Big Data Demands Big Context

Business history is replete with stories about well-thought-out, exhaustively researched initiatives that totally, utterly, humiliatingly bomb. 

New Coke usually gets pride of place in this Hall of Shame, a product developed precisely in response to customer data implying that Coca-Cola's loyal followers actually wanted something that tasted a little more like, well, they didnt want to come out and say this exactly, but like, gulp, Pepsi. So Coke delivered what they thought would be the knock-out punch to a determined but competitor. That KO, as most business veterans know, was delivered not to Pepsi, but to Coke itself, as loyal customers rebelled at the introduction of the new product, despite what they had told researchers.

The tech world is now in for its share of such epic misinterpretations. Social media generally and the Facebook IPO in particular provide evidence of the disconnect between hucksterism and the need to demonstrate a credible propensity to generate cash money. Microsoft, too, with its Windows 8 'tiles' adapted what it thought was the genius of a smartphone design to computer screens, failing to understand, as the following article explains, that what works in brief spurts on a small screen is not necessarily relevant or especially helpful in addressing the tasks performed on a larger, less mobile one.

Context is the missing ingredient in the Big Data revolution. Advocates extoll the virtues of information as if the mere presence of such data will solve problems all by itself. Just as rational explanations of the solution to socio-religious-political differences will result in world peace...

Context is about accurate interpretation of the numbers depending on how they are being used and for what purpose. It provides awareness of the sources, uses, and impacts or outcomes that can and should be expected, as well as how those affected by these changes will react. Context, in other words, enables the user of the data to place it in an appropriate framework. This optimizes the variables in order that desirable outcomes are more likely to be achieved.

Steve Jobs taught us that the customer is often the last person you want to ask about product or service improvements because they just want what they have to work a little better rather than re-imagining what might be possible. Every enterprise should be figuring out how to apply that insight to big data. JL

Jess Neill comments in Harvard Business Review:

When we entered the age of big data, many of us assumed we had left the age of big risk. We didn’t have to guess anymore. We didn’t have to go out on a limb. We’d follow the numbers, the “truths.”
When Microsoft built Windows 8, its goal was to move beyond operating-system conventions that were based on outdated user-behavior assumptions and create an OS for the way people really use computers today.
Microsoft’s engineers discovered that people were doing less of the time-consuming writing and creating that had once been the norm. Increasingly, users were socializing for short bursts.
The research also showed that people loved having “touch” functionality and were avidly consuming small pieces of live information.
Consequently, Microsoft decided that Windows 8 should feature navigation that enabled multitasking and quick interactions, and that it should also have touch and live tiles.
None of this was wrong. And yet these decisions, so carefully researched and thought through, all contributed to the failure of Windows 8.
How does this happen?
But time and time again we’re finding that it’s not that simple. No matter how good the research is, big data is nothing without big context.
To keep context in mind, there are a few questions I ask myself while designing research, analyzing data, or, most important, making decisions.

• What underlying assumptions am I making?

In Windows 8, I think Microsoft’s engineers made a fundamental assumption that led them astray: that users want one interface for all machines, one machine for every part of their lives. The research went into what this single interface should be. Not whether it should be.

What if users don’t want just one device? What if they’re embracing and owning many specialized devices?

It’s easy to get into a bubble and focus our thinking too quickly. So whenever I approach data, I first ask myself what assumptions I’m making.

• What emotions will be at play?

When asked what they want in the hypothetical, people answer rationally. They make “good” decisions. They pick cheaper, faster computers that are less attractive. Or they say that they’d try an exciting new platform without considering the frustration of a learning curve.

But what people say and what they do are two very different animals. Which of us hasn’t been seduced into a less-savvy purchase because of a shiny case?

When designing research, I try to probe for the emotional drivers as well as the rational drivers. I want to know if consumers see a product as a utility or a luxury. Do they identify with a store or brand as part of their persona? Do they see it as a friend? I can use these answers to temper and check the purely rational responses.

•  How can I better learn about context?

Yes, people are using touch daily on smart phones and tablets. It’s intuitive. Just look at YouTube videos of babies trying to “swipe” pages in physical magazines.

But when Microsoft put touch at the forefront of its operating system for PCs, consumers didn’t bite, partly because touch computers were more expensive than non-touch. But the bigger problem was that although touch is great for the social interactions and brief browsing that people do on smart phones and tablets, users are relegating PCs to work and productivity, and in that context, they don’t see the value of touch.

Microsoft had the right information. But it was missing the larger contextual picture.
The details are easy to measure. They give you clear-cut data and answers. Context is harder—it gets mushy. The methodologies are more complex and the results are open to interpretation.
Most disconcerting of all, data about context won’t give you an answer; it will only help inform your answer. Compounding that, contextual data can seem superfluous, so fighting for the money to research it can be hard, and selling ideas based on it even harder.
But we take a bigger risk when we ignore the context.
Microsoft’s research points to an increasingly diverse device landscape, with each device being used for specific and differentiated uses and behaviors. Probably the ultimate PC OS leans into PC behaviors, letting the smart phones and tablets specialize and optimizing the user’s journey between devices.
But the real lesson is that we always need to consider context. Otherwise, we too can have all the right answers to all the wrong questions.