Specification of Requirements

We can now start making our specification of requirements for the performance management dashboard. The goal is to drive 60 miles in 60 minutes. We can now place in our budget a goal line, as shown in Exhibit 3.2, which is a straight line and a function of time. In other words, we choose a goal that is to drive with the same speed all the way. To do that, we must be halfway through after 30 minutes.

Our KPI must specify key elements of the performance and give us an idea of the degree of success with the project. An obvious choice for KPI will therefore be the relationship between what we have achieved and what we plan to achieve.

Visually, this means that the graph with actual miles is lying above the target curve, when our KPI is more than 1 (see Exhibit 3.2).

In addition to the graph, we could set up a “cockpit” or performance management dashboard, consisting of a number of simple indicators for the process. Here we have made a status indicator showing our current KPI, and this is more than 1 when the status line is over our target line.

We have also added a trend meter, which points downward if the speed in the current period is lower than the speed in the previous period. The situation at the black dot in Exhibit 3.2 is therefore that we are doing well overall, but that we should be aware that we are losing speed. Further, we have added a smiley face on this dashboard with information about whether the summerhouse will be reached on time given the current location and acceleration. This last KPI is illustrated by a smiley that is happy, neutral or unhappy, depending on a projection of whether we will reach our destination on time, might reach our destination on time, or won't reach our destination on time, based on current statuses. At an overall level, we just have to keep an eye on the smiley.

Technical Support

So what do we need in terms of technical support to realize this specification of requirements? Exhibit 3.3 shows a section of our base table.

Our data in the start time, budget, and time columns is fixed before we begin the trip. These values are static. The column with actually driven miles is updated on an ongoing basis by program code that reads the number of driven miles. A new figure is added to the table in the column with actual miles every minute. Then the graphics on the data‐driven and dynamic dashboard are updated. All the data in the table is read every minute to the graphical object that shows the curve and actual and expected miles. Then our KPI is computed by dividing the latest number of actual miles by the number of expected budget miles, and the change can be seen in real time on the dashboard, along with any replaced GIF arrow (graphic image of an arrow) and smiley. If we drove faster in the previous minute (actual miles/time) compared to the minute before, the GIF arrow pointing upwards is loaded. If the latest KPI is computed to be more than 1, the happy smiley is loaded for our performance dashboard in Exhibit 3.2.

Off We Go to the Summerhouse

We start our journey, and the first couple of points on our status curve in Exhibit 3.2 appear on the dashboard after a minute, along with the other graphics. We're driving in the city and are therefore under the target line. Our KPI is under 1, and our smiley is unhappy. Our trend arrow, however, is pointing upwards most of the time as we slowly, minute by minute, increase speed on our way out of the city. Performance monitoring encourages us to drive faster, but that just won't do on city roads—and isn't that annoying! Once on the highway, we finally speed up and move over the target line; the KPI is now over 1. Out smiley becomes neutral and then happy, and the trend arrow is still pointing upwards, as we continue to increase our speed. But then we run into traffic on the interstate. Our KPI falls back through 1, as we're now getting under the budget line. Then the smiley is unhappy and the trend arrow begins to point straight ahead, as we have to stop the car!

However, the traffic quickly dissolves, and we increase speed significantly to get to our summerhouse on time. Our KPI goes from 0.9 and breaks through 1. The smiley is happy again (just as we are), and our trend arrow has pointed upward ever since the traffic became lighter. When, after a while, we leave the highway, our KPI is 1.1 (10 percent over target or budget). For the last bit of the trip, we'll be driving on smaller roads and our speed will therefore fall. The trend arrow points downwards, the status speedometer slowly drops towards 1. But we're feeling optimistic, because we know that we have enough margin for the last part of the trip, as the smiley and KPI both show us.

By means of the above example, we've tried to illustrate the idea behind KPIs and performance management. The example may seem trivial, but it does provide a useful insight into key concepts and how to monitor a business process.

Lead and Lag Information

The summerhouse example also gives us an understanding of the two types of information used by the BA function—lead and lag information.

Lag information is retrospective information, which we choose to register on an ongoing basis in our data warehouse in connection with performance management. In the summerhouse example, the lag information is the actual number of miles. Lag information is typically stored in tables in the business's data warehouse and is used for analyses to create a learning loop back to the strategy (see Chapter 2 on strategy) or for new lead information.

Lead information has a completely different character than lag information. Lead information is used to improve existing business processes or initiate ones. Lead information in the BA framework is typically created on the basis of an analysis of lag information and is therefore usually not stored in tables, since this information, as already mentioned, is the outcome of an analytical process. Lead information will typically have the character of “breaking insight,” which can be used to improve overall business processes and provide learning loops back to the strategic level. An analytical process using, for instance, a data mining methodology on our base table in Exhibit 3.3 (naturally, after we've done the trip to the summerhouse several times) would be a useful tool for uncovering key factors to provide us with knowledge about why we tend to arrive at the summerhouse early () or late (). This knowledge will, in future, help us arrive at our target on time, thereby achieving success. Our breaking insight, which is the outcome of these analytical processes on our historical data, could be a statistically significant correlation between the value of our KPI, when we reach our target, and the start time of our journey. The correlation is illustrated in Exhibit 3.4.

The trip to the summerhouse is usually successful if we start driving before 2∼PM or after 7∼PM (). The worst time to start is between 4∼PM and 5∼PM. If we start within this time interval, our chances of reaching our target on time are minimal. The explanation is, of course, that it takes longer to get through the city in rush‐hour traffic, and we'll almost always end up in a slow line on the motorway. Our breaking insight or lead information, which we could also call our critical success factor, will be: We must start our trip to the summerhouse before 2∼PM or after 7∼PM in order to optimize/improve our operational process and be successful in our endeavor.

Note that the new important lead information identified by analytics obviously works to provide a learning loop back to the strategic level (see Chapter 2) to be used next time a strategy is developed for the coming year. In this section, we'll take a closer look at what KPIs are, how they are generated, and what they can be used for. The creation of KPIs is normally intuitive as illustrated in the above summerhouse example.

Generally speaking, KPIs describe the relationship between the organization's activities and its main objectives. KPIs can be financial key indicators, index figures specified for the occasion, or other SMART (specific, measurable, agreed, realistic, time‐bound) objectives. What is required of KPIs is simply that they on the one hand set some standards for how business processes must perform (lag information) and on the other help us define which activities have “gone wrong,” if the process does not meet its objectives. This means that if we have a KPI, and we are below target, we always know which consequences this will have in the long run. This knowledge enables us to adjust activities and thereby ensure that the overall targets in the corporate strategy are achieved.

KPIs therefore work as warning signals. Generally speaking, if some KPIs are not achieving their targets, we must look into why. Is it a question of a lack of strategic focus (i.e., the organization for some reason is not focused in its efforts to meet the strategic objectives). Is it a case of correct execution of the desired activities, but with a lack of competencies or resources, which means that the activities do not reach the desired level? Did something change that we did not plan for, or was the strategic target too stretched, as they often are? Every year companies plan for growth above the market growth, else they would be planning to lose market presence. But not all companies can grow more than the market.

