Commentary
If User-Level Data is Too Costly, Look into Alternatives
- by Michael Kaushansky , March 20, 2012
Everyone is racing to connect all their online data to a single user, to provide a view into that consumer’s behavior and engagement with a product/brand across a plethora
of touch points and channels. There are many marketing benefits in having all the information about your consumer at your fingertips; however, is it cost-effective to embark upon multi-source
consumer-level data integration? Are there alternatives?
When we talk about consumer-centric data, we are really talking about three categories of data types: behavioral, attitudinal, and demographic. Behavioral data will tell you what your consumer is doing online, typically starting with ad-serving tracking. Attitudinal data will tell you why your consumers are behaving a certain way, by uncovering their perceptions/opinions of your product/brand. Finally, demographic data will tell you who these consumers are. As you can imagine, having this granular data can be extremely telling and powerful -- especially if the data is not limited only to online media, but also includes mobile and social. This is powerful because it will enable multi-screen, multi-touch, and cross-channel measurement (including websites) at the consumer level. Add demographics and even attitudinal feedback, and you have the full picture.
Several players have claimed success in this space, specifically the largest DMPs (data management platforms). Combining the three types of data usually involves taking verified offline data, such as mailing lists, with demographics, and connecting it using emails to online consumer-level behaviors. Some data aggregators leverage online panels or surveys to refine and include attitudinal response, and some simply extrapolate the three types of data by applying statistical models to online patterns. Having data aggregated for you is extraordinarily helpful, though I urge you to dive deeper into the services these providers offer. You may find in some cases their data is extremely thin, incomplete, subject to aggressive assumptions and in some cases flat-out inaccurate -- not too surprising, since this is true for many marketing data sources and you need to be smart in the way you select and use third-party data. You may also look into doing it yourself or look for alternatives.
Using Big Data to provide a single view of a consumer has become Big Business! There is a mad rush to be more granular and detailed in the data we collect, though in some cases I would argue it can be over-kill and overwhelming. Acquiring consumer-level data has its set of challenges:
1. Privacy -- there is increasing pressure for regulation, with questions about the safety, security and ethics of data usage and control in interactive marketing.
2. Complexity – integration across data sources at a consumer level is often very complex.
3. Lack of data standards – data is often fragmented without any unique or common identifiers.
4. Expensive – although there is some universal integration happening in the
industry, for the most part we still lack a uniform framework for data integration. Custom work is often expensive, and time- and resource-intense.
5. Panel/survey bias – there are two challenges with panel/survey data:
a. A recent study conducted by an online panel taskforce concluded that a vast majority of online panels do not conform to using probability-based recruitment; therefore the findings and insights off that panel may be biased and not representative. We all know survey error does exist, but if it is larger than expected, it will certainly skew the overall results.
b. Panels and survey data may not be scalable for smaller or niche campaigns due to sample size limitations.
Don’t get me wrong - I do believe that consumer-level data is very powerful and offers rich insights. If you can overcome the challenges, it will work for you. Let me also suggest two alternatives that just may get you the answers you need
1. Marketing econometric modeling (MEM) – a method of aligning data based on a time-series versus at the consumer level. Traditionally this approach has been used to develop offline media mix models; however, this will also work well in evaluating daily online activity across multiple channels, devices, media, etc. Although data is still not always easy to come by, MEM integration is much simpler. MEM may not address multi-screen measure but can be used for multi-touch and is already being used for cross-channel measurement and analysis.
2. Agent-based modeling (ABM) – agents are autonomous decision-making entities (consumers), and agent-based models are made up of agents and a framework for their interactions. It is a virtual representation of your consumer base matched up with your segmentation. Each agent is assigned behaviors and representative patterns. These models provide sufficient levels of detail about the behaviors of your consumer. The approach can be used to address very detailed questions and allow for multi-screen, multi-touch and cross-channel measurement and analysis.
Whether you rely on consumer-level data, time-series or simulated agents, one thing is clear: marketers today need the ability to track a consumer across the complex web of marketing communications and touch points to better understand marketing ROI.
This is a nice breakdown of some alternatives that might get marketers to a better place than they are in today.
There are a few things that I would question prior to making a decision to steer away from a strategy that brings you to the consumer-level nirvana described (accurately) as ideal.
First, I would encourage marketers to consider not just "what is the cost to get there" but also "what is the opportunity cost" for not getting there. What if your competition gets there first and they are operating on a higher plane than you? Can you afford that long-term? Or worse, what if your customers develop a thick callous to your brand message because you continue not to deliver the best experiences across your touches with them? How much more will you have to invest in marketing to achieve a level of effectiveness that wins for you? I don't mean to make it a fear-based insurance sale but its a real consideration for those who only look at it from one side.
Second, thanks to innovation and market demand, there are lower-overhead alternatives to the big DMPs that are built on newer technology foundations and with the customer-centric approach in mind (not a bolt on tech integration or a rebrand of an archaic platform that is "serviceable" but expensive). Here is a link to a Forrester paper that digs more deeply into the topic from a messaging perspective (http://www.knotice.com/whitepapers/forrester-new-messaging-mandate/) and he explores some of the material impact that can be expected with a change in orientation.
Its no secret that this approach is incredibly powerful and that it works. The big secret seems to be that actual customer-level data is more accessible than people think- especially to mid-size or larger agile companies able to consolidate and motivate for the change across the enterprise.
As the author pointed out, we live in a highly connected world where real-time access to news, data, and information has transformed the way both people and businesses make decisions. To navigate this ever-increasing complexity, marketers must shift the focus from data-intensive tactical planning to a smart data approach for strategic, integrated marketing that is centered around real people they are targeting. Many of the insights, analytics, and processes used to inform marketers today are based on aggregations of consumer data, boiling all the diverse demographics, media consumption, and behaviors into one "average consumer." This makes it impossible to connect the very people they target to specific marketing activities and messages. An agent-based modeling approach to marketing helps marketers to understand how their consumers behave in today's evolving and dynamic marketplace so that their marketing activities will get them to buy their products. It is a more accurate, agile and effective way to measuring and forecasting the ROI across consumers, channels, and tactics to make more informed marketing decisions.
Michael, thanks for bringing up an important topic. The comment below from David Lawson is very valid here. What is the opportunity cost for not doing the right thing? As David mentioned the Big Data technology is advanced a lot in the last 5 years and analyzing several billion impressions a day is done these days very cost effectively. Author’s concern about some 3rd party data being biased, extremely thin, incomplete, subject to aggressive assumptions and inaccurate, is also true for his own recommendations. Marketers should be worried about the black-box type analytics where the marketer has to put a lot of trust and jump off the cliff with them. I would encourage marketers to embrace methodologies that can back up the math using marketer’s own data. Not sampled or simulated - the full set of data. Sometimes, there is no shortcut to doing things right vs. being wrong.
I agree that the costs of managing Big Data have been significantly reduced, but tracking/tagging it across the entire campaign (paid/earned) is complex and involved. Unless you are using Big Data to target individuals a sampled file will do....FDA approves drugs every year based on small statistical sample. Google Analytics, the free version uses sampled data. The point is that sometimes the effort in trying to collect everything does not outweigh the benefit (caveat - it all depends....) ;-)