This section of Smart (Enough) Systems will discuss why decision logic is difficult and why some past approaches to technology and systems have failed. You'll also learn about the data tipping point, how serviced-oriented architecture (SOA) and related technology is helping companies create smart systems and why a smart system must automate operational decisions. You'll also be introduced to SmartEnough Logistics, a company that ships packages around the world for its business clients and applies smart enough systems to maximize returns and minimize costs.
Table of contents:
- The importance and benefits of operational decision making
- How to make operational decisions and data corporate assets
- Benefits of operational, real-time capabilities in smart systems
- How to create automated operational decisions
Current Approaches Fail
"We have established that you cannot code your way into the future."
Different approaches to technology and systems development have failed to deliver smart enough systems so far, but why?
"The problem is that computers and the software that runs on them . . . are notoriously difficult to change with any speed or accuracy."
First, decision logic (policy rules, formulas, thresholds, regulatory mandates, and other elements used to make decisions) traditionally has been hard-coded into operational systems. As a result, development is time-consuming and costly. Developers have to translate business requirements ("If this condition is encountered, respond in this manner") into abstract representations in programming. This is a laborious process full of possibilities for error through misinterpretation. Developers have to try to anticipate all possible requirements and conditions because any changes after deployment could affect other parts of the program and require unraveling a good part of their work. Businesspeople requesting a change usually have to wait weeks or even months for the change to be coded and deployed, and because the hard-coded decision logic is buried in a system, it must be written (and rewritten) for each new platform or channel.
In addition, decision logic is difficult to understand. Because it's lodged in application programming code, business managers often have difficulty saying exactly how decisions are made. Different programmers might have coded layer after layer of policies and other types of rules in various ways. Some companies have tens of thousands of rules coded into their systems, including many that are irrelevant because they're based on market conditions and business requirements that no longer exist. Also, as organizations have moved from proprietary programs and applications to packaged applications from independent software vendors, the range of available decision rules and criteria has shrunk to those that could be "configured" with software system tools and workbenches.
Second, good decision making requires insight, especially into the probability of specific outcomes. Retail banking and other credit-extending companies have used this type of analysis extensively in automated decision systems. These "predictive analytics" are equally valuable—and still largely unused—for decision making in other industries. Business managers who want to bring predictive models into their decision processes might be daunted by the complexity of the data and analysis, however. Additionally, there's the impact of analytics deployments on IT resources. Predictive analytics, like decision logic, must be programmed into application code.
Third, although many companies can capture data from front-line systems and have invested heavily in data warehouses to store it, too much time might go by before they draw insights from the data. Most companies, in fact, often operate on stale data, partly because of what must be done to turn the data into a form useful for gaining insight.
Massive investment in business intelligence (BI) and data warehouse technology has undoubtedly helped management understand the impact of their decisions and detect trends in their business. What this technology hasn't done is improve the way employees and information systems that interact with customers make operational, front-line decisions. The purpose of using BI is to put it in the hands of people who can use analytic and business operations skills to understand what it's telling them. No matter how much visualization or smarts are embedded in these tools, they remain focused on knowledge workers who aren't the people making most of the decisions involved in day-to-day operations. These decisions are made by customer service representatives, counter staff, drivers, Web sites, or telephone support staff.
The Data Tipping Point
More and more data is now available to organizations about their operations, their customer interactions, and their Web sites. With the arrival of radio frequency identification (RFID) chips on every palette, case or box of products, and eventually on every individual product, organizations will have more data about their supply chain than ever before. With mobile devices that are always on and fitted with global positioning system (GPS) chips, every vehicle and employee will be a source of a continuous stream of data. Customers, too, as their mobile phones interact with organizations' systems, will deliver constantly updated information about their whereabouts and activities. Growing sensor networks and the integration of massive external consumer databases with enterprise and government databases will only add to this increase in information. We will, if we haven't already, reach the tipping point where the volume of data overwhelms current data reporting and analysis systems.
Two other factors complicate an organization's ability to take advantage of this embarrassment of riches. First, most tools, techniques, and methods for managing data are largely for transactional, relational data. Much of the new information that's available isn't. It might be unstructured text, as in e-mail or blogs, or structured but not semantically understandable. Social network software accounts for some of the most popular Web sites and can be a gold mine of information, if it can be extracted and understood. Second, all these types of data require technology such as voice recognition, image recognition, and text analysis to turn previously unusable data into information. Bigger volumes of unfiltered data, however, won't be valuable to organizations unless the data can be turned into useful insight.
Modern IT Architecture Is Helping
The recent growth of standards-based service-oriented architecture and related approaches and technology, such as business process management and event-driven architecture, is moderating the negative impact of many of these trends. The approach in this book is complementary to these changes in how IT tackles problems. How this approach interacts with and complements a modern IT architecture is explained in Chapter 10, "EDM and the IT Department."
Fourth, much money has been spent on customer relationship management (CRM) and other enterprise application technology. Too few CRM implementations have successfully created a unified view of customers, identified their preferences, rewarded them for providing information, and then marketed to them and interacted with them the way they want. Many companies fail to respond to customers in a consistent, focused, targeted way, despite massive investments in CRM. Too many call centers and other groups of front-line workers have been neglected. These agents don't have what they need to help customers solve their problems, too much cross-selling and up-selling are done without sensitivity to customers, and the feedback loop to improve interactions is broken. The move to outsource call centers has only exacerbated this problem.
Current practices in coding, data analysis, data capture, and data management and the priorities represented by enterprise applications have resulted in systems that just aren't smart enough anymore. Chapter 3, "Why Aren't My Systems Smart Enough Already?" gives more details on this problem. But if the current approaches to information systems don't work, what can provide smart enough systems?
