Here’s what’s wrong with your data model
September 11, 2017
Tricia Wang is a global technology ethnographer who’s spent a lot of time helping businesses grow by discovering new and unknown things about their customers. In addition to working in R&D at Nokia and spending a year as an expert-in-residence at IDEO, she co-founded Constellate Data. The consultancy provides organizations with research and training services that help them better leverage data to understand people, and ultimately grow their business.
She’s also the world’s most renowned matador, an ex-FBI agent, and holds a Ph.D. in “cooking while submerged in water,” from the Aquatics School of Culinary Studies, according to her LinkedIn profile.
“LinkedIn is trying to create a certain identity for me. It wants me to present myself as this totally together professional,” Wang says. “And I resist any form of boxing me in.”
Wang explains that when a company is working with what’s called a singular identity data model, it assumes that by simply collecting enough data on any consumer, the company can know everything about that consumer. But according to her research, that’s not true. And she wants to help companies understand that.
“I’m saying, ‘No, I am many different peoples,’” Wang says. “This whole Western concept is that we have one authentic self, and we work all our lives to discover who we really are. And then you find out who you really are and then you present that. But we actually present different identities in different contexts. And this matters a lot for how businesses relate to, speak to, and target their customers.”
Shift your perspective.
At one point, the singular identity data model did work for companies trying to understand consumers, Wang says. When the internet was young, there weren’t as many spaces for people to express themselves. So, people presented more stable, singular identities online, because that was all they could do with what limited spaces there were. But now that the internet is a more complex and diverse ecosystem, people present multiple identities on the internet—and they’re all different. Who you are on LinkedIn may be very different from who you are on Twitter, or on Facebook. You may have multiple Twitter accounts and behave very differently on each one, depending on the purpose of the account.
The problem with the singular model is that it means that the software a company uses to aggregate the data doesn’t acknowledge this diversity.
“I know what the software is doing,” Wang says. “The software is trying to link my LinkedIn, to my Facebook, to my Instagram. From all the data I present on the internet, it’s trying to create this “person,” and then they want to put me into demographic boxes. They want to build data models about my behavior… But these data models are being built on assumptions of what the human interaction model is. When in reality, the human interaction model is not one-to-one. I don’t fall into any one category. I, like others, are part of multiple communities, and it’s becoming harder to throw us into boxes. We are networked, complex, and elastic, and it’s only recently that we’ve had a range of virtual spaces to reflect that. But we haven’t conceptually understood this yet, so companies keep building faulty data models.”
If your data models are off—if they don’t reflect real human behavior—your assumptions will be off, too, and the problem will manifest at scale, Wang says. Take software innovation as an example. After spending gratuitous time and money researching, designing, and deploying a business or consumer application, customers or employees may still grumble about the user experience because the app was designed based on incomplete research.
Ask the right questions.
The solution for companies, Wang advocates, is to make sure that you are asking the right questions before you ever try to answer them. This requires the collection and use of both quantitative and qualitative data, and sometimes requires an organization to step outside of its existing business model. Wang experienced this firsthand during her time at Nokia—which she details in her TED talk, The human insights missing from big data.
At Nokia, as in many enterprises, decision-makers had a business model that placed all of the emphasis on quantitative data—numbers, numbers, numbers. Meanwhile, Wang collected qualitative data, and didn’t have a way to scale her insights, because it was outside of the company’s business model. Decision-makers would want to decrease churn, for example. Wang argued that before they could understand how to decrease churn, they needed to study how people actually used their product.
“Really, we were both right,” Wang says. “The decision makers I was working with, they were saying, ‘Look, we’re trying to prioritize for tomorrow—the quarterly report.” And I was saying, ‘I get it. I’m not touching that. I am talking about in two years. But we do need to start now.’”
Create a shared language.
Wang decided that resolving these conflicts started with all parties truly understanding each other. And to do that, they needed a shared language.
“We didn’t have a way to talk about our roles in the companies in a depersonalized way,” Wang says. “I don’t think they felt that I truly understood their [data scientists’ or analysts’] needs. And I didn’t think that they understood my needs.”
So, Wang created a common language for talking about questions of conflicting interest when it comes to data. She calls it Integrated Data Thinking©.
“The framework teaches people to talk about it in a low risk way. To try out different things,” Wang says.
The framework starts by nailing down the questions that really need to be answered by data, and then asking, “What’s the quickest way to get the best answer?” For example, some questions are most quickly and accurately answered by ethnography or exploratory big data analysis. Alternatively, some questions might work better for qualitative focus groups. This shared language allows non-experts to discuss questions, and for qualitative and quantitative teams to collaborate on the right deliverables. It also reveals whether resources are aligned to the goals, and helps teams prioritize responsibilities.
Integrated Data Thinking proved so successful, in fact, that it became the basis for Constellate Data, the company Wang co-founded.
“That is very important to me, to teach a shared language,” Wang says. “Giving businesses the skills they need to integrate, to be data-literate and to identify the unknown, but also communicate that within the context of the known—we love teaching companies how to do that.”