Ultimately, if we're going to link non-financial metrics to financial performance, we've got to figure out how do we take this analytics results and put them into a financial model. Because ultimately, what we're trying to do is predict if these non-financial metrics go up and down, what's going to be the impact on financial results? So, here's an example of this. We've got a computer hardware manufacturer, major supplier of equipment, and again, senior management, they want to know whether these customer initiatives this company was putting in, actually paid off or not, and more important than that it's not just did they pay off in the past. If I need to develop action plans to increase value in the future, what action plan should I take on? And obviously, what I would like is what's called the biggest bang for the buck. If I put a dollar in something, where am I going to get the biggest return? Which non-financial metrics do I want to improve, and by how much? Now, the first thing you need to think about is, when you do this, what are the financial outcomes that you care about? Because I need to predict that. And it doesn't have to necessarily be profits. There could be intermediate financial results you want. So, if you take something like customer satisfaction, there's lots of reasons why you want higher customer satisfaction. You think it's going to keep the current customers you have, a retention. You think they're going to recommend your product, and not do bad recommendations. You're going to get a bigger share of wallet, a bigger proportion of the expenditures by your customers. I mean there's various reasons why, and you need to lay out when you're going to estimate these models. What are the outcomes that you care about? And ideally what you would like to do is what are the intermediate outcomes? If you think about customer satisfaction again, there is no reason that higher customer satisfaction necessarily leads to higher performance. It's only if satisfaction goes up and you can retain customers, and they buy more and you get bigger share of wallet and they make positive recommendations, that's when it pays off. So, you need to set up, what are these intermediate outcomes that you want if these things go up. So, this is what the company picked. Now, the idea here is yes, customer satisfaction is going to impact these things. So, we're going to try to estimate what the impact is so we can build a financial model. But if you're a manager, customer satisfaction actually doesn't really mean anything. Okay, they're satisfied. What does it mean to say a customer is satisfied? What can you as a manager do to actually change satisfaction, and which dimensions do customers really care about? So, here's where you have to put the front end of the model on this. Customer satisfaction is in the middle. What are the things that customers care about that's going to increase their satisfaction in such a way that they buy more from you, that they stay with you, that they make positive recommendations? So, here what the company did is let's go out and do some marketing research. Let's do some focus groups with customers. Let's see what they say would cause them to actually buy more from us. And there's various things that could happen, right? Do I think I got a good relationship with my supplier? Do I trust them? Do I think the value is high? Something like a personal computer, right? Do I think the thing actually works? Can I change over when I put the new software onto this program? One of the big things is can I take my old files, and push it over there? So, what you want as a manager when you estimate these models, yes, I want the financial results. But what I want to know is, what are the specific actions I can take of these over on the other side that will impact satisfaction in such a way that my financials go up, and where is the biggest bang for the buck? Of these many dimensions I have over on the left hand side, if I had a dollar to invest, which one would I want to invest in such that I get the biggest return? And that's the model we're going to estimate. So, what we have here is basically a regression model. First of all we're going to say if satisfaction goes up, how much does retention go up, how much does word of mouth go up, how much do these ultimate financial outcomes go up? On the other end you've got which one of these dimensions are drivers actually cause customer satisfaction to move up and down and by how much? So, these are the estimates we have here. Now, if you look at this diagram, we've got two things, we've got things inside circles, and we have little numbers next to arrows. So, here's what these things are. The numbers in the circles, that tells you how is this company doing on a zero to 100 scale on that dimension. 100 being the highest. So, if I get a 70, I'm doing 70 out of 100, if I get 80 on that dimension I'm doing 80 out of 100, or 80 says I'm doing better. The numbers next to the arrows, those are the coefficients from a regression model. Basically what it says is, if I can increase the score on those drivers to the left by one unit, here is the impact I'm going to have on my dependent variable. Either on customer satisfaction, or if I'm estimating economic outcomes, here's how one unit change in satisfaction impacts these economic outcomes. So, here's an example. If you take those numbers, if you increase the relationship score by 10, what this says is the customer satisfaction score is going to increase by three. You take that 0.3 times 10, so it's going to go up by three. Now, if I can increase my customer satisfaction score by three, what this predicts is it's going to increase my retention score by three. From 73 to 76. Now, based on that if I can figure out what's the financial benefit from keeping a customer, I can come up with a financial model. Now, again if you go back to the diagram, something I always ask my students is if you had a dollar where would you invest it given this? Now, a lot of times what people say is well invest it in trust. Because that coefficient, that number next to the arrow, is biggest for trust. It's 0.8, which says if I can increase my trust score by one point, my customer satisfaction score is going to go up by 0.8. That seems to make sense. But then the question is, do you really believe you can increase your trust score by one point or 10 points, and how much is going to cost you? It may be the case. Remember, these are personal computers. They trust you as much as they're going to trust you. Okay. They don't trust anybody perfectly. It may be extremely hard to move the trust score above 0.8 where you're already doing fairly well. It also may be really expensive. Alternatively, if you look at your shopping score, you're doing so badly on that one. Something in the 60s, and it may actually be easier to go after that because if you're doing badly it might be easier and cheaper to improve that. So, even though it has this coefficient of 0.1, if it's easy to move that one by ten points, and very hard to move trust by even one point, you should probably be going after the shopping. So, what you need is both that model we just estimated which tells you the benefit from moving the shopping or trust score, and what do you think the cost would be to actually move this, and how easy is it to do that? Now, based on that, you can start coming up with a financial model, which is ultimately what we want. If I increase trust, or if I increase shopping, would do I think the ultimate impact is on retention, recommendations, and the other financial outcomes that we care