Another important function of KPIs is that they are able to stop activities again. It is not uncommon for CRM departments to have to take on many troubleshooting tasks. When we face a problem, we solve it by starting a new and corrective process. But when do we stop these processes again? If we fail to do so, the organization's CRM strategy will become a patchwork quilt of historical troubleshooting exercises. If we are constantly patching things up, more and more resources will be needed over time to maintain these stopgap measures throughout the organization.

When systematically collected, KPIs also provide the organization with a memory, which means that learning can be derived from successful projects. This learning may come by means of analytics, which we will cover in the following section, but also by holding people to their promises. It is quite a common phenomenon to have people who are extremely good at convincing management that they have a great idea for a campaign. And then there are people who make great campaigns. As the two are not necessarily the same, measuring KPIs will tell the organization which campaigns are working. In the long run, this means that we get an organization where the focus is on results, rather than on what sells internally.

More about Lead and Lag Information

As mentioned in the introduction and in the summerhouse example, BA often distinguishes between lead and lag information, where lead information is the knowledge we need to initiate or improve a process. If we take our point of departure in our trip to the summerhouse, there are two possibilities; either it's our first trip there on that route, or we've gone that way before.

If it's our first time on that route, we must initiate a new process, because we're doing something for the first time. This also means that we have no historical knowledge about how long it takes to take that route, and we therefore have to plan our trip based on other information, such as directions from the Internet or general experience with how long that kind of trip takes. What we're talking about here is lead information, the information that will get us to the summerhouse using the correct route with arrival on the correct time. Therefore, it is information that we need to have before we start our trip.

As we are driving toward the summerhouse, we receive a large amount of lag information. The nature of lag information is that it monitors our process. We can react to it and adjust our actions by driving faster or slower, but it will not change the actual process we are in, based on this information. This is information we collect and use in the course of the process.

If we get fed up with the route we've chosen and want to find a new one, we have to start looking for new lead information to find out whether there is another route that may be quicker and easier. If we then choose to go by the new route, we will again start generating lag information, based on which we will create expectations to whether we reach the summerhouse on time. If it is not the first time we take the given route to the summerhouse, we already have knowledge about the usual course of the trip. We have, in other words, lag information telling us how long the trip usually takes, whether the traffic is different at different times of the day, week, and year. Based on the lag information we're already in possession of, we can generate new lead information because we know when we want to reach the summerhouse, and how the traffic usually is at the given time. We can therefore count back and find out when we need to leave, and plan whether there is time for other activities before we go. We can, in other words, learn from our internal knowledge and optimize the process, which is the trip to the summerhouse. This is exactly what we do in connection with process optimization, where we are not just using lag information to monitor whether the process meets its objectives. Rather, we are also saving this lag information for future analyses to improve this process via lead information that has the character of being breaking insight.

Since the subject of this book is how to optimize business processes based on strategic requirements, we have chosen to include two perspectives. One perspective is the establishment of a process for the first time, which includes identifying which lead and lag information is required in the organization, so that we can initiate and manage the given process.

The second perspective is taking its point of departure in a strategic demand for the optimization of given business processes. Since this process is already established, we can use saved lag information describing the correlation between the process, the way in which it has been influenced, and the effect it had on the process, and derive knowledge about how we can optimize the process. So, based on lag information, we can generate lead information, and if the nature of the process is not being completely restructured based on learning from the new lead information, this learning cycle can be maintained.

In terms of our trip to the summerhouse, we can continue to cut back on minutes and seconds of the drive while potentially minimizing petrol consumption, if we measure that, too, and thereby improve our process on an ongoing basis. This is a case of optimizing processes by improving the use of resources (less petrol consumption, for instance, if we observe traffic regulations and do a bit of shopping, which should otherwise have been done separately) and optimized user satisfaction with the process (the fact that we arrive exactly on time without stressing, and maybe get out and stretch our legs on the way, if that is what the users ask for at the beginning). The reason for separating the two is the fact that passengers in the car don't always want to get out to shop, or to start the trip early and drive at 50 miles per hour to minimize fuel consumption. Passengers are not likely, either, to appreciate the excellent service it would have been if we could get them to the summerhouse in half the normal time by driving 160 miles per hour. The same thing occurs in a restaurant where service has been cut back too much, or the level of service has been raised so high that we don't want to pay for it. To optimize resources in a business process, it is necessary to take user satisfaction into consideration, which is a fundamental rule in performance management.

Level 1: Identifying the Objectives

Imagine working in the marketing department of a large telecom enterprise. A marketing department has two overall purposes: to attract new customers via campaigns on television, in magazines, and other media, and to hold on to the existing customers.

To achieve good customer relations, many businesses adopt dialogue programs that inform customers of new prices, stores, products, and so forth. Some of this communication has the purpose of educating the customer, but most of the communication is aimed at expanding and strengthening the customer's involvement. For instance, a telecommunication provider could teach customers how to use social media and encourage them to listen to music or surf the Internet via a mobile device, thereby creating more data traffic and potentially higher income per customer. Another way of expanding a customer's involvement is to make sure he or she does not jump ship and join the competition. Finally, imagine that this is December and that the boss has just given us next year's target for our department (see Exhibit 3.6).

The targets are: At the end of the year, the telecom company's customer base must be 10 percent bigger, and the average income per customer must have gone up by 10 percent.

Level 2: Identifying an Operational Strategy

Since we are responsible for CRM and not customer acquisition, this means that the strategy must be based on the creation of growth in our customer base by becoming better at holding on to them. As the one who is responsible for CRM, we know that if we can hold on to our most valuable customers, average income will also go up as a result of a more valuable customer portfolio. Therefore, we decide as a starting point to concentrate on retaining the most valuable third of the customer base. This is based on the fact that the company has already performed a value‐based segmentation, dividing its customers into gold, silver, and bronze customers, and that each of these segments constitutes about a third. Based on this, it is relatively easy to determine the average value of the customer base, if we were to add the additional 10 percent gold customers, as is our intention. After some calculations, we find that if our gold customers grow by 10 percent, our average revenue per customer will go up by 5 percent. As the average market growth is 3 percent, we still need a strategy for how to create growth in the average revenue per customer of the remaining 2 percent. We decide to create the 2 percent via added sales to our existing customers. All this is strategic lead information, as it is used for shaping our overall CRM strategy.

As illustrated by Exhibit 3.7, there is now a two‐part strategy: Retain the gold customers and initiate added sales activities to our customer base.

Level 3: Identifying the Critical Success Factors

Defining the critical success factors before starting on a project is always subject to discussion, and that discussion may continue throughout the project. In this context, we take the BA perspective and focus only on what may be of importance to our information strategy.

This step in the process toward an information strategy is based on the fact that when we implement a strategy, we initiate a large number of activities, and some of these activities are more critical than others. It is, for example, key to retention activities that we find out why customers leave our company, and that we develop retention offers that are at least as good as those of our competitors.

In addition to this, our budgets tell us that we cannot afford to send out valuable retention offers to all of our 100,000 gold customers. It is, therefore, essential to our CRM strategy that we obtain knowledge about which customers intend to leave, when, and why. With that knowledge, we only have to contact customers who are likely to leave. We want information about when we need to contact this group of customers, as well as knowledge about which offer we must give each individual customer. If we hold in one hand the knowledge about which of the customers intend to leave, and in the other an effective retention offer, then we have some excellent tools for carrying out a retention campaign. It is therefore a critical success factor that we can offer the right customers the right retention offers at the right time. Otherwise, our retention campaign will fail.

It is the same thing with cross‐selling activities. We want to communicate only offers in which our customers will have an interest. For example, we don't want to spend resources promoting the use of social media on the move to our fixed‐line customers, because they'll ignore it at best. At worst, they'll be annoyed that they, our customers, are paying for and spending time on misplaced communication.

It is therefore a critical success factor for cross‐sell activities that we know which customers can be assumed to be interested in the various added sales offers. In other words, it is a critical success factor for our cross‐sell campaign that we are able to give the right customers the right offer at the right time. Otherwise our added sales campaign will fail (see Exhibit 3.8).

We will call this tactical lead information, as this is the knowledge that we need to design our campaigns.

Level 4: Identifying Lead and Lag Information

So the knowledge we're after in connection with our customer retention strategy must answer the question: Which customers are leaving us, when, and why? Once we've got that knowledge, we can carry out campaigns with the right retention offer for the right customers at the right time. At the same time, we want to fulfill our added sales strategy. We want to know which customers will buy what and when.

All this is lead information. In other words, it is information or knowledge that is necessary for even beginning new business activities. We also want to collect lag information because it's important to be able to monitor the processes to see whether we are going to fulfill our strategies. If it looks as if that will not be the case, we want to be able to act as quickly as possible to make adjustments.

Therefore, we want to receive information on an ongoing basis about how the individual campaigns are going. Are there some of them, for instance, that are doing better than expected and could therefore be rolled out further, or are there some that should be canceled altogether? Of course, we also want to receive continuous information about the size of our customer base along with the average income per customer. All of this is summarized in Exhibit 3.9. In regard to the operational information, let us wait with this for a while until we, based on the lead information, have designed how the operational processes should look in the first place. Then, afterward, we can look into which operational data we need to run this new process.

In Exhibit 3.10, we illustrated lead and lag information seen from a process perspective. We want to repeat that lead information does not necessarily have to come solely from the BA function; it just does in this example. There will be cases where the BA function is unable to support the decisions, for if there is no relevant information to support the solution of the problem in the existing data warehouse or other data sources, there is no point in involving conventional data warehouse analysts as part of the project until the need for lag information is discussed. To further exemplify the difference between initiating new processes and monitoring existing ones, it is said rather wryly in controller/accountancy environments that “possibly the most valuable person in the entire organization is the one who is able to start up new and relevant business initiatives (via lead information). But the second most valuable person in the organization is without a doubt the one who is able to stop all the wrong initiatives (via lag information, which the controllers themselves are managing).”

Our department can now carry out the new business initiatives, and the BA function can support with lag information that informs us on an ongoing basis about whether the process is meeting the defined objectives. This information is typically delivered as conventional KPIs to management and process owners, more frequently and in more depth than lead information.

For details about how to generate information about which customers will leave, when, and why, see the “Data Mining with Target Variables” section in Chapter 4. Here we will, among other subjects, explain decision trees. These trees show us, customer by customer, what risk there is of him or her canceling their customer relationship in the coming period of time. In addition to this, the trees enable us to interpret the reason behind the given risk profile, and thus also what can be done to retain the customer.

In the section about data mining, we will also be looking at cross‐sales models, which identify consumer patterns based on historical information. Based on these, we will suggest, for each individual customer, what he or she should be offered and when. Finally, the Web site BA‐support.com contains a case study that describes how one telecom company, via analytically driven CRM strategy, went from a significant financial deficit at the end of one year to an even larger profit the following year. In that case, data mining was a driving force for the entire project.

As is evident from the case study, it often makes sense to analyze whether lead and lag information is on a strategic, tactical, and operational level. The following example will show these distinctions in more detail and see some examples as they could play out in a digital marketing department.

When we refer to lead and lag information on a strategic level, we refer to it on a marketing strategy level. So the question would be: Which information might be needed to create a digital marketing strategy? The strategic lead information. One of the first questions to answer is what the overall objectives are; if they are not clearly defined by the company strategy, look into whether sales turnover should be supported through getting new customers, selling more to the existing ones, or focusing on customer retention. We would also look into whether we want to focus on specific customer groups more than others. A digital marketing organization would also like to look into which campaigns in the existing campaign landscape yield the highest returns on investment, and whether we should look into optimizing our digital marketing landscape. Other sources of decision support could come from customer feedback, competitive analysis, or reputation reports purchased from marketing agencies. All of this information would give us a good idea about how the overall marketing plan for the next period should look, and based on our planning, we will set some expectations for the results of this marketing plan. These results often call for KPIs, and they could have to do with how many new customers we aim to acquire, how many products are sold, the expected earnings per product category or segment, or the like.

To see whether we achieve our targets, we will start measuring for these KPIs. Are we getting the amounts of new customers that we were planning for, or selling the amount of products that we hoped to? This new information, what we will call strategic lag information, is needed to monitor and steer our overall marketing strategy over the period to come.

Tactical lead information would in this case be information about how to create and optimize our individual campaigns and marketing activities. So let us say that we wish to create a new digital campaign executed through our multichannel or omnichannel marketing system. We would look into who should receive this campaign, when the campaign should be entered into our landscape of ongoing campaigns, and whether it should be time limited. Also, we would like to examine the campaign cannibalization effects, since some of the customers who before would have received campaign A will now instead get the new campaign B. Hence, we must also expect to sell less of what campaign A promotes. When all this is done, we can adjust our expectations to our individual campaigns and create a new set of KPIs. This is what we would call our tactical lead information.

As in the strategic case, we will also start creating lag information, which is the information that will tell us whether our individual campaigns lives up to our expectations. Next to campaigns result measure, we might also set up other performance measures, such as how many of the campaigns people click on or respond to. We might also create some test groups to benchmark our campaigns against. Some of these test groups do not get the offer in the first place, so we can see the overall lift of our campaigns. Some of these test groups we base on random customers and compare them against the targeted customer segments, so that we can see the effect of our targeting, whether it may be analytically based or simply based on subjective assumptions. We might also want to track our budget spending per campaign, to make sure that we allocate our marketing budget correctly.

Operational lead information would be the concrete lists of customers to contact via e‐mail and which offers we should make, or the models we deploy into our real‐time campaign landscape that place the right banners on the right customers' personal pages; additionally, we may direct the call center agents in regard to what should be promoted to the individual customers when they call in.

Similarly the operational lag information would be the day‐to‐day results per sales agent, how much time is used per sale, or how the customers responded to banner ads or e‐mails.

Concept of Performance Management

First and foremost, however, we want to introduce the concept of performance management and what it covers. Very briefly, it can be described as the optimization of processes. As a point of departure, a process can be optimized in two ways: first, by ensuring a better use of resources deployed in keeping the process going, and second, by improving the result of the process. This is also the essence of what is known to some as Lean, in which we try to reduce waste in our processes, or Six Sigma, in which we to a larger extent seek to minimize the variation in our processes in order to make better use of our resources and provide our process users with a more predictable output.

