5

COMPETING ON ANALYTICS
WITH EXTERNAL PROCESSES

Customer and Supplier Applications

ANALYTICS TOOK A GREAT LEAP forward when companies began using them to improve their external processes—those related to managing and responding to customer demand and supplier relationships. Once kept strictly segregated, the boundaries between customer relationship management (CRM) processes such as sales and marketing and supply chain management (SCM) processes such as procurement and logistics have been broken down by organizations seeking to align supply and demand more accurately. Unlike internal processes that lie completely within the organization’s direct control, externally focused processes require cooperation from outsiders, as well as their resources. For those reasons, managing analytics related to external processes is a greater challenge.

Despite the challenge, many companies in a variety of industries are enhancing their CRM and SCM capabilities with predictive analytics, and they are enjoying market-leading growth and performance as a result.

Many companies generate descriptive statistics about external aspects of their business—average revenue per customer, for example, or average order size. But analytical competitors look beyond basic statistics and do the following:

• They use predictive modeling to identify the most profitable customers—as well as those with the greatest profit potential and the ones most likely to cancel their accounts.
• They integrate data generated in-house with data acquired from outside sources for a comprehensive understanding of their customers.
• They optimize their supply chains and can thus determine the impact of unexpected glitches, simulate alternatives, and route shipments around problems.
• They analyze historical sales and pricing trends to establish prices in real time and get the highest yield possible from each transaction.
• They use sophisticated experiments to measure the overall impact or “lift” of advertising and other marketing strategies and then apply their insights to future analyses.

Strange Bedfellows?

At first glance, supply chain management and customer relationship management would seem to have little in common. Supply chain management, on the one hand, seems like a natural fit for an analytical focus. For years, operations management specialists have created algorithms to help companies keep minimal levels of inventory on hand while preventing stock-outs—among other supply chain challenges. And manufacturing firms have long relied on sophisticated mathematical models to forecast demand, manage inventory, and optimize manufacturing processes. They also pursued quality-focused initiatives such as Six Sigma and kaizen, tools for which data analysis is an integral part of the methodology.

Customer relationship management, however, seems less amenable to analytical intervention—at least, that might be the common perception. The traditional focus in sales has been on the personal skills of salespeople—their ability to form long-term relationships and to put skeptical potential customers at ease. And marketing has long been viewed as a creative function whose challenge has been to understand customer behavior and convert that insight into inducements that will increase sales.

But the roots of analytics in business come from the customer side as much as the supplier side. Twenty years ago, consumer products firms like Procter & Gamble began using analytical software and databases to analyze sales and determine the parameters of product promotions. These companies invented the discipline of marketing-mix analytics to track the impact of individual investments such as trade promotions and coupon offers. They collected and analyzed data from vendors like ACNielsen and Information Resources, Inc. (IRI) to understand how their customers’ (grocers) and consumers’ behavior was influenced by different channels. These early innovators are being joined today by companies in virtually every industry, including retailers such as 7-Eleven Japan, manufacturers like Samsung, phone companies such as Verizon and Bouygues Telecom, and pharmaceutical companies such as AstraZeneca. More recently, marketing organizations have radically increased their analytical orientations with the rise of campaign management software. Quantitatively oriented marketers can now use these tools to experiment with different campaigns for different groups of customers and learn which campaigns work best for which audiences.

Analytical competitors, however, take the use of analytics much further than most companies. In many cases, they are pushing not only data but also the results of analyses to their customers. Our survey data suggests that they are also integrating their systems more thoroughly and sharing data with their suppliers.¹ As companies integrate data on products, customers, and prices, they find new opportunities that arise by aligning and integrating the activities of supply and demand. Instead of conducting post hoc analyses that allow them to correct future actions, they generate and analyze process data in near–real time and adjust their processes dynamically.

