Let's go through some of the steps on linking non-financial metrics to financial performance. Again, here's some fundamental questions. We're going to have this causal business model, so the first question is, to what extent does this business model actually incorporate the right drivers of financial success? Obviously, they're predictions. That's what a strategy is. It says, "If I do this, this is how I think financial performance is actually going to end up happening." Do we have the right measures in there? Again, you could have a measure that you think is good. It just doesn't capture what you think it's capturing on this. The second is, okay, given that, now that I've got this business model, I think I have some measures that I want to use. How can I use that business model to allocate resources? Because ultimately that's what we need to do. Now, allocation resources could come through. I said budget's based on this. It could come simply because this is how I'm going to evaluate managers, and if I evaluate you on something, that's what you're going to give me. I don't specifically have to tell you what to do. Finally, as I've said before, we have to worry about how you set targets for these measures, which is going to be one of the most difficult problems that we have with non-financial measures, because more is not necessarily better. Let's start with step one, identifying the right drivers. Here's the steps. We're going to develop this causal business model. First of all, again, it has to be linked to organizational strategy. What are you trying to achieve? What the causal business model is trying to do is saying, "Here's what I'd like to achieve. What are the steps that I need to take to successfully implement this strategy, see if it works, and see if ultimately, at least, a financial performance." The way we're going to do that is we're going articulate these hypothesized drivers. This causal thing, A leads to B leads to C, those are hypotheses. For each one of those A, Bs and Cs, we're going to come up with financial or non-financial performance measures and see if ultimately, do they translate through the way we think it's going to. Second thing we have to do is construct what are called reliable and valid measures of the key drivers. Basically, do these measures pick up what you think they're picking up? Our strategy says, "Okay. Do A. Do B. Do C." The question is, how do I measure A? Again, this is not always such an easy thing to do on this. Here's a little pop quiz. If you think of bad fast food hamburger chains in the United State, there's lots of ways I could evaluate, "Are you satisfied with the fast food hamburger chain". I could say, "Is it the quality to the hamburgers?" I could say, "Is it the cleanliness?" I could say, "Is it how easy is it to get in and out of the parking lot?" If you think about most customer satisfaction surveys you've seen, they ask you a whole bunch of questions. I can answer all of those, but what I really want to know as a company is not just, are you satisfied, are you satisfied on a dimension that's going to make you come back and buy more, and that's the key thing. It turns out, for fast food hamburger chains in the United States and some research we've done, the quality of the hamburger is actually quite low. Now it turns out the quality of the French fries is really important, but it has very little predictive ability of whether they're going to come back. Number one, and only those of you who actually have small children are going to understand this, the number one predictor of whether somebody says they're going to come back is was their child happy with the toy and the Happy Meal last time you went there. Because as you're going to realize once you have small children, the only time you go out for fast food is when your kids want you to. Now if you think about that, if I ask you something about, did you like the hamburger and I'm using that to evaluate managers, they're going to improve the quality of the hamburger, but if that doesn't cause you to come back and buy more, that's not a good strategy. What we need are constructs or measures that are reliable at capturing what you think they're going to capture and valid. They are going to be predictive whether you come back. The other thing it was what you want is, you don't want measures that bounce up and down all the time on this stuff. What you get a lot of times with these measures is because you use very small sample sizes. They bounce up and down, they don't actually tell you much. What we want is a measure that's valid and reliable, it predicts what you want it to predict, and it's not what's called, too noisy, which means it just bounces up and down, we don't know why on this. Next step, I've got a causal business model, I have some constructs or measures that I think actually capture the dimensions I care about. Let's verify the linkages, let's see if it's actually true. Because again, we're guessing here. The strategic plan may be based on your intuition or whatever. It is a hypothesis about how things are going to work. Let's verify, if I said A should lead to B, B should link to C, let's verify this, let's do some analytics on this to see if these linkages that you hypothesized actually show up. Again, I want to know why because if they aren't, either I got to change my measures, change my strategy, or figure out what are the organizational barriers that are blocking this from happening. Here's an example, we've got a major fast food chain, company's got 6,000 plus stores in the United States, and they offer both in-store purchases, they also offer delivery. Their overall profitability