Being focused on the optimization of a process does not mean that the strategy is irrelevant, because the requirement for an optimization might well be derived from the strategy. Besides, strategy will always restrict our scope for what we can implement. If, for instance, we are an organization that competes on good service, there is a limit to how long we can let our customers wait when they call us. Firing some of the staff in the call center might be an easy way of saving money, but it would create an imbalance in relation to customer expectations. This balancing is at the very core of performance management (i.e., the optimum weighting of the resources that are used in a process in relation to the expected outcome of the process). In this call center example, we can cut back on staff to the extent that we are making our customers unhappy. On the other hand, a time will come when the number of staff is so high, with costs to match, that customers do not notice the difference and do not appreciate the increased costs, which they will ultimately be paying. It's important to acknowledge, too, that good service is an expected quantity. On the one hand, we expect to line up at the post office and the discount store, but on the other, we would find it unacceptable to have to line up for an ambulance after a traffic accident.

In the example, we must imagine that in connection with the telecom company's strategy, two requirements have been put to the organization. The first is that the CRM department has interviewed customers who have canceled their mobile phone subscriptions. Based on these interviews, it was found that a large number of the customers are dissatisfied with the service they received from the call center because they had to call several times to get a problem solved.

The other strategic requirement to the organization is that the call center needs to reduce its overall costs by 10 percent without compromising the level of service given to customers. The strategy function has computed that if the two initiatives are successful, they will contribute significantly to the bottom line. The reason for this is that it is expensive to acquire new customers to replace the ones who left in frustration, because they must be given financial incentives, they incur initial set‐up costs, and bonuses must be paid to the stores and channels that land these new customers. All in all, the cost of landing a new mobile phone customer is approximately U.S. $400 in Denmark at the time of writing—money that is wasted if the customer cancels his or her subscription because the company must then reinvest in order to maintain its market share. The choice of the call center as an area to save money is based on the fact that it is a salary‐heavy department, and it is thought that there is room for improvement in this area.

So imagine in this example that we're working with a call center in a telecom company. The purpose of a call center is to deal with inbound calls from customers as well as to make calls to customers. Incoming calls are generated by customers who have questions or want something done. This may concern their bills, setting up their mobile devises, subscription cancellation or the like. Outgoing calls may be concerned with campaign offers or the answers to questions that couldn't be answered when the customer called.

If we take a closer look at the process for incoming calls, they start with a customer calling the call center. He or she might be waiting on hold and subsequently transferred to an agent. The agent then needs to clarify the nature of the customer's request or problem and deal with it. The customer is then expected to hang up, and the request/problem is expected to have been met or solved. During the entire process, automatic logging is performed. It records the time of the call and which phone number called in. The agent also logs what the problem is and whether it has been solved, or whether someone else must get back to the customer later.

Sometimes a customer calls to have a problem solved, but it is not possible. This becomes clear when the same customer calls back soon after with the same problem, and the agent then has to deal with the customer again. This has two consequences. One is that further resources are used to deal with a “repeat call,” and another is that there will be a customer who is not satisfied with the advice received to begin with. If we are therefore looking at optimizing our process, we are presented with an opportunity to minimize the resource input and optimize customer satisfaction if we can minimize the number of repeat calls.

To reduce the number of repeat calls, we need to obtain some information: Which agents generate many repeat calls, and in connection with which problems? With this knowledge we know who to train and in what. To be able to measure the number of repeat calls, we must define that as a KPI. We also need to set a goal for the KPI, which could be a 20 percent reduction, to be able to achieve both strategic targets. We won't discuss why we've set the goal at 20 percent and not more or less. Let's just assume the figure is an estimate based on various calculations.

We can turn the number of repeat calls into a KPI because we know which phone numbers ring in at which time and with which problems. We can therefore define a repeat call as having happened if the same phone number rings in with the same problem within one week. Here we are assuming that the problem was not solved in the course of the initial call. We also know which agent did not solve the problem the first time around, as we can connect an agent name to each call. We will also make a measuring point: How many calls does this agent take on an average basis per hour? This is because the total number of customer problems that are dealt with must not drop as a result of our future activities.

In a couple of days' time, an analyst can now produce two types of reports for continuous production, which will constitute our lead information. One report will focus on which types of problems typically generate repeat calls. Is there a specific type of problem that is generally difficult to solve? If this is the case, we can train all staff in how to solve these, and update internal manuals. The other report focuses on which agents generate many repeat calls. The recipient of the report is the team manager and agent him‐ or herself, so that an individual training program can be set up. To motivate the agents, a bonus program is introduced that rewards staff who generate relatively few repeat calls, while still dealing with relatively many incoming calls.

As illustrated by Exhibit 3.12, we have now established a considerable number of measuring points and related them to each other. We'll start with some of the activities at the bottom of the model, which are built on employee interviews, training, and bonus systems. These activities influence subsequent processes. The improved processes mean that not only is the call center making better use of its resources, but customer loyalty is up. These developments, then, seem to render some positive financial results for the company. But is this actually the case? We think so, but we don't know.

This is why, when using performance management, we are usually not content with just having the measuring points at the time when we need them; we save them, too, for future analyses. For is it actually possible to measure a connection between the general training of call center staff and fewer repeat calls? Or is it solely on the basis of the individual training that we can detect a correlation between training costs and fewer repeat calls? We are therefore very interested in identifying which activity gives the best effect per invested dollar. It is equally valuable to know whether the company via its activities is also achieving its target of reducing call center costs and the number of unhappy customers who cancel their subscriptions. We also want to know how long it will take for the effect to be noticeable. All of this can be linked to financial goals to analyze which activities render the biggest return.

The objective of performance management is therefore also to systematically accumulate experiences based on performed activities by systematically saving and analyzing lag information. This puts us in a position where we can obtain detailed insight into our own processes, an insight that in time means that we can gain a more holistic picture of our organization. What is the correlation between the profiles we employ and how they perform in a call center? Do students perform best? Or do we want to focus on the older generation, because we know they'll stay in the job longer? However, are they performing as well as the younger people, if we give them individual training? And do our bonus systems mean that the employee with the biggest bonus stays longer, which would reduce the overall costs of hiring and thereby contribute positively to our HR budgets? All these questions are about striking the right balance. On the one hand, balance is about minimizing the resources used to keep a process going and, on the other, about ensuring that the process meets the user's expectations.

Customer Relationship Management Activities

CRM activities are one of the processes that historically has been supported by BA. There are several reasons for this. First of all, these are processes that can visibly add value in the very short term, and that own a lot of information— that is, customer data. Primary industries in this field are organizations with stable and long‐term customer relations, such as banks, insurance companies, and telecoms. The names of individual customers are known, and it is possible not only to measure their consumption month by month but also to identify who cancels his or her customer relationship. In the following example, we will take our point of departure in the three focus areas of CRM: getting valuable customers, increasing the value of existing customers, and keeping customers. Digitalization in the form of apps or purchases from the Internet portal, and the subsequent ability to generate customer data and use it for customer dialogue, will however mean that more companies will focus on CRM data. The fact that customers increasingly expect that companies know them and only communicate relevant information to them will also drive the importance of the use of customer data in all channels.

All in all, CRM is however about optimizing a customer's lifetime value, which equals the average consumption times or the number of months this person is a customer, minus costs that are associated with getting this customer. See “the whale” in Exhibit 3.14. Note that the three types of activities—get, increase, and keep—are something companies have always done, but when the company grows to a certain size, their customers become an undistinguishable mass. Analytical CRM can, so to speak, color customers red, yellow, and blue, and on the basis of these colors, we can carry out individualized initiatives designed to meet the needs of different customer groups. To read more about the whale concept, please go to BA‐support.com or examine the book Business Analytics for Sales and Marketing Managers.

When we talk about getting valuable customers, we assume two things: that low cost is associated with acquiring the given types of customers and that these customers generate high revenues. A typical analytical technique can therefore be an analysis of which customers have the highest lifetime value: average consumption times average number of months as customers. Next we need to find out through which channels we got the customer, and which campaigns attracted him or her to penetrate this segment further. This is an analytical method used by telecom companies when they wish to optimize earnings from prepaid subscribers (such as those they get from scratch cards, where the customer's name is not known). In this context, we are interested in learning from which distribution channels the valuable customers buy their scratch cards with a view to focusing sales through this channel.

Another frequently used technique in connection with the optimization of customer lifetime value in the early phase of the relationship is based on new sales via named campaigns. Perhaps we have sent out 1,000 letters or made 1,000 phone calls to get new customers. When the campaign is completed, we can make a profile of the prospects who said yes to our offer. Based on this profile, future activities are focused, on, say, midsized financial institutions, as they seem to be most susceptible to our message. In subsequent campaigns, the message is adapted further to the segment, and advertising resources focus on this target group, as this is where the greatest return is to be obtained. This is an analytical method that is essential to sales departments with limited budgets.

If we want to increase our customers' spending, we do so through added sales activities aimed at optimizing customer lifetime value by increasing their average consumption. There are a number of analytical methodologies that support added sales activities. We will look at all of them in the next chapter, which takes its point of departure in analytical methods. A popular method is cross‐sales techniques, which look for multiple purchasing patterns. A classic example from the United Kingdom describes that men often buy canned beer, frozen pizza, and baked beans together. A clever businessman will therefore position these three products next to each other to remind the segment of this culinary combination. If he, at the same time, chose a slightly up‐market version of one or more of these products, he would secure a bit of extra earnings that way.

Up‐sell sales activities are about knowing our customers' consumption development. From banks we know the financial services that follow a customer's life cycle: children's savings account, youth account, family account, pension schemes, and savings plans. Up‐sell models are about finding out what to offer the customer next and when, based on his or her last purchase. In addition, these analyses can answer the question of who will typically upgrade to new software versions, or which model of car the customer should be offered next.

Optimization of wallet share is about trying to get the customer to make all his or her purchases in one place (i.e., with us). For example, telecom companies know their corporate customers' consumption. They can compare it with an estimate of what the customer ought to be consuming based on, for example, Dun & Bradstreet information. Then the number of employees in each of the customer companies is identified, timed with the average consumption per subscriber in the given segment. If we then combine the actual consumption with the estimate of what the customer can be expected to consume, we can identify which customers are likely to be buying from somewhere else, too, and we can then focus on becoming the sole supplier—before the competing telecom company does this first.

Based on details their customers have given about themselves when receiving their loyalty cards or when logging in on an app, Tesco, a U.K. supermarket chain, has computed the individual customer's family's “stomach share.” This calculation estimates how many calories the customer buys for his or her family in its store. If the number of calories is insufficient to nourish the family, then the store concludes that the customer must be shopping somewhere else, too. Tesco then tries to target more campaigns toward this customer.

When we talk about keeping customers, BA is able, via data mining models, to deliver information about which customers will discontinue their shopping and when. Based on this information, the organization can then come up with some retention products meeting the needs of the individual segments, and thus contact these bargain‐hunting customers. BA solutions can also systematically monitor the different ways in which customers are lost: Some customers are happy enough when they leave, but they just had a better offer, while others really are dissatisfied. Sometimes companies themselves reject bad customers. More information about this is available at BA‐support.com and in The Loyalty Effect by Frederic Reichheld (Harvard Business School Press, 1996). Customer relationship management activities are usually built on value‐based segmentation. This makes sense when we consider the 80/20 rule, which says that a business makes 80 percent of its profit from 20 percent of its customers. A company will therefore do a lot to retain this 20 percent, and will run retention strategies for this group of customers. For example, consider the activities of a large telecom company that let their less‐valuable customers wait in phone queues, while the best customers were put straight through. The company also made different retention offers to customers based on their value segment. Some customers were given a free phone along with cinema tickets, while other had to make do with 100 minutes of free phone time.

In the middle there is a group of customers that the company will typically try to keep, while at the same time increase their value. Toward this group, added sales techniques are used. Finally, we've got the group of least value. This least valuable 20 percent usually delivers 1 percent of the sales. If we then add the fixed costs associated with having these customers, we may well be losing money doing business with them. A business should simply opt out of these customers, or at least minimize all costs when dealing with them.

Campaign Management

This type of analytical process is closely related to CRM activities, but we have moved it a bit toward the operational excellence perspective, because it contains large elements of process optimization in relation to customer dialogue. In more practical terms, this means that CRM has to do with making the right campaigns whereas campaign management has an element of optimizing the existing campaigns in terms of automating and developing analytical tools that monitor the cost and performance of the individual campaigns.

We could call these reactive and proactive CRM activities, where the reactive to a great extent is about minimizing problems, such as the retention of customers. This may seem like spending good money on bad customers. It can be necessary, but should rather be looked at as a troubleshooting exercise and a phase the company just has to get through. Proactive CRM activities are more concerned with rewarding and educating good customers and thus investing in loyal customers.

Often loyalty programs will be created that build on an automated dialogue with the customer at the time and with the content that is most valuable to the customer. A customer who is moving can be registered via a change of address in the data warehouse. In such a case, the company could assist with information about the nearest store in the customer's new local community. An alternative, which is used by one of the largest telecom companies, is a continuous update to the customer about which subscription currently delivers the biggest advantages to the customer. In short, these dialogue programs have the aim of making the customer approach our brand and find his or her shoes, rather than customers approaching the shoes and finding our brand.

For this approach to succeed, we need a high degree of automation of our dialogue with customers and analytical competencies to define which customers must have which messages at which times.

In addition, we need to follow up on existing elements in the dialogue program. What gives a positive, negative, or no response? How can we improve our dialogue?

Product Development

This discipline has its obvious place in the product innovation corner of Exhibit 3.13. Innovative processes can be driven by many forces. An example would be strong creative forces, as we know from people in the arts. At other times, processes are initiated by registered customer needs, which are matched with what we expect to be financially optimal. These processes are, as is known from strategy development processes, a combination of art and science which must form a synthesis, because each of these on its own has a tendency to create only mediocre results.

In this context, BA represents the fact‐based element, which sets the scene for the creative processes and the element, which subsequently validates the quality of the creative results via business cases and simulations. We wish to mention, too, data mining in continuation of customer cancellations. For when we have defined which customers typically cancel and described them in a number of dimensions, we are presented with the creative questions: How can we keep customers with a similar profile? And in this way, how can companies continually, and maybe somewhat reactively, adapt their range of services to the needs in the market? The conjoint analysis is another analytical technique, which by means of relatively few customer interviews can deliver information about how many segments we should be developing new products for, and which characteristics each of these products should have, including the optimum price.