At Harrah’s casinos, for example, customers use loyalty cards that capture data on their behavior. The data is used in near–real time by both marketing and operations to optimize yield, set prices for slots and hotel rooms, and design the optimal traffic flow through the casinos. Harrah’s chief information officer, Tim Stanley, describes the change in orientation: “We are in a transition from analytical customer-relationship management, where customer data is analyzed and acted upon at a later time, to real-time customer analytics at the point of sale in the casino, where . . . action is taken on data as it is being collected.”²

How does this work in practice? One example can be seen when a customer loses too much money too fast. Harrah’s systems can identify this problem and almost immediately send a message (electronically or through a service representative) to the customer at a slot machine, such as, “Looks like you’re having a tough day at the slots. It might be a good time to visit the buffet. Here’s a $20 coupon you can use in the next hour.” Harrah’s is also experimenting with real-time marketing interventions over cell phones and PDAs that help customers manage their entire vacation experience in Las Vegas, where the company owns several adjacent properties. “There are two seats left for the Celine Dion concert tonight,” a text message might say, “and we’re making them available at half price because of your loyal play at Harrah’s! Text ‘yes, 2’ if you’d like both tickets.”

In the remainder of this chapter, we’ll explain how other companies are taking advantage of their analytical abilities to optimize their customer and supplier processes.

Customer-Based Processes

Companies today face a critical need for robust customer-based processes. For one thing, acquiring and retaining customers is getting more expensive, especially in service-based industries such as telecommunications and financial services. And for another, consumers are harder to satisfy and more demanding.³ To compete successfully in this environment, analytical competitors are pursuing a range of tactics that enable them to attract and retain customers more effectively, engage in “dynamic pricing,” optimize their brand management, translate customer interactions into sales, manage customer life cycles, and differentiate their products by personalizing them (refer to “Typical Analytical Applications in Marketing” box on the next page).

Attracting and Retaining Customers

There are, of course, a variety of ways to attract and retain customers, and analytics can support most of them. One traditional means of attracting customers has been advertising. This industry has already been, and will continue to be, transformed by analytics. Two factors are most closely associated with the transformation. One is the econometric analysis of time series data to determine whether advertising is statistically associated with increased sales of a product or service. The other is the “addressable” and relatively easily analyzed nature of Web-based advertising, as exemplified by Google. We’ll describe each of these briefly.

Econometric analysis has begun to address the age-old problem with advertising in traditional media, as described by department store pioneer John Wanamaker: “Half the money I spend on advertising is wasted; the trouble is, I don’t know which half.”

Sir Martin Sorrell, CEO of WPP Group, one of the world’s largest advertising agencies, calls econometrics the holy grail of advertising. He noted in an interview, “There is no doubt in my mind that scientific analysis, including econometrics, is one of the most important areas in the marketing-services industry.”⁴

Several advertising agencies have created groups of econometrics experts to do such analyses for clients. One example is DDB Matrix, which was spun out of the agency DDB in 1999. DDB Matrix gathers data for its clients, builds data warehouses, and analyzes the data to find answers to a variety of questions on advertising effectiveness. The questions include such issues as which medium is most effective, whether the additional cost of color is worthwhile in print advertising, and what days of the week are best to run ads. Obviously, a great deal of data must be gathered to rule out alternative explanations of advertising lift. CEO Doug Hughes describes the data for one client: “For example, with Dell we have, over the last seven years, built a warehouse with more than 1.5 million records containing data on all ads, print, radio, cable TV, etc. The application now has thousands of lines of custom code. This database started out quite small. We expect that firms will increasingly view analytics about advertising as a necessary adjunct to embarking upon any campaign. Otherwise, advertising resources can hardly be considered well spent.”⁵

The other dramatic change in advertising is the rise of online, Web-based ads. These are revolutionary, of course, because whether someone clicks on an ad can be tracked. There are a variety of Web-based advertising approaches—banners, pop-ups, search-based, and more—and the effectiveness of each can be easily tracked. One of the most powerful forms of online advertising is the search-based ad exemplified by Google. By having the industry-leading search engine, Google can serve ads that correspond to search terms (AdWords) used by a potential customer. Google also serves ads onto other companies’ online properties through its AdSense network. The popularity of search-based advertising has fueled Google’s rapid growth in revenues and profits.