was not growing enough to meet either their internal or external expectations. Like a lot of companies, this is public. You've got the analysts saying, here's an earnings per share target that we have to hit, they weren't hitting that, and they weren't hitting their internal targets either. What happens, they have a series of meeting, we've got the senior level executives, all the functional areas, we're going to get together and we're going to build a consensus business model. We're going to actually go through this process. Let's take our strategy, let's lay out what we think the consensus business model is and then we're going to come up with performance measures based on that consensus business model. Now, like a lot of companies, and even you in a lot of cases, you have to start out somewhere. This consensus business model was developed using management intuition, there really wasn't any data analysis upfront, we've been in this fast food business a long time, we know our customers, we know our competitive environment, here's what we think it's going to take to meet our financial objectives. Here's their Consensus Business Model. They're going to use something called the customer service profit chain, which is very common in retailers. The idea is, I've got to start out looking at my employees so if I've got very good employees the idea is that is going to lead to a better customer experience. If I got a better customer experience, what's going to happen? Will customers going to buy more, they're going to come back, they're going to tell a friend, and ultimately that is going to get me the financial results I want, which in their case was growth and financial returns. Now if you look at this, there is actually no direct link between employee selection and staffing and financial performance. Lots of intermediate things have to happen, so if I was just going to look at employee selection and staffing and try to use predictive analytics to predict financial results, it's going to be very difficult, there's too many other things that are going to have to happen before you get that end result. The other thing is, it's going to take a while for this to actually go through between me hiring employees and actually getting better financial performance. But in the interim, if I get better employees, that's going to predict that there's more productive, the employees are going to add more value, if I've got better employees who are more productive and work harder, that's going to lead to a better customer experience, which ultimately is going to lead to the financial performance they want. Now again, think about this, it's a fast food company and we've got delivery here. Here being an example, say you got a pizza delivery company. How many times have you actually seen that same pizza delivery person come to your house? Not very often. So, you need to start out with the intuition, why might it be the case that if I have better employees it's going to lead to a better experience and lead to better financial performance. Well, it could be the case that if the person knows my neighborhood, my food actually arrives on time, they get the order right on this. The person taking the order on the phone is actually more productive, can get the order taken faster, they get the right things down. If I got brand new employees, they're probably very unproductive, I have to train them so I probably don't want to have to hire new ones. So, all of that should lead to a better experience. If I liked the pizza delivery person I might order more from them, I might tell my friends this is a great pizza delivery firm. And ultimately, that's when you're going to get what you want at the end. So, that's the idea behind a causal business model. The ultimate financial results were growth and financial returns, but to get there, how exactly were going to get there.? Now, if you look at this diagram, under each one of these round circles which is the basic causal model, there are some more specific things that we think capture that action. That's where the performance measures are. So, if you look at under these you could come up for each one of the things in the rectangles at the bottom you could come up with a performance measure that is related to the business model which is the round circles there, and we can test that. If I start gathering measures for the things in the rectangles, let's see the measures in the blue buckets go up. What happens to the customer measures? Do I see them going up or not? Because that's what we're hypothesizing what we'll happen. So, here's what actually happened based on this model. First of all, if you believe this model, the key thing I care about is employees first. If I don't get that right my entire strategy is going to fall apart. So what they did is take employee turnover as one of their primary metrics for decision making a performance evaluation. And again, this is based on intuition. We just know this must be true. And again, think about employee turnover, it's costly. You've got to hire new employees, I have to do the recruiting. They're very unproductive when you first start. You might have to buy them a uniform. So, employee turnover is very, very expensive. And it impacts the customer experience, because if I have an employee who doesn't know the business or is not trained or doesn't do it well, I'm not going to like going there. So, that became a very important performance measure, it also entered into the bonus plans for the store managers. We're going to evaluate you on overall employee turnover. In addition of that, were going to put in these human resource programs retention bonuses