By deploying this somewhat more proactive route, we can deliver input to the creative process about which attributes the new product must contain, as well as subsequently assess the question about the right price in terms of the costs that are associated with the given proposals.

A last powerful source of lead information for product innovation is customer input. This input can be gathered from customer surveys, but complaint logs from call centers or social media blogs can also provide insights into where to make product improvements. During the digitalization waves in the mid‐teens of this first millennium, we are also experiencing an unprecedented rise of methodologies based on Lean and Agile, which are increasingly fueled by customer interviews and surveys.

Web Log Analyses

Web log analyses, and thereby the way in which users click their way around a company's Internet pages, resemble CRM in many ways, as the purpose of the analyses is to understand where our users come from and how to increase their value. We also want to continue to make them use the portal and thereby keep them as customers. The Web log that constitutes the basis for this type of analysis is a file that has one row per click, created on the Web site. This row has information about when the click happened and where the user came from. Often it is possible via cookies or other forms of user ID to recognize and follow the same user over time. Despite its many resemblances with CRM, we have chosen to place this analytical method between customer relations and operational excellence, because Internet portals and apps may have many others purposes, too. These include customer self‐service, storage of public information, or as an intranet for staff. If a company has a commercial Internet page where it sells its products, it must—just as any other store—make the market aware of its existence and its offers. Therefore, we need to perform marketing activities. In continuation of these marketing activities, we would like to understand where our customers came from, and what we did right or wrong.

If this is done on the Internet, there are a number of ways of attracting customer attention. We might buy banner ads, which are ads that pop up in other places on the Internet, where the Internet user who clicks on the advertisement is sent to the advertiser's page. Search engines are another way of attracting attention. The company can search on some key words that the company wants to be associated with and, based on the Internet pages that then appear, decide what to imitate. And we can imitate existing Internet pages to gain a high placing on the results pages of the search engine. A company can make collaborative agreements with other Internet pages that offer complementary products, and mutually refer to each other. Chat rooms and social media that are used by our customers are other places to attract customer attention and, naturally, we can also choose to run our campaigns via the “old” media.

This also means that in order to know what works and what doesn't, businesses selling their products via Internet stores will be interested in measuring how customers/users are referred to their Web site. Web logs can deliver this information if users reach the portal via clicking on links, since this will simply be registered in the Web log. From this we can see which marketing activities via Internet referrals deliver good response.

When we are talking about getting customers to increase their usage, it can be helpful to divide them into groups as shown in Exhibit 3.15.

Via the Web log we can see which users are likely to be visiting for the first time and, as we have not registered their identity before, this could be the result of marketing activities. We can see whether the same surfers visit our page several times and, from this background, ensure that they are shown a page that reflects this. They are obviously interested; they have come back, so now we want to try to develop them into what we may call users, which could be achieved by getting them to create a user profile. On the basis of this user profile, we now know much more about the user, and he or she may even subscribe to our newsletter. We will be able to see whether our emails are attention‐getting enough for our user to forward them to others in his or her network who might be interested. Finally, we can develop the user into a customer, which happens when he or she buys from us. Now we know even more about our customer, because we know his or her physical address, which products he or she likes, and we know his or her behavior on our Web site.

From now on we interact with a given customer by running traditional CRM activities to keep and grow him or her. In addition, we have information about which marketing activities result in an actual sale. We can now follow the entire development, from banner advertisement to purchase, and learn from this. Are there any geographical areas where we should advertise more? Are there places on the Internet where we should advertise more, and the like? How did the customer develop over time, and which pages did he or she look at prior to purchasing? How many times do people visit our Web site before they make their first purchase, and how can we improve this process?

When we talk about retaining customers, it's largely the same analytical methods that are used as in conventional CRM. However, the key issue here may be when we lose a given customer. Is it when he or she canceled his or her newsletter, or when there has been no purchase in a year, or when the customer moves down into a lower value segment, or perhaps something else?

At the time of writing, we are amazed at how few resources are actually used to analyze Web log information. It's true that the dot‐com time is dot‐gone, as the saying goes, but the vision behind the dot‐com trend is as real, as it is reborn in the world of apps. The Internet and world of apps are still a parallel universe to physical distribution, with enormous potential for scalability, both independently and in combination with physical distribution. The large shift did happen; consumers have gotten used to using the Internet and apps. So the big revolution represented by the dot‐com trend has been more of a trickling process, with businesses waiting for people to start shopping on the Internet. Now the consumers are ready and, ironically, we now seem to be waiting for stores to get ready to provide unique shopping experiences beyond drop lists, baskets, and PayPal.

Social media data is becoming increasingly interesting, though traditional commercials placed on leading platforms notoriously underperform. However, the data can be used in other ways; for example, we see that measuring the social media response to campaigns at a very early point, will give a strong indicator of the overall success the campaign will achieve. This approach offers an alternative to traditional campaign tracking based on costly customer interviews.

The trend is also moving in the direction of storing social media data together with Web log information per customer in CRM systems. Based on the language that customers use on their social media platforms, language‐based algorithms similarly analyze and segment customers based on whether they can be characterized as analytical, concrete, passive, and the like. Banks, for example, scan social media for significant life events, such as looking for new houses, comments on pregnancy, or job changes, so that the next campaigns can address these new life events.

Pricing

We have chosen to place pricing in the product innovation corner, because we feel that price is an essential element of a product. Versace and Rolex would, for instance, not be the same if they could be purchased for $20 in the nearest supermarket.

Pricing is a subject in its own right in the field of marketing. Price is hugely connected with the position the company's brand has or wishes to have in the market. We can work with this analytically, too, but analytical optimization is first and foremost about identifying which price a given product with its given characteristics, for a given segment, should have. The product may be new, so we have little or no historical information to work with, and we therefore want to use a conjoint analysis as mentioned in the previous section on product development. Alternatively, the product (or a similar product) may have been on the market a number of years, and we therefore have a lot of historical information to work with.

The techniques used in this context are called forecasting techniques. These techniques consist of two phases. The first is about understanding the correlation and, based on this, developing a model. The next phase is about using the model to find the optimum mix of price, amount, and marketing activities. Questions to be answered in the first phase could be:

• How big an impact do my competitors' prices have on my revenues?
• If I carry out a campaign, how big an effect will it have on sales?
• When will I see the effect of the campaign, and when will it stop?
• How do I adjust my activities in connection with seasonal, weekly, or daily variations?
• What are the synergies between advertisements in different media?
• What will my sales be in the coming quarters?

When we get to phase two, we'll have knowledge about the market mechanisms under which we operate, which means that we are in a position to adjust our business behavior. Specifically, this can be done via simulations such as these: What happens if I employ another salesperson, if I increase or reduce my price, or if I carry out more and/or combined campaigns? We will, in other words, be in possession of analytical input to support the optimization of the company's marketing mix. It all sounds rather complex, but the market offers software that makes it surprisingly easy and cheap to deliver these correlations.