One of the reasons Google has been successful with advertisers is its extensive use of analytics. Because Google advertising is done for a large client base for small payment increments (a few cents per click-through), much of the analytics must be automated and highly scalable. Google employs self-learning algorithms that are constantly analyzing the efficacy (typically in conversion rates) of different keywords (the primary advertising medium on Google’s own properties), placement on the page, creative material, and so forth. The learning is input for an optimization engine that develops suggestions for advertisers without any human intervention. Advertisers can see the suggestions when they look at the reports for activity relative to their ads. The suggestions may differ for different types of sites, such as entertainment versus publishing. Large Google advertisers also have account managers who can work with the advertiser and provide analytics-based advice. Google’s philosophy is that analytics and metrics will make advertisers more successful in working with the company, so they try to provide as much analytical sophistication as advertisers can use. Thus far the analytically intensive approach seems to have paid off in advertiser loyalty and Google’s growth.

Other approaches to customer analytics primarily focus on retention and cross-selling. For example, the Norwegian bank DnB NOR has built analytics on top of a Teradata warehouse to more effectively build customer relationships. The bank uses “event triggers” in the data warehouse to prompt customer relationship analysts to offer one or more tailored services based on the event. For example, if a customer receives a substantial inheritance, a bank representative will call on the customer to offer investment products. DnB NOR has a set of automated tools that match customer profiles and events and then generate a set of suggested products. Based on customers’ past experience, DnB then chooses the most effective channel through which to contact a customer about the most appropriate products. Using these tools, the company has achieved a conversion rate on cross-selling between 40 and 50 percent and has halved its marketing budget while increasing customer satisfaction.⁶

One of the most impressive users of analytics to retain customers is Tesco. Founded in 1924, Tesco is now the largest food retailer in the United Kingdom and one of the world’s largest retailers. Located in thirteen countries, it operates in every form of retail food channel—convenience, specialty, supermarket, and hypermarket. Tesco’s spectacular transformation began in 1995, when it introduced its loyalty card, the Clubcard. The card functions as a mechanism for collecting information on customers, rewarding customers for shopping at Tesco, and targeting coupon variations for maximum return. Customers earn points that are redeemable at Tesco at a rate of 1 percent of purchase amounts. Tesco estimates it has awarded points worth £1 billon.

The results are impressive. While the direct marketing industry’s average response is only 2 percent, Tesco’s coupon redemption rate is 20 percent and ranges as high as 50 percent. The company’s CEO, Sir Terry Leahy, believes the Clubcard program is also responsible for the company’s success with its Internet business. The world’s largest Internet grocer, Tesco has delivered food to more than a million homes and serves four hundred thousand repeat customers. All online Tesco customers must have a Clubcard, so Tesco can know what they purchase and target online promotions accordingly. By analyzing Clubcard data, combined with a rigorous program of experimentation, Tesco’s Internet business has seen sales surge for nonfood items including home furnishings, music downloads, and homeowner and automobile insurance.

Tesco uses the data it collects on purchases to group customers according to lifestyle. For example, a female shopper who makes weekly purchases, buys what’s on sale, and uses coupons sent in the mail is considered a value-conscious customer. A male shopper who makes three to four purchases per week, all for prepared meals, and rarely changes his purchases regardless of price promotions is considered a convenience-conscious customer. With this kind of insight into its customers, Tesco then tailors its promotions to its customers’ priorities and interests. Tesco says that it issues 7 million targeted variations of product coupons a year, driving the coupon redemption rate, customer loyalty, and ultimately financial performance to market-leading heights.⁷

Firms also use analytics to avoid bad customers while attracting the few customers who defy conventional measures of suitability or risk—an approach known as skimming the cream off the garbage. As we mentioned in chapter 3, Progressive Insurance and Capital One both eschew the traditional industry-standard risk measures. At Progressive, for example, instead of automatically rating a motorcycle rider as a high risk, analysts take into account such factors as the driver’s employment history, participation in other high-risk activities (such as skydiving), and credit score. A driver with a long record with one employer who is also a low credit risk and avoids other risky activities will be rated as a low-risk customer.

Capital One has improved upon conventional approaches to attracting so-called subprime customers—those individuals who by their credit rating are considered to be a high risk for bankruptcy or default. Capital One employs its own proprietary consumer creditworthiness assessment tool to identify and attract those customers it sees as less risky than their credit scores would indicate.

Pricing Optimization

Pricing is another task that is particularly susceptible to analytical manipulation. Companies use analytics for a competitive advantage by pricing products appropriately, whether that is Wal-Mart’s everyday low pricing or a hotelier’s adjusting prices in response to customer demand. Analytics also make it easier to engage in dynamic pricing—the practice of adjusting the price for a good or service in real time in response to market conditions such as demand, inventory level, competitor behavior, and customer history. This tactic was pioneered in the airline industry but now has spread to other sectors.

Retail prices, for example, have historically been set by intuition. Today, however, many retailers are adopting analytical software as part of “scientific retailing.” Such software works by analyzing a retailer’s historical point-of-sale data to determine price elasticity and cross-elasticity (a measure of whether one good is a substitute for another) for every item in every store. An equation is calculated that determines the optimal price to maximize sales and profitability.

Retailers usually begin by using pricing analytics to optimize mark-downs—figuring out when and by how much to lower prices. Some then move on to pricing for all retail merchandise and to analysis of promotions, category mix, and breadth and depth of assortments. Most retailers experience a 5 percent to 10 percent increase in gross margin as a result of using price optimization systems. Some yield even greater benefits. According to a Yankee Group report, “Enterprises have realized up to 20% profit improvements by using price management and profit optimization (PMPO) solutions. No other packaged software can deliver the same type of top-line benefits and address bottom-line inefficiencies. PMPO is the best kept secret in enterprise software.”⁸

JCPenney was one of the earliest large retailers to adopt price optimization software and processes. A few years ago, the company began an analytically intensive program that integrated merchandising, pricing optimization, and the supply chain. This approach helped the company add five points of gross margin, increase inventory turns by 10 percent, and grow top-line and comparative store sales for four consecutive years (2001 through 2004). Operating profits also grew at double-digit rates.⁹

Analytically based pricing software is spreading to other industries as well. In fact, 77 percent of the large business-to-business U.S. enterprise respondents to a 2006 Yankee Group survey that are not already using price management software reported that they have developed a business case to purchase that software.¹⁰

When combined with other analytic data, dynamic pricing can provide a strategic advantage in the face of changing market conditions. At Dell, the chief financial officer used the company’s forecasting system to anticipate the economic downturn of 2000 and 2001. While competitors like Hewlett-Packard and Compaq remained optimistic, Dell cut costs and slashed prices. Dell was able to weather the recession with only a slight downturn in sales—2.3 percent in 2001. Compared with Hewlett-Packard (sales down by 32 percent), Compaq (down 36 percent), and Gateway (down 62 percent), Dell was a success, and its increased market share poised the company for greater success when the economy rebounded.

One cautionary note: most consumers are used to the idea of dynamic pricing in the context of changing market conditions—resorts that lower room prices during the off-season and raise them during peak demand, for example—and probably find it fair. However, companies can face a backlash when they use demand elasticity (the fact that loyal customers will pay a higher price for something than fickle customers) to make pricing decisions. For example, for a time, Amazon.com priced its DVDs higher to people who spent more. When that practice became known to the public, Amazon.com was forced to retreat by the resulting outcry.

Brand Management

Just as analytics bring a heightened level of discipline to pricing, they also bring needed discipline to marketing activities as a whole. Leading companies have developed analytical capabilities that enable them to efficiently design and execute highly effective multichannel marketing campaigns, measure the results, and continually improve future campaigns. Many are using econometric modeling and scenario planning to predict performance outcomes depending on overall budget levels or how much is spent in different channels.