to keep people in place. Because if what we believe is employee turnover is one of the key drivers of financial performance, we want to figure out how to keep them there. Well, what they did is if you stayed there for at least a year, they started paying some of your educational expenses for you. So, now there's an incentive for you to stay there. So, we're going to put the performance measurement, and that's a bonus for people. We're going to put in this expensive program of retention bonuses, and based on this causal model and that's how we're going to run our business. Well, they did that. Spent a lot of money on these retention bonuses, gave bonuses to the store managers based the turnover and then they said "Well, why don't we go back and actually see if this model works". Because we're basing all of this on a prediction that if these measures, the rectangular boxes at the bottom of those circle, at the bottom of that diagram as they go up and down we expect that to translate all the way through to the end to the financial performance. Well, they found out some very interesting things. First of all, you had stores that had very different turnover that had absolutely no correlation with financial results. Well, that was our prediction on some of this. You had some stores that were extremely profitable with very high employee turnover, and you come to these and turnover really was not all that predictive when you started doing this. Now, here's where you want to be careful about. We're going from data analysis to interpretation. Now, it turns out I was one of the consultants working on this. And their first response, and when you'll get a lot of times when it doesn't match intuition, it's got to be wrong. And what they did was they hired a statistician to replicate our work and the statistician came back and says, "Well, that's exactly right. That's what the statistics stay". Now, the question is and a key part of analytics, let's try to understand why. Well, when they actually went out and did some qualitative assessments which I would encourage you to add here, it turns out that one of the reasons why stores with higher turnover might have better financial performance, they had better managers. The better managers were in the store more often, they were actually looking at their employees, making sure they were following all the procedures, they were more willing to fire people because they knew whether they were doing a good job or not. That was part of it. So, it's necessary if you've got a good manager maybe they will fire people faster because they know let's get rid of them or replace them. The other big issue was, it depends on what location you're at and whether you can actually replace employees. It turns out in a lot of low income areas, you don't have anybody to replace them with. So, why would I fire them when it's not necessarily true that I could pick up an employee who is going to be at least as good or better than the one I currently have. So, part of this is going to depend on your labor market. So, it actually had to happen then was, let's do two sets of analysis, one for stores in low income areas, and one for stores in high income areas. So, part of predictive analytics is you need to keep pulling back the onion. Here's the initial results I get. Let's try to understand this which may lead you to more analytics work you need to do. And here it was let's estimates separate models for different socio-economic areas because it has a big impact on what you think the relationship is, has a big impact on what you would set the bonuses at, and a big impact on how you would spend money on this stuff. What actually happen here, the management intuition was partially correct because it turned out after further analysis what they found was, it's actually not overall employee turnover that matters, it's turnover of the supervisors of the store. Those are the people that do all the scheduling, they're the ones who are actually making sure people are following the procedures. So turnover was right, but only turnover of a specific type of employee which was supervisors. So, this is where you want to make sure you got the right measure. Your model was right. Turnover impacts satisfaction, impacts other things but what measure of turnover do you want? Here the measure was turnover of supervisors. So, based on that, and again peeling back the onion, they change this performance measure to not overall turnover, but turnover by employee category. And that's what we're trying to get here. It's not just performance, it's not just turnover, it's which measure of turnover actually predicts something we care about here. Further analysis led them to say, "Okay, now that I know the relationship between supervisor turnover and financials, I can use that to estimate what the benefit might be from the sizes retention bonus. How much am I willing to spend on it? I want to spend too much. But based on this relationship I could say okay, if I can increase my supervisor retention or lower the turnover by 10 percent say, how much am I willing to pay them in a bonus to actually make that economically feasible?" So, it not only changed the performance measured, it changed your estimate of financial performance. And finally, after that again, if you think about that causal model, it's not just employees to financials, it's employees to satisfaction to financial performance. After that, then they started estimating this bigger business model where you start looking at all these linkages. But it's usually a good idea to start small. Let's look at one linkage at a time, before you start trying to estimate all of this.