I guess I’m going to start out sort of moti­vat­ing some of this. And when we talk about per­for­mance and pro­duc­tiv­i­ty, com­pa­nies are so data-driven—probably a lot of your com­pa­nies are so data-driven—when it comes to cus­tomers. And I could ask you ques­tions about where your cus­tomers buy prod­ucts, or what sort of prod­ucts they buy. And you could give me very detailed answers.

But I could ask rel­a­tive­ly sim­i­lar ques­tions about what goes on with­in your com­pa­ny that you can’t answer. So, if you’re an engi­neer­ing com­pa­ny, how much does man­age­ment actu­al­ly talk to the devel­op­ers? If you’re a retail­er, how much should you talk to a cus­tomer at a store? Now, you think about how basic, how fun­da­men­tal those ques­tions are. And one of the rea­sons that we can’t actu­al­ly answer those ques­tions is because we don’t have very good data about what actu­al­ly goes on inter­nal­ly. And in fact, there’s real­ly only one indus­try in the world that is total­ly data-driven about the way its peo­ple behave at work, and uses that data to make deci­sions. And those orga­ni­za­tions are base­ball teams. 

Photo: Philadelphia American League Base Ball Team via Library of Congress

For 150 years the way you built a base­ball team is you had a bunch of old guys who knew a lot about base­ball watch peo­ple play base­ball, and then based on their sub­jec­tive eval­u­a­tions they’d build a team. Now, some­times they’re right, some­times they’re wrong…that was viewed as the best way to do things. And I’ll point out for those of you who are not famil­iar with the game of base­ball, it’s a game where you hit a ball with a stick, and it’s a game that’s a lot more fun to watch if you’re drunk. So, here we go. 

What’s fas­ci­nat­ing is this very sub­jec­tive way of man­ag­ing these teams con­tin­ued into 2001. And then one day you get this guy Billy Beane (or Brad Pitt, if you like) who said, No, we’re going to use behav­ioral data to build our orga­ni­za­tion. We can use data about what peo­ple do on the field—” and every­one thought he was crazy. But if you saw the movie Moneyball, if you read the book, they went on a record win­ning streak, they made the play­offs. An now every sin­gle orga­ni­za­tion builds their team in this way.

The ques­tion is what is the equiv­a­lent of that in indus­try? Because of this I’ll put up one ques­tion here which might seem a lit­tle strange. Why is orga­ni­za­tion­al change hard?” The thing is you know, things like M&A, restruc­tur­ing, we know they’re hard. But we fail a lot at those things. I mean, M&A fails con­ser­v­a­tive­ly about 60% of time. 

Slide depicts a company organization chart on the left, and two people chatting at a water cooler on the right

And a big rea­son is because we focus on the stuff that’s easy to cap­ture. We focus on the stuff over there, the for­mal stuff. Because it’s easy to under­stand. I can point to the per­son at the top and I can say that’s the most impor­tant per­son in the company. 

On the oth­er hand if I ask you, again, how much does man­age­ment talk to a divi­sion? If I ask you who’s the social cen­ter of a com­pa­ny, that’s much hard­er to answer. Because we use sur­veys or con­sul­tants, which are use­ful for cer­tain pur­pos­es but can’t give us the same gran­u­lar­i­ty dai­ly, even month­ly, about what’s actu­al­ly going on. But now because we have sen­sors, we’re wear­ing next-generation ID badges, we have dig­i­tal data, email, IM, phone calls, all of a sud­den we can under­stand at a mil­lisec­ond lev­el what is actu­al­ly going on with a com­pa­ny. And then we can use that to first, actu­al­ly under­stand what hap­pens, but also under­stand what actu­al­ly dri­ves the out­comes we care about.

And this kind of data gives rise to this whole area of peo­ple ana­lyt­ics. And I’ll have the stan­dard plug for my book, which it makes a great gift. You can go buy mul­ti­ple copies. But the idea behind peo­ple ana­lyt­ics is real­ly using behav­ioral data to under­stand what’s going on and chang­ing the way your com­pa­ny is man­aged. I’m hap­py to say that now there are well over a hun­dred com­pa­nies that have peo­ple ana­lyt­ics divi­sions. What they do is they take the lessons we’ve learned about ana­lyz­ing cus­tomer data and applied it internally.

I want to give you a fla­vor of what that actu­al­ly can look like. So, one of the things that we do is we use sen­sors, sort of next-generation ID badges, to mea­sure how peo­ple talk to each oth­er, using a micro­phone. Who talks to who. And where peo­ple spend time, and how they move around. We don’t record what peo­ple say. We don’t give indi­vid­ual data to com­pa­nies. And I know we’re going to talk lat­er about some of the pri­va­cy impli­ca­tions of this kind of tech­nol­o­gy. But what I want to show you is data from a major European bank. 

What we did is we deployed these sen­sors across actu­al­ly hun­dreds of loca­tions. These are retail loca­tions, where have peo­ple sell­ing loans to cus­tomers. And they have some loca­tions where they real­ly out­per­form the pro­jec­tions that the com­pa­ny made, and they have oth­er loca­tions that real­ly don’t do such a good job. From a for­mal per­spec­tive, they’re exact­ly the same. People are trained in the same way. You have sim­i­lar employ­ee demo­graph­ics. Just very dif­fer­ent performance.

