Time magazine just published a fascinating account of how President Obama's campaign team used data to microtarget voters. At HBR, we've been tracking the rise of Big Data in the private sector for
some time, and see this as a useful case study of how one organization
actually implemented those analytic principles to get results. I spoke
over the phone with MIT's Andrew McAfee, a regular contributor and
principal research scientist at the Center for Digital Business in the
MIT Sloan School of Management. What follows is an edited version of our
conversation.
Even though there's a push for data transparency, it seems like Obama's key to success was the opposite: to keep his data and algorithms as secret as possible. Organizationally, does it make sense to keep your data team siloed off from the rest of the team, or do you want them to work together more closely?
In this case, it's easier to identify the downsides of sharing that data more broadly. It could leak out, leak to the press, leak to the other candidate. If there is some secret sauce, you want to keep that close. If either the data asset itself or the algorithm on top of it seems to be cutting edge or proprietary, I wouldn't go shout it from the rooftops. You want to have the data team say to the volunteers, "Call up these people, knock on these doors, go to these neighborhoods." The volunteer doesn't need to know why; they just need to know they're knocking on the right doors.
That seems to be a move away from this movement we've been reading about about information wants to be shared, data wants to be free, everything should be transparent —
From any rational standpoint, that line is nonsense. That's like saying, "All money needs to be free," or if you're a trucking company, "All trucks need to be free." Data is an asset, like everything else.
In the spotlight on Big Data that we just ran in the magazine and online, we focused heavily on making the case that executives really do need to use it. We focused less on the execution piece of it. But it seems to me that having the data is only one part of the problem; you really need to know what to do about it.
There are two different questions to the execution piece. First, there's executing with the data resource and the computing resource itself. A lot of organizations aren't good at that because our data is fragmented. The amount of centralization and rationalization to take real advantage of big data is pretty daunting for a lot of organizations. To take the Obama campaign as an example, look at their move from having two databases in 2008 to one in 2012. I guarantee you that was not easy. That was technically pretty challenging, and it was organizationally super, super difficult — it means you have to convince one group to let go of their database and let go of being the sole keepers of that data. That's hard.
The second side of the execution, though, is the question of what do you do with the data. Which ads do you serve up, which doors do you knock on, which streets do you go down? Campaigns have been doing this for a long time; they know how to knock on doors. That's not the hard part. But to my eyes, what big data can do is help them be more efficient, by allowing them not to knock on every door on that street. You can say to the volunteer, "Just go knock on house numbers 14 and 18, and then skip the next six houses and go knock on house number 30."
Big data lets you be small. It lets you do really precise, targeted interventions. I see evidence of this over and over, in all kinds of organizations. It started with marketing, but that has led the way to operations and the supply chain also using those kinds of precise data-driven interventions. And then you can do that at scale.
So you're saying it lets you be small, but at scale? Are you then big again? [Laughs]
It's a very different kind of big; it's not mass production. "Mass customization" is an overused phrase and it's been around for a while, but in this era of big data we can really do that. It's no longer just configuring my car so it has the upholstery I want. It's that campaigns no longer need to do the kind of mass mailings they're always relying on; they can get the right kind of mailing to the right kind of voter. Maybe this one cares about women's rights, but that one cares about economic policy.
For the Obama campaign, that was at the heart of their ground game advantage. In our HBR article, we included an example from the health care sector. Over and over we're seeing this ability to be precise and differentiated, at scale and repeatedly, with a lot more efficacy.
How would this work in the business world, say in health care?
Take the example of Aetna. They have around 20 million lives under care. Because they're the company that pays the doctor, the lab, the pharmacy, and so on, they have all this information about my health. There are huge confidentially concerns, of course, but let's put that aside for a moment.
What that means is that they have better data about my health than any other player in the system; than my doctor, my hospital, my pharmacy. What Aetna can do at scale, in real time, over and over, is check across the millions of lives in their system and say, "Do we see any obvious gaps in care here?" They can then send a message to the doctor or to the patient. Now, we need to be careful about confidentiality, and we need to make sure the messages are phrased the right way so they're not ignored, but we can use this database to do mass scale interventions into health care delivery. And that means we can improve outcomes without having to rejigger the entire system.
Okay, so say everyone starts doing that. Then do you get to a place like we see now in baseball — where a few years ago, crunching the numbers gave some teams a big advantage, but now everyone does it and it confers less of one? Look at the Oakland A's. Using data used to give them a big advantage, but since every team started to do it, it really doesn't seem to give them much of one.
Yes, you've got to move on. It's not going to give you an advantage forever, but if you are analytically oriented, you can push ahead and get finer-grained insight and advantage. In baseball, the science of sabermetrics has moved on. They're doing increasingly sophisticated things. Billy Beane and the A's were able to pick the low-hanging fruit and do really well for a while because they were able to do the simple things better. Now everyone has those insights so they've got to work harder. The marginal benefit might be less, but you've still got to do it. If you just go back to scouting players like you did in the 1990s, that's a great way to have a terrible team. Standing still is a very bad strategy.