Consider the challenge faced by Samsung in 1999. At the time, the company sold 14 product categories in more than 200 countries in a total of 476 category–country combinations. Amid that complexity, the company struggled to allocate marketing resources effectively. It collected data sporadically and unsystematically and analyzed it only at the country level. Further, its definitions of data, collection procedures, and reporting conventions were not standardized, so interregional analysis was impractical, if not impossible.

This was the situation awaiting Eric Kim when he joined Samsung in 1999 as executive vice president of global marketing. With a worldwide marketing budget of $1 billion, Kim launched an eighteen-month project to gather detailed information on those 476 combinations; assemble and integrate country-level data on a single, easy-to-use site accessible from Samsung marketers’ desktops; use brand data to better set marketing objectives by country and product; and employ the analytical power of software to predict the impact of different resource allocations.

The project resulted in the development of a system called M-Net that not only houses reams of data but also provides the analytic tools needed to make sense of it. Now Samsung managers can assess primary marketing objectives, analyze the results of recent marketing investments around the globe, and build predictive models and what-if scenarios to test future investments. Among other things, M-Net has saved the company millions of dollars by revealing mismatches between some of the company’s marketing investments and the maximum returns those investments could ever yield. Today Samsung allocates marketing resources after systematic analysis of product or geographic market potential, not historical performance.¹¹

In Samsung’s case, analytics were required to maximize the company’s return on its marketing investment. In other cases, industry limitations require creative solutions to acquire the data needed to stay ahead of competitors. Beer, wine, and spirits producers, for example, compete in an industry that is highly dependent on in-store promotions to boost sales, increase market share, and defend against competitors’ encroachments. Each company has access to some sales data from its distributors and external information providers, such as Information Resources, Inc. But these companies have little visibility into the hundreds of thousands of retailers that sell their products to consumers.

Anheuser-Busch Companies successfully surmounted this obstacle through the creative use of mobile technology and an analytical system called BudNet. The brewer derives financial benefits from BudNet by using this information to optimize product mix and price at the local retailer level. The company also uses information from the system to develop local marketing and advertising strategies that are unique to the market.

How does BudNet work in practice? Anheuser-Busch guards details about the system carefully and treats it as a mission-critical weapon. In brief, however, the company arms both company-owned and independent distributors with data entry devices. The devices allow the reps to update inventory levels, but they are also used for a more strategic objective. A sales rep will walk around retail stores looking at the shelf space, displays, and inventory levels of Anheuser-Busch products, as well as the displays, pricing, and shelf space of competitor products. The rep uses a cell phone to upload this data.

Through BudNet, data from thousands of distributor sales reps gives Anheuser-Busch unparalleled insight into what happens to its products at the retail level. The company knows whether a six-pack of Bud Light was warm or cold, whether it was on sale, and what the price of Bud Light was at other stores in the area. In combination with data from IRI (which supplies data on all scanned alcoholic beverages), the BudNet data helps Anheuser-Busch offer even more effective marketing. The company knows that St. Patrick’s Day promotions work poorly in Atlanta, for example, but well in St. Louis. The result has been consistent growth for Anheuser-Busch while major competitors like Coors Brewing Company and Miller Brewing Company have experienced flat or declining sales.¹² Other beverage manufacturers have taken notice of BudNet’s impact and have begun to imitate its analytical capability. Gallo, for example, has a program called the Gallo Edge that allows its retailers to optimize the profitability of the shelf space devoted to wine products.

Converting Customer Interactions into Sales

The strategies described so far relate to arm’s-length interactions between companies and customers, but it is also possible to use analytics to improve the face-to-face encounters between customers and sales-people. Consider how Capital One Health Care, which sells financing services through medical practices for uninsured medical procedures (like cosmetic surgery), outsmarts competitors. Most financing firms market their credit services to doctors the same way many pharmaceutical reps do—known in the business as “pens, pads, and pizza.” By stopping by at lunchtime, representatives hope they can entice the doctor out for a quick lunch break and an even shorter sales pitch. At Capital One, however, reps don’t randomly chase down prospects and hope that a few freebies will clinch the deal. Instead, analysts supply the company’s reps with information about which doctors to target and which sales messages and products are most likely to be effective.

Best Buy is another company that is acting on knowledge gained through customer interactions to improve those interactions (and, not incidentally, to boost sales). Over the last five years, the company has collected data on 60 million U.S. households. To maximize financial performance in each retail store, Best Buy used data-driven insights to develop profiles of eight customer segments.

To translate their insights into increased sales and market share, however, Best Buy needed to understand the best way to serve each segment. It began by establishing a few stores as laboratories. CEO Brad Anderson describes these stores as “our R&D arm for researching customer segments and the value propositions that matter to them.”¹³ The

company used analytics to determine, for example, the impact of pricing changes not only on short-term sales velocity but also on the overall customer experience and its long-term impact on customer perception and sales. It even studied the behavior in each segment of frequent returners—people who commonly seek to exchange products or return them as defective—to learn how to better satisfy these customers.

Incorporating the insights from data analysis and testing at the lab stores, Best Buy developed new store formats for each segment. A “Barry” store, for example, is targeted to young, male audiophiles and videophiles and contains a home theater store-within-a-store. “Jill” stores are oriented to practical, short-on-time mothers. In this format, personal shoppers are available since they are both useful to “Jill” and are a feature that dramatically increases average spending per customer. Other changes are more subtle; for example, the volume on background music is lower than it is in other formats.

As stores convert to these formats, employees are educated about the customer segments who shop at their stores and the best way to serve them. Anderson sees the changes to employee behavior as critical: “We encourage employees to ask customers lifestyle questions and engage in fresh dialogue so that they can recommend suitable solutions. Then we train employees to hypothesize, test, and verify new ways to meet specific needs of the local population.”¹⁴

Best Buy also trains employees to understand financial metrics, such as return on invested capital, so that they can gauge for themselves the effectiveness of customized merchandising displays. Specialized sales-people, such as home theater experts, get additional training that may last weeks. While exact figures are not available, “customer-centricity” based on analytics has delivered significant business results to Best Buy—the new stores formatted around specific customer segments are generating sales at twice the rate of Best Buy’s traditional format.¹⁵

Harrah’s is also able to use the information it collects to improve the experience of the customer while simultaneously streamlining casino traffic. Customers hate to wait; they may be tempted to leave. Worse, from the casino’s perspective, a waiting customer is not spending money. When bottlenecks occur at certain slot machines, the company can offer a customer a free game at a slot machine located in a slower part of the casino. It can also inform waiting customers of an opening at another machine. These prompts help redirect traffic and even out demand. According to Wharton School professor David Bell, Harrah’s is able to tell “who is coming into the casino, where they are going once they are inside, how long they sit at different gambling tables and so forth. This allows them to optimize the range, and configuration, of their gambling games.”¹⁶

Managing Customer Life Cycles

In addition to facilitating the purchases of a customer on a given day, companies want to optimize their customers’ lifetime value. Predictive analytics tools help organizations understand the life cycle of individual customer purchases and behavior. Best Buy’s predictive models enable the company to increase subsequent sales after an initial purchase. Someone who buys a digital camera, for example, will receive a carefully timed e-coupon from Best Buy for a photo printer.

Sprint also takes a keen interest in customer life cycles. It uses analytics to address forty-two attributes that characterize the interactions, perceptions, and emotions of customers across a six-stage life cycle, from initial product awareness through service renewal or upgrade. The company integrates these life cycle analytics into its operations, using twenty-five models to determine the best ways to maximize customer loyalty and spending over time.