So what do they do dif­fer­ent­ly? I’m going to show you data from three branch­es. I added some noise to the data so it’s not actu­al­ly the orig­i­nal data. But it gives you a fla­vor for what you can see. And I’ll show you just a very sim­ple cut of the data.

Wee see three branch­es here. Each dot rep­re­sents a per­son. The lines between them rep­re­sent how much they talk to each oth­er face to face, which we can detect with these devices. So we see some­thing very inter­est­ing. There’s a num­ber of branch­es like Branch 1, where pret­ty much every­body talks to every­body else. We see branch­es like Branch 2, where we have these two clus­ters here, which is sort of inter­est­ing. And then you’ve got branch­es like Branch 3, where you’ve got a real­ly tight knit core, but you’ve got three lone­ly peo­ple on the outside. 

And so the ques­tion is which branch has the high­est per­for­mance? They look very dif­fer­ent. We found hun­dreds of branch­es that clus­ter into these dif­fer­ent areas. So we’re going to have a lit­tle vote here. These peo­ple all have the same job; they’re sell­ing small busi­ness loans. And I can mea­sure their per­for­mance very con­crete­ly: how many loans did they sell per per­son? Alright, so who thinks Branch One had the high­est per­for­mance per per­son? Alright. Branch 2? And Branch 3.

Okay, pret­ty close. You guys did pret­ty well. So, Branch 1 did have the high­est per­for­mance. And to give you a sense of the order of mag­ni­tude of that effect, Branch 1, the aver­age employ­ee there would sell about 250% times the num­ber of loans as some­one in a branch like this, like Branch 2

Now, Branch 3 was also quite close, but these out­liers brought every­one’s per­for­mance down. So what’s hap­pen­ing? Why does this hap­pen? Well, now we’ve iden­ti­fied that there are behav­ioral dif­fer­ences in these dif­fer­ent kinds of branch­es. So we can actu­al­ly start to zoom in, and the bank sent peo­ple to see what do they do dif­fer­ent­ly. People are paid based on how much you sell. But it turned out that at branch­es like Branch 1, man­agers had imple­ment­ed an infor­mal bonus sys­tem. They had a tar­get for the whole branch, and if you hit it every­body got a bonus. They incen­tivized peo­ple to share, and they do.

Branches like Branch 2, we saw lots of them like this. Exactly two groups, and every time it was two groups. Any hypothe­ses about why that would be? First floor, sec­ond floor. We timed it. It takes less than ten sec­onds to walk from one floor to anoth­er. But nobody does it. And it had a huge effect. You save mon­ey on rent, you lose it in performance.

Branches like Branch 3, any guess­es for these out­liers, who they are? Work from home? Not quite. Bosses. I wish; no. These are new employ­ees. But they’d all been in the branch for over a month, and they still only talk to their man­ag­er. Actually, they only talk their manager. 

Now, what they did is they ran tests. Took this group per­for­mance sys­tem, rolled it to out to half of their branch­es. They start­ed switch­ing peo­ple’s desks, first floor and sec­ond floor, in two-floor loca­tions. And now they give €100 a week to man­agers to take their new employ­ee out to lunch with some­body else. We can get into the details lat­er. Long sto­ry short, with con­trols, that increased sales by over 11%. This is worth over €1 bil­lion a year. Using rel­a­tive­ly sim­ple changes. 

And so the future of this tech­nol­o­gy, a lot of peo­ple think of Star Trek or Star Wars, and I like to think about some­thing a lit­tle dif­fer­ent. I like to think about Harry Potter. If you remem­ber the Harry Potter movie, they have these stair­ways that move on their own; I always real­ly liked that idea. And imag­ine with this kind of data com­ing in all the time that you can get into an ele­va­tor, and you don’t press any but­tons but you get offered a cer­tain floor, an algo­rithm lets you off there. Because you’re prob­a­bly going to have an inter­est­ing con­ver­sa­tion there.

Imagine the cof­fee machines move around at night; they’re robots, and actu­al­ly do that to change the inter­ac­tion pat­terns in an office. And lit­er­al­ly one of our cus­tomers does that, so I actu­al­ly have data from this. Groups in the Netherlands are mak­ing robot­ic walls. And even­tu­al­ly, of course it’s going to apply to real­ly big busi­ness­es. But it could even­tu­al­ly apply to small busi­ness as well, where you could under­stand what makes my restau­rant effec­tive, and how can I com­pare that to the best restau­rant in the world? Or what could I learn from a gro­cery store in Africa? 

And then how could I not just learn from those com­pa­nies, but big com­pa­nies learn from small com­pa­nies? Companies in Russia learn from com­pa­nies in the US. And real­ly rather than wait­ing a decade for an HBR case study to come out, being able to do this lit­er­al­ly in a mat­ter of days. So real­ly that’s the poten­tial. And I appre­ci­ate every­one for lis­ten­ing. Thank you very much. 

Further Reference

Annual Meeting of the New Champions 2016 at the World Economic Forum site