What's interesting to me is that the volumes of data are exploding terribly quickly. The toolkit is also expanding by leaps and bounds. This is a real new arms race. You might not love it, and you might wish the world was predictable and calm and that Excel would get you through — but that would be a recipe for disaster.
http://blogs.hbr.org/hbr/hbreditors/2012/11/the_analytics_lesson_from_the.html
Even though there's a push for data transparency, it seems like Obama's key to success was the opposite: to keep his data and algorithms as secret as possible. Organizationally, does it make sense to keep your data team siloed off from the rest of the team, or do you want them to work together more closely?
In this case, it's easier to identify the downsides of sharing that data more broadly. It could leak out, leak to the press, leak to the other candidate. If there is some secret sauce, you want to keep that close. If either the data asset itself or the algorithm on top of it seems to be cutting edge or proprietary, I wouldn't go shout it from the rooftops. You want to have the data team say to the volunteers, "Call up these people, knock on these doors, go to these neighborhoods." The volunteer doesn't need to know why; they just need to know they're knocking on the right doors.
That seems to be a move away from this movement we've been reading about about information wants to be shared, data wants to be free, everything should be transparent —
From any rational standpoint, that line is nonsense. That's like saying, "All money needs to be free," or if you're a trucking company, "All trucks need to be free." Data is an asset, like everything else.
In the spotlight on Big Data that we just ran in the magazine and online, we focused heavily on making the case that executives really do need to use it. We focused less on the execution piece of it. But it seems to me that having the data is only one part of the problem; you really need to know what to do about it.
There are two different questions to the execution piece. First, there's executing with the data resource and the computing resource itself. A lot of organizations aren't good at that because our data is fragmented. The amount of centralization and rationalization to take real advantage of big data is pretty daunting for a lot of organizations. To take the Obama campaign as an example, look at their move from having two databases in 2008 to one in 2012. I guarantee you that was not easy. That was technically pretty challenging, and it was organizationally super, super difficult — it means you have to convince one group to let go of their database and let go of being the sole keepers of that data. That's hard.
The second side of the execution, though, is the question of what do you do with the data. Which ads do you serve up, which doors do you knock on, which streets do you go down? Campaigns have been doing this for a long time; they know how to knock on doors. That's not the hard part. But to my eyes, what big data can do is help them be more efficient, by allowing them not to knock on every door on that street. You can say to the volunteer, "Just go knock on house numbers 14 and 18, and then skip the next six houses and go knock on house number 30."
Big data lets you be small. It lets you do really precise, targeted interventions. I see evidence of this over and over, in all kinds of organizations. It started with marketing, but that has led the way to operations and the supply chain also using those kinds of precise data-driven interventions. And then you can do that at scale.
So you're saying it lets you be small, but at scale? Are you then big again? [Laughs]
It's a very different kind of big; it's not mass production. "Mass customization" is an overused phrase and it's been around for a while, but in this era of big data we can really do that. It's no longer just configuring my car so it has the upholstery I want. It's that campaigns no longer need to do the kind of mass mailings they're always relying on; they can get the right kind of mailing to the right kind of voter. Maybe this one cares about women's rights, but that one cares about economic policy.
For the Obama campaign, that was at the heart of their ground game advantage. In our HBR article, we included an example from the health care sector. Over and over we're seeing this ability to be precise and differentiated, at scale and repeatedly, with a lot more efficacy.
How would this work in the business world, say in health care?
Take the example of Aetna. They have around 20 million lives under care. Because they're the company that pays the doctor, the lab, the pharmacy, and so on, they have all this information about my health. There are huge confidentially concerns, of course, but let's put that aside for a moment.
What that means is that they have better data about my health than any other player in the system; than my doctor, my hospital, my pharmacy. What Aetna can do at scale, in real time, over and over, is check across the millions of lives in their system and say, "Do we see any obvious gaps in care here?" They can then send a message to the doctor or to the patient. Now, we need to be careful about confidentiality, and we need to make sure the messages are phrased the right way so they're not ignored, but we can use this database to do mass scale interventions into health care delivery. And that means we can improve outcomes without having to rejigger the entire system.
Okay, so say everyone starts doing that. Then do you get to a place like we see now in baseball — where a few years ago, crunching the numbers gave some teams a big advantage, but now everyone does it and it confers less of one? Look at the Oakland A's. Using data used to give them a big advantage, but since every team started to do it, it really doesn't seem to give them much of one.
Yes, you've got to move on. It's not going to give you an advantage forever, but if you are analytically oriented, you can push ahead and get finer-grained insight and advantage. In baseball, the science of sabermetrics has moved on. They're doing increasingly sophisticated things. Billy Beane and the A's were able to pick the low-hanging fruit and do really well for a while because they were able to do the simple things better. Now everyone has those insights so they've got to work harder. The marginal benefit might be less, but you've still got to do it. If you just go back to scouting players like you did in the 1990s, that's a great way to have a terrible team. Standing still is a very bad strategy.
What's interesting to me is that the volumes of data are exploding terribly quickly. The toolkit is also expanding by leaps and bounds. This is a real new arms race. You might not love it, and you might wish the world was predictable and calm and that Excel would get you through — but that would be a recipe for disaster.
http://blogs.hbr.org/hbr/hbreditors/2012/11/the_analytics_lesson_from_the.html
No comments:
Post a Comment