Sprint’s goal is to have every customer “touch point” make the “next best offer” to the customer while eliminating interactions that might be perceived as nuisances. When Sprint discovered, for example, that a significant percentage of customers with unpaid bills were not deadbeats but individuals and companies with unresolved questions about their accounts, it shifted these collections from bill collectors to retention agents, whose role is to resolve conflicts and retain satisfied customers.

According to Sprint, the group responsible for these analytics has delivered more than $1 billion of enterprise value and $500 million in revenue by reducing customer churn, getting customers to buy more, and improving satisfaction rates.

Personalizing Content

A final strategy for using analytics to win over customers is to tailor offerings to individual preferences.

In the mobile network business, companies are vying to boost average revenue per user by selling subscribers information (such as news alerts and stock updates) and entertainment services (such as music downloads, ringtones, and video clips). But given the small screen on mobile devices, navigating content is a real challenge.

O2, a mobile network operator in the United Kingdom, uses analytics to help mobile users resolve that challenge. The company pioneered the use of artificial intelligence software to provide subscribers with the content they want before they know they want it. Analytical technology monitors subscriber behavior, such as the frequency with which users click on specific content, to determine personal preferences. The software then places desirable content where users can get to it easily.

The vast majority (97 percent) of O2’s subscribers have opted to use personalized menus and enjoy the convenience of having a service that can predict and present content to match their tastes. Today, O2 has more than 50 percent of mobile Internet traffic in the United Kingdom, and the company continues to explore new ways to use analytics; for instance, it is investigating new collaborative filtering technology that would analyze the preferences of similar customers to make content suggestions. Hugh Griffiths, O2’s head of data products, believes that “personalization is our key service differentiator.”¹⁷

Supplier-Facing Processes

Contemporary supply chain processes blur the line between customer-and supplier-oriented processes. In some cases, customers penetrate deep into and across an organization, reaching all the way to suppliers. In other cases, companies are managing logistics for their customers (refer to “Typical Analytical Applications in Supply Chains” box on the next page).

Connecting Customers and Suppliers

The mother of all supply chain analytics competitors is Wal-Mart. The company collects massive amounts of sales and inventory data (583 terabytes as of April 2006) into a single integrated technology platform. Its managers routinely analyze manifold aspects of its supply chain, and store managers use analytical tools to optimize product assortment; they examine not only detailed sales data but also qualitative factors such as the opportunity to tailor assortments to local community needs.¹⁸

The most distinctive element of Wal-Mart’s supply chain data is its availability to suppliers. Wal-Mart buys products from more than 17,400 suppliers in eighty countries, and each one uses the company’s Retail Link system to track the movement of its products—in fact, the system’s use is mandatory. In aggregate, suppliers run 21 million queries on the data warehouse every year, covering such data as daily sales, shipments, purchase orders, invoices, claims, returns, forecasts, radio frequency ID deployments, and more.¹⁹ Suppliers also have access to the Modular Category Assortment Planning System, which they can use to create store-specific modular layouts of products. The layouts are based on sales data, store traits, and data on ten consumer segments. Some suppliers have created more than one thousand modular layouts.

As Wal-Mart’s data warehouse introduced additional information about customer behavior, applications using Wal-Mart’s massive database began to extend well beyond their supply chain. Wal-Mart now collects more data about more consumers than anyone in the private sector. Wal-Mart marketers mine this data to ensure that customers have the products they want, when they want them, and at the right price. For example, they’ve learned that before a hurricane, consumers stock up on food items that don’t require cooking or refrigeration. The top seller: Strawberry Pop Tarts. We expect that Wal-Mart asks Kellogg to rush shipments of them to stores just before a hurricane hits. In short, there are many analytical applications behind Wal-Mart’s success as the world’s largest retailer.

Wal-Mart may be the world’s largest retailer, but at least it knows where all its stores are located. Amazon.com’s business model, in contrast, requires the company to manage a constant flow of new products, suppliers, customers, and promotions, as well as deliver orders directly to its customers by promised dates. With one of the most complex supply chain problems in business, Amazon.com recruited Gang Yu, a professor of management science and a software entrepreneur who is one of the world’s leading authorities on optimization analytics, as the head of its global supply chain.