Decision Management Is Required
Fundamentally, a smart enough system must automate the operational decisions that drive your business. If you identify and automate operational decisions, you can separate them from the rest of your applications so that they can be managed and reused.
Managing decisions isolates the logic behind the decisions, separating it from business processes and the mechanical operations of your applications. Treating decision logic as a manageable enterprise resource means you can reuse it across applications in different operational environments and treat your decisions as a corporate asset.
Managing decisions means applying analytics to make decisions more precise. Using analytics in this way makes it possible to ensure that your decisions are informed by the data you're capturing. Indeed, with experience, you can apply more advanced analytics, take market and economic uncertainties into account, and arrive at optimal decisions.
Managing decisions makes it easier for you to improve decisions over time. You can focus efforts on improving decisions and be certain that improvements will be spread throughout your organization. This focus means your return on investment (ROI) is higher, because any improvements in decisions improve results in all applications that use them, essentially multiplying the value of your investment by the number of applications used.
Managing decisions has a cultural component. By recognizing and separating out operational decisions, you can focus your business thinking and investment on these decisions more easily. You can apply your strategic vision and management approaches to achieve optimal decisions.
So what might an organization that used smart enough systems to run its business look like? How would an organization that managed its decisions act?
Introducing SmartEnough Logistics
SmartEnough Logistics is a glimpse into the immediate future. SmartEnough is a company that ships packages around the world for its business clients and applies smart enough systems to maximize returns and minimize costs.
Customers interact with drivers collecting packages, who can price, up-sell, and cross-sell effectively, thanks to their handheld devices. These devices can record customer preferences and needs, predict whether the service being purchased is more or less than required, match this prediction against the customer's contract and established preferences, recommend cross-sell and up-sell opportunities, and get the price right, given the relevant contracts.
Packages are tracked using RFID and trucks with GPS, so the tracking system knows where each package is (in which truck and at what location). The dispatch system also uses this information to make adjustments. For example, it predicts that a driver's truck won't have the capacity for the second-to-last scheduled pickup because of the volume of orders at that location, so it changes the truck's route dynamically and transmits this information to the anticipated driver, who's used to this kind of just-in-time changes to routes. The historical data from the GPS allows the dispatch system to predict which trucks can make the pickup (by predicting the additional time needed) and select a different truck by balancing this data with each truck's open capacity.
As another example of making just-in-time changes, while a truck is in transit, a customer realizes that one of her packages was shipped with too low a priority. She logs on to the company's Web site to change the shipping priority. The site uses the same decision engine as the driver's handheld device, so it gives her exactly the same information about likely delivery dates, times, and pricing as the driver did, reassuring her that everyone involved knows what they're doing. The customer requests expedited delivery online, and the system responds immediately with available upgrades based on the rules and analytics built into the scheduling system. New pricing is displayed, and the customer accepts; because her relationship with the shipper is established, she doesn't need a credit check or additional credit card.
When the truck arrives at the cross dock, the unloading crew takes each package off the truck. The RFID tags trigger an automated sorting belt that routes them to the correct loading areas for different delivery schedules and destinations. This system routes the changed package differently at this point, given its new information. The manager of the loading area notices this change because he received an automated alert from his activity-monitoring system when a label configuration unusual for the load being assembled was scanned. When he checks, the system can tell him exactly why the routing has been changed.
While the plane carrying the package is in flight, the customer calls the call center from her cell phone on the drive home, worried about her package. The customer service representative (CSR) sees the shipping information, the rerouting and the reason for it, and he has access to the same system for predicting delivery time. This delivery time, of course, now reflects the impact of the rerouting. Despite reassurances from the CSR, the customer is still concerned the package won't arrive on time. The CSR asks the system for other options on the package, and it shows that no additional rush options are available (given the package's current routing). The CSR also sees that an extra notification offer is available for free; this offer is based on the package and its delivery location, and the pricing is based on the customer's status as "concerned" (entered by the CSR) and "long established" (from the system). The customer accepts the offer and goes home to bed.
This package must also go through customs at its delivery location. SmartEnough has another system that applies current rules for the destination (a combination of rules established by the locality and rules from the federal government about exporting the items in the package) and generates the correct customs paperwork. This system ensures that the package won't be held up in customs. Some rules were added just today when new export rules were announced unexpectedly. Fortunately, SmartEnough's system could be updated directly by the legal department as soon as they understood the new rules' implications.
During the flight, bad weather closes an airport on the route, forcing a diversion. This information enters SmartEnough's system directly from the airport's system. Automated routines run in response to see what rerouting options exist for packages on the flight (package by package) and determine that some won't make their scheduled delivery, no matter what. The system immediately notifies these customers of the delay and gives new delivery times. It also makes retention-oriented "we're sorry" offers based on calculated retention risks for those customers and the kind of service they ordered. Packages that can be rerouted are.
Two options are available for the concerned customer's package. The first means it might arrive on time but has the risk of a lengthy delay. The second means it won't be on time but guarantees that it will be only a little late. Given the kind of service ordered, customer preferences, and the package's transit history, the system chooses the first option and informs the customer. At this point, there's time for the customer to change the option if she notices the notification in time (it's now night), but the system has made the best decision it can for now.
When a package is delivered in a foreign country, a third-party delivery company is used. Because SmartEnough makes all its systems available to its extended enterprise, the delivery company's driver has access to the same information and same decisions as the driver of a SmartEnough truck. When the package is delivered—on time, as it happens—the third-party delivery staff are notified that acknowledgment of receipt is important for this package, so they double-check with the hotel staff at the delivery location. The system prompts for a name and phone number from the person signing for the package and transmits this information to the customer.
Smart (Enough) Systems
by James Taylor and Neil Raden ISBN 0132347962
First printing June 2007
Prentice Hall Professional
More on accessing data: