now possible and starting to happen.

Dad, in the auto-intender bucket, is exposed to a 15-second pre-roll ad before logging into his newspaper subscription on his tablet in the morning. The message is reinforced by more equity display ads he sees in the afternoon at work. And while checking messages on his mobile phone on the way home, he receives an offer for $500 off with a qualified test drive. After Dad hits the dealership and checks in through the CRM system, he receives an email thanking him for his visit and reminding him of the $500 coupon he earned.

These tactics are not possible without tying user identity and systems together. Doing so not only enables sequential messaging, but also the ability to test and measure different approaches through A/B testing.

Cross-Channel Attribution

How about attribution? It’s impossible to perform cross-channel attribution without knowing who saw what ad. At the end of the day, it’s really about the insights.

Procter & Gamble is famous for spending millions of dollars every year to understand the “moment of truth,” or why people choose Tide over another detergent. Although they know consumer segmentation and behavior better than anyone, even the biggest brand marketers struggle to gain quality insights from digital channels.

Data management is starting to make a more reliable view possible. Brand advertising is just another form of investment. Money is the input. The output is sales and, just as important, the data on what drove those sales. In the past, brand marketers relied on panel-based measurement to judge campaign effectiveness. Now, data management helps brands understand which channels drove results and how each contributed.

It is early days for truly reliable cross-channel attribution modeling, but we are finally starting to see the death of the “last-click” model. Smart marketers use data to author their own flexible attribution models, making sure all channels involved receive variable credit for driving the final action. In the near future, machine learning will help drive dynamic models, which flex over time as new signals are acquired. We will then start to see just how effective – or not – tactics like standard display advertising are for driving upper-funnel engagement.

Is 2015 the year for programmatic branding? For marketers that are leveraging data management to enable the best practices outlined above, the answer is yes. The more accurately marketers can map online user identity and understand results, the more investment will flow from linear to addressable channels.

[This post originally appeared on 1.4.2015 in AdExchanger]