Self‐optimizing dynamic pricing is also a new trend which is on the rise in processes that sell tickets to travels or concerts, or access to hotel rooms or container slots on ships. All these business processes have in common that they must sell a service, and if they fail to, the opportunity is lost; for example, if we do not sell a room to a customer for a certain night, that earning possibility will never come back. Dynamic pricing builds on the assumption that at, for example, 30 days before a departure, 50 percent of the tickets to holiday should be sold; if they are not, the prices will be dropped—and vice versa. The general idea is that all tickets, rooms, or slots on the ship must be sold at the highest possible price.

Human Resource Development

In the middle of Exhibit 3.13, we have placed human resource development (HRD), which consists of processes, excluding the purely administrative, that spring from the human resources (HR) function. The reason is that the three competitive disciplines described by the model are never going to be any better than the people who perform them. Qualified and motivated employees are often the scarcest resource in modern organizations. In this context, it's a wonder that more major organizations, including wholly or partly public institutions, don't employ analytical HRD. We are talking about very large volumes of data that describe employees based on dimensions such as illness, education, target achievement, gender, age, manager, department, and career route, so the data material is already there.

The biggest hurdle here may be the culture that exists in HR departments, with its strong focus on creativity and soft values. Thus, it can seem provocative to present analytical facts, however qualified. We therefore have to approach this professional area via strong sponsors high up in the organization, since a noncommitted culture, as we know, can eat any strategy for breakfast.

If we look at analytical CRM, the approach from this field can be used in connection with analytical HRD, because we need to be able to attract the right employees, optimize their performance, and retain the best—that is, get, increase, keep. The objective is therefore comparable with the one we know from CRM. The means are, however, completely different, because in HR we can develop employees via new hiring and talks with existing staff, with focus on satisfaction surveys, professional or personal development, and/or reward systems.

In terms of attracting the right employees, we can take future desired states into consideration by analyzing which groups of employees perform above average.

When we are talking about optimizing performance, this could be achieved by associating the motivation of individual employees with their absence due to illness. Employees in the municipalities of Copenhagen and Aarhus in Denmark have, for instance, accrued more than 20 days of absence due to illness per year, whereas employees in bordering local municipalities have accrued less than half that. The profiling of which groups of employees have a lot of absences due to illness can therefore provide a good basis for targeted initiatives adapted to the individual employee's need for motivation. To quote from another context: “If our employees are the organization's most important resource, then it is management's most important mission to ensure they turn up for work again tomorrow.”

Employee retention is also a discipline that can render huge value creation. Losing and hiring a typical skilled professional leads to costs of around $50,000. So what must the company offer its employees to ensure that they are still there same time, next year? Is it possible to offer employees a month or two of working only 80 percent of a 40‐hour week, with a corresponding salary reduction, while their house is being renovated? Is it possible, with no change in salary, to offer employees a company car, share options, or professional or personal development?

Questions such as these can be answered via analyses of questionnaire data and interviews with employees, who are leaving the company. The company can, with good reason, view its employees as customers who pay with their work. And as a thank‐you for their work, employees should receive individualized counter services: a mix of salary, leisure time, personal and professional development, and the like. For more about this subject in connection with the presentation of the SIPOC model, see Chapter 8.

In regard to hiring procedures, companies are often using intelligence and person profiling tools. With the rise of multiple publicly available data sources, ranging from social media and public data to endless sources of written documents, in the future we can expect personal profiling based upon this Big Data. Already the first tools that make a personality description of people based on, for example, Twitter data are available. The next step will simply be to join all data sources and present them in a format that is optimized from hiring procedures.

Behind this trend, we also can start to understand how data in itself is increasingly becoming a sellable commodity, including the need for discussion about the individual's right to own and impact all existing data about him‐ or herself. For example, consider this: Would you employ a person who, based on genetic analysis, has a 5 percent likelihood of mental illness within the next 3 years, even though her person has no previous illness history? If you say yes to this, what about 50 percent? What if you had an equally good candidate with a likelihood of 0.5 percent? What if you were the owner of a little three‐person company and you know that this would set you back?

Another area to be aware of when relying only on analytics when creating campaigns and for customer retention is that these algorithms are designed to react to the market. So if there is a lot of churn in a certain market, this is what the churn prediction model will direct its attention toward. Similarly to sales, the algorithms aim to sell what is sold the most. These are of course very valuable capabilities to have in an organization; however, we must also be aware that these mechanisms will direct our go‐to‐market model toward a market‐follower approach. Everyone responds to the market. However, if we are strategically oriented toward being a market leader, we must also consider other approaches. They could simply be the ability to see whether campaigns and other customer events have an impact within hours or minutes via dashboards on the wall, and this way, the marketing go‐to‐market approach will become less reactive and more based on hyper‐innovation, meaning that from hour to hour we can test new messages, product bundles, prices, channels, and the like.

Corporate Performance Management

We will only mention this subject briefly here, as it is covered in the section on optimizing existing business processes earlier in this chapter. CPM, or corporate performance management, is about measuring processes (performance management) in order to understand the correlation between process‐improving activities and their effect, with a view to further improving the processes. We have chosen to place CPM in the middle of the triangle to indicate that the desired learning may relate to innovation, customer relationships, and operational excellence.

If we focus solely on the Six Sigma and the Lean approaches, focus will move toward the operational excellence corner, depending on the extent to which we want to include the customer's needs when establishing new processes. Through so‐called control charts, we also get some useful alternative tools to monitor our processes, and they will tell us whether we have managed to influence them positively via our process‐improving initiatives.

Finance

Activity‐based costing (ABC) is about being able to allocate the company's costs to the processes that are generating them. The purpose of this is to enable us to subsequently assess which products or customer groups are profitable. Generally speaking, it gives the company a clearer idea of where its significant costs are. If some processes represent great costs, and these processes are not essential to the company's competitive situation, outsourcing of these should be considered. If, for instance, a company produces designer goods and has a significant profit margin on this, but at the same time is running a number of shops with a small profit margin, this could obviously prompt considerations as to whether it might be an idea to sell the shops and focus resources on the most successful area. Similar considerations may have resulted in Shell Denmark selling all its shops to 7‐Eleven in 2007. Shell is good at creating high‐profit margins in the energy industry, but less so in retail—and vice versa with 7‐Eleven.

If we combine ABC, which is about knowing our cost structure, with conjoint analyses or other pricing methods, which are about sales potential, we get some strong tools that enable the company to optimize based on profit margin.

Lean is another approach to minimizing costs. Here we constantly work toward what has been defined as an optimum process. Whatever is between the actual process and the optimum one is described as a waste of resources, and the cost should be cut. Critics of this method maintain that this is about process optimization, but not necessarily well‐being, which is why Lean measurements benefit from being supplemented by employee satisfaction surveys (analytical HRD) to give the full picture.

Other kinds of financial analysis could be about documenting the actual value of customers (known as value‐based segmentation) to let the company know which customers they do the most to retain. Similarly, in industries where contracts are negotiated individually, a yearly review of the actual earnings per customer should be considered a standard input to the yearly contract negotiations. This analysis will also expose which key client managers should be rewarded for being the best negotiators.

Inventory Management

We are now far down in the operational excellence corner of Exhibit 3.13, where our optimizations can be difficult to link to direct customer relations and the future product development. This applies to inventory management, too, as this field is concerned with ensuring that the people who draw on the inventory must always be able to get what they want. If they can't, production will grind to a halt, whether it's the spare part or raw material inventory that has gone into back order. If the product inventory is empty, customers will be waiting. Being overstocked, on the other hand, represents an unfortunate situation of tied‐up capital, and the risk of stocked items becoming outdated and losing value.