about. Now, again, this is looking a straight linear relationship. It just keeps going up in a nice straight line. But we found from some of the earlier analyses that a lot of times, there's non-linearities. It doesn't keep going up in a straight line. In fact, maybe you want to be really good at customer satisfaction or maybe you want to stop somewhere like 75. So, here's where you want to dig down deeper. So, here's what the company did. Let's not stop there. Let's look a little further. Do we have any of these non-linearities? It turns out this is one of those companies where you really want to increase satisfaction. Because what they found out is, if you could increase that retention score to 90 or above, the customer bought the same brand again 56 percent of the time. If it was below 90, they only bought it 30 percent of the time. That suggests you really, really want to spend money to get customers above 90. That's when they start buying a lot more of this stuff or they stick with you. Same thing with the recommendations score. If it was 90 or above, they recommended about one and a half times. If it was below 90, it was less than once. Those are fairly big changes. So, here's a company where I would say, if I can push my customers above 90, there's a huge payback because there's a big difference once you get past this 90 in terms of whether you're going to keep them or recommend them. So, now the question would be, I've got to pick action plans because ultimately, that's what I want. Can I use analytics to pick out which action plans are going to result in this higher financial performance? So, based on this, what you might say is, let's try to pick action plans that are going to move customer scores above 90 because that's where the big payoff is, if I can move them above 90 on here. So, let's make an assumption. And you would have to do some market research to figure this. Let's assume that you could probably move a customer score by a maximum of 10 points. And that's probably true because customer satisfaction scores are really hard to move. They're easy to move down. If you had a disaster, people get less satisfied. They're really fairly hard to move up. So let's say the most on a 100 point scale you could move them is 10. What that would say is if I want to move people above 90, there's only a small number of customers that I could move 10 points that are going to push me above 90. So, at that point, what you want to do is figure out which customers do I want to focus on. It's not all the customers. What I want are the ones that I could do something that might move them 10 points and move them over this 90 threshold. That's where you're going to get the biggest bang for the buck. So here's what you're going to do. If you start looking at all the customers, and here's just some diagrams of how many customers you have at each score, the people you want to focus your action plans on, the one based on the analytics you're going to get the biggest bang for the buck, are the people in the red bars. Those are the people where I can move them about 10 points over that 90 threshold. And based on my analytics, I believe that they're going to recommend a lot more and they're going to stick with me. So, roughly, it's about 16 percent of the customers you want to focus on. Not all of them. Now, based on that, let's see if we can come up with a financial model. So, what we're going to do is, say, okay, if I focus on these 16 percent of the people, how much is it going to cost me to put a customer initiative in that's going to move them up 10 points? If I move them up 10 points, what do I think the financial benefit is from retaining them and from having them recommend people more? Well, the way we can do that now is take the analytics and basically, we're going to do a little Excel spreadsheet. We're going to come up with what's called a net present value. Again, net present value because some of the benefits are going to come out in multiple years, and I need to discount those back because I would rather have a dollar now than a dollar in the future. So, here's the little spreadsheet we're going to have here. So, here are some of the assumptions we're going to make. First of all, and again, you'd have to base this on what your company does. Let's assume we've got a five-year time horizon. What happens now with satisfaction is going to have no impact after five years especially with something like technology. New players come in, new technology. Let's assume we have a 15 percent discount rate. There's a cost to capital in your organization. How much do I have to go out and borrow the money from? How much do I have to pay to get money from my shareholders? Let's assume 15 percent, which right now is very high, but we'll use it for argument's purposes. Now, the margins on a personal computer are pretty small. Let's say, on average, you could sell a $2,000 high-end personal computer. The margin you're going to get out of that's is $145. You're not going to make much on each one of these sales relative to the selling price. So, let's assume what we're thinking about is, can we invest five million dollars in some kind of customer satisfaction improvement initiative, where we've got a five-year horizon and 15 percent discount rate? Let's see what the benefit would be there. Now, here's the estimates we did before. We estimated that 25.99 percent improvement in retention if I could move people over 90. We estimated a 0.65 change in recommendations if you could move them above 90. Now what you need to think about is, of those people, how many do you think you could actually move 10 points? Because obviously, you're not going to be able to move all of them. So, here's why you want to do market research. Based on some more analytics that you do with market research, what's the odds that we could move somebody up at least 10 points? If you think it's 20 percent, 20 percent. If it's 30, it's 30 percent. Because that's going to have an impact on whether you think this works or not. So, based on those assumptions, which are in the spreadsheet, we're going to compute the net present value. Again, with the net present value, you've got cash going out. Well, that's the five million dollars right now. Then I've got cash coming back in. Well, the cash comes in either because I stick with you when I buy your computers later or I recommend to somebody else and they buy it. And again, based on this analytics in market research, we can estimate that. And based on that, we can come back and just with this simple scenario, you get a huge payback. Now, obviously, these are all estimates. But the nice thing here is, you use the analytics to start the estimates. Based on that, you come up with a spreadsheet. Now, you can start doing a lot of what-if analysis. What if I change my discount rate to only five percent, which is getting closer to what it is right now? We have almost no interest rates. What happens if my margins go up to $200? What happens if they go down to 100? So, once you set up this spreadsheet that's based on the analytics, you can do a whole bunch of what-if analysis by changing some of these parameters. Saying this is an estimate, how comfortable are you if this estimate is off by 10 percent, positive five percent or negative? And come up with a comfort zone. Based on that, how comfortable are you investing five million dollars to improve this dimension? But it all starts with the analytics because you have to come up with some parameters to start estimating these financial models. Then you can start doing the what-if analysis on this.