An analytical approach delivers decision support to the people responsible for inventory management in terms of identifying the optimum number of items in stock. Of course, there's always the risk of the occasional empty shelf, but that is a calculated risk. Any loss that is incurred because of empty shelves is compensated for by less tied‐up capital by means of overall smaller stock. As described in connection with CPM, analytical tools deliver the possibility of continuous monitoring of which stock items are unavailable too often, enabling the company to continuously adapt its inventory.

Given the Internet of Things (IoT), these processes are increasingly being fully digitalized, robot‐operated, and in many production industries increasingly integrated with supply chain management.

Supply Chain Management

Supply chain management (SCM) is about managing the company's relations to its suppliers. These relations vary from company to company and from supplier to supplier. At one end of the spectrum, we've got strategic collaborations where the companies' processes become as one, and we're working with joint development projects of products and logistics. At the other end, we find relationships characterized by sporadic transactions, where alternative suppliers can always be found. In terms of strategic collaborations, the analytical methods will reflect the same methods that are used for internal optimizations. These could be ABC, Lean, or CPM, as shown earlier.

With regard to the more casual supplier relationships, in which price is negotiated from deal to deal, the required information will be more about giving the buying organization complete details about its suppliers. This can be seen as a countermove to the CRM information held by the selling organization that is used to optimize its sales processes.

It may sound trivial, but this ultimately has to do with the fact that large companies such as A.P. Moeller (APM), a conglomerate within the energy and shipping industry, have thousands of suppliers worldwide, and these suppliers are specialists in their fields. APM is therefore in a weaker negotiating position than the seller. The countermove here is to ensure that the buying organization has information about how to get the same service from somewhere else. It would work to strengthen APM's bargaining position, too, to know whether the company is a major buyer of services from the individual supplier, making the company a key customer that the supplier wouldn't lose for anything. For the relatively large company, analytical SCM is about obtaining all the details about the many suppliers, so that the company has in‐depth knowledge about the pricing in the supplier market, with the result that it could exploit a strategic customer position. As large organizations also have the greatest saving potential, we must expect that these organizations soon will start to invest, targeted into information systems that, based on all the available public data, use this information to their benefit. Also we must expect that the suppliers, to be visible in these Big Data searches, will provide data that fit into these searches. Ultimately this can be expected to give the buying organization more price visibility, as in the business‐to‐consumer (B2C) market where consumers can use tools like PriceRunner, Amazon, or eBay to compare prices.

Lean

It could be argued that Lean should be placed in the middle of Exhibit 3.13 since product development processes can also be Lean, and since when we buy a service we are essentially a process based on our needs. However, we placed it in the right side of Exhibit 3.13 because Lean essentially is the discipline of balancing internal resource utilization with what the customers want.

There are many ways to optimize processes essentially; however, they all have the same purpose of optimizing the balance between how we as a company deploy our costs and what makes the customers of the process satisfied. There is no universal balance that in itself is the optimal. It depends on our organizational strategy and customer expectations. For example, it's possible to buy a meal at both the Waldorf‐Astoria and Burger King; however, these are very different experiences. This also means that when we design the way customers must line up in a Burger King restaurant, they are quite unlikely to get any free champagne, as opposed to what they might receive at the Waldorf‐Astoria.

There are several terms used for this way of improving our business model, such as business process reengineering, Lean, Six Sigma, operational excellence, process excellence, and the like. The essence of this approach is that the changes that we set out to do, via the different tool boxes, are specifically approached from a process perspective as opposed to other disciplines which are more departmentally focused (e.g., procurement, inventory, finance, marketing, etc.).

In brief, analytics primarily support this way of making organizational improvements in three ways. First, learn what the customers want before starting to change the process; this can be done through surveys that ask customers about the basics that the Waldorf‐Astoria has to get right when customers are waiting, like indoor seating and a good social atmosphere. Find out what the customers are willing to pay. Finally, find out what would delight the customers and could be a business differentiator, which could be the champagne mentioned before. As a starting point, however, analytics could also text mine customer comments or social media in order to get an idea about what should be included in the questionnaire in the first place. As a general rule, text mining in the form of complaints will give an indication of what basics are occasionally being handled poorly, whereas the farther we move toward the delight factors such as champagne in the waiting line, the more we have to rely on creative and non‐data‐driven sessions.

To learn whether the customers actually like the newly envisioned process, questionnaire information can help process owners to see how it affects customer satisfaction. The critical question is whether the company has achieved the right balance between how they spend their resources and what is rewarded by an increased willingness to pay a higher price by the customers. Alternatively, some organizations seek the right balance between how they spend their resources and the change in customer loyalty, under the assumption that a mathematical relationship between the customer loyalty and what the customers spend has been established. The critical questions could be: Have we made the process so cheaply designed that it does not live up to our customers' expectations set? Are we spending money on something that the customers don't really value anyway?

One analytical tool that can help balance customer satisfaction with the cost of implementation and of running the new process is called a customer value calculator. It can be as simple as a spreadsheet in which we enter the expected increase in customer satisfaction that we have learned from interviews or questionnaires. The output of the spreadsheet is then the financial results that the company can expect to see over the next years. These financial results could be based on whether customers will give a higher wallet share, stay longer or be willing to pay more for our services. How exactly we estimate the increased returns is up to our kind of business, what data we have and which statistical correlations we can find in our data warehouse. We have learned when making a value calculator is that input from yearly customer satisfaction surveys can be extremely valuable data, as it will identify, for example, whether satisfaction has gone up compared to last year, what the impact is on how much people buy (wallet share), and whether customers with high customer satisfaction have longer customer lifetimes measured with survival curves, or are customers with higher satisfaction scores willing to grant better contracts (e.g., average price per container shipped).

To learn about the processes on a continuous basis, analytics also monitor existing processes for changes via control charts that indicate whether the process performance has changed significantly from a statistical perspective. This can answer questions like: Has the process improved? Has it become more stable and predictable? Companies with a great focus on Lean will often have a process owner who in turn has a control chart of the particular process which is revisited every day. The purpose of a control chart is to see whether there are any changes in the performance of the processes, or, alternatively, how the company can benefit from or minimize the damage of the change. Text mining through a customer satisfaction survey could also be relevant for process owners since a control chart only sends a signal telling that something is wrong. However, written input from customers might reveal information about what causes problems from a customer perspective.

As Lean has to do with designing and optimizing the way we work in terms of processes and the current overarching trend in the business world today has to do with digitalization, analytics has also become a key success factor. The link is that the people who used to take the decision within a process now also have to be replaced with algorithms. In fact, one of the next business functions that we expect that will be increasingly digitalized is customer service, where we soon will simply speak or write to a customer service robot or other kind of front end, which, based on our input, will translate our words to text and, based on natural language processing, will seek to understand our mood and needs, and respond accordingly. The rise of apps on mobile devices for self service is another area where we already have seen digitalization of processes that have helped organizations Lean the way they work and where we, to an increasing extent, must expect to see analytics that, behind the scenes, continuously will try to make the self‐service process easier for the user and more efficient for the company.