Yu and his team began by integrating all the elements of their supply chain in order to coordinate supplier sourcing decisions. To determine the optimal sourcing strategy (determining the right mix of joint replenishment, coordinated replenishment, and single sourcing) as well as manage all the logistics to get a product from manufacturer to customer, Amazon.com applies advanced optimization and supply chain management methodologies and techniques across its fulfillment, capacity expansion, inventory management, procurement, and logistics functions.

For example, after experimenting with a variety of packaged software solutions and techniques, Yu concluded that no existing approach to modeling and managing supply chains would fit their needs. They ultimately invented a proprietary inventory model employing nonstationary stochastic optimization techniques, which allows them to model and optimize the many variables associated with their highly dynamic, fast-growing business.

Amazon.com sells over thirty categories of goods, from books to groceries to industrial and scientific tools. The company has a variety of fulfillment centers for different goods. When Amazon.com launches a new goods category, it uses analytics to plan the supply chain for the goods and leverage the company’s existing systems and processes. To do so, it forecasts demand and capacity at the national level and fulfillment center level for each SKU. Its supply chain analysts try to optimize order quantities to satisfy constraints and minimize holding, shipping, and stock-out costs. In order to optimize its consumer goods supply chain, for example, it used an “integral min-cost flow problem with side constraints”; to round off fractional shipments, it used a “multiple knapsack problem using the greedy algorithm.”

Logistics Management

Sometimes a service company uses analytics with such skill and execution that entire lines of business can be created. United Parcel Service (UPS) took this route in 1986, when it formed UPS Logistics, a wholly owned subsidiary of UPS Supply Chain Solutions. UPS Logistics provides routing, scheduling, and dispatching systems for businesses with private fleets and wholesale distribution.²⁰ The company claims to have over one thousand clients that use its services daily. This approach, captured in the “Don’t You Worry about a Thing” campaign, is enabling UPS to expand its reputation from reliable shipping to reliable handling of clients’ logistics value chains.

Of course, UPS has been an analytical competitor in supply chains for many years. In 1954 its CEO noted, “Without operations research we’d be analyzing our problems intuitively only.”²¹ The company has long been known in its industry for truck route optimization and, more recently, airplane route optimization. The current UPS CEO, Mike Eskew, founded UPS’s current operations research group in 1987. By 2003 he announced that he expected savings from optimization of $600 million annually. He described the importance of route optimization: “It’s vital that we manage our networks around the world the best way that we can. When things don’t go exactly the way we expected because volume changes or weather gets in the way, we have to think of the best ways to recover and still keep our service levels.”²²

FedEx has also embraced both analytics and the move to providing full logistics outsourcing services to companies. While UPS and FedEx both provide customers with a full range of IT-based analytical tools, FedEx provides these applications to firms that do not engage its full logistics services, leading one analyst to observe, “FedEx is as much a technology company as a shipping company.”²³ UPS and FedEx have become so efficient and effective in all aspects of the logistics of shipping that other companies have found it to their economic advantage to outsource their entire logistics operations.

Another company helping its customers manage logistics is CEMEX, the leading global supplier of cement. Cement is highly perishable; it begins to set as soon as a truck is loaded, and the producer has limited time to get it to its destination. In Mexico, traffic, weather, and an unpredictable labor market make it incredibly hard to plan deliveries accurately. So a contractor might have concrete ready for delivery when the site isn’t ready, or work crews might be at a standstill because the concrete hasn’t arrived.

CEMEX realized that it could increase market share and charge a premium to time-conscious contractors by reducing delivery time on orders. To figure out how to accomplish that goal, CEMEX staffers studied FedEx, pizza delivery companies, and ambulance squads. Following this research, CEMEX equipped most of its concrete-mixing trucks in Mexico with global positioning satellite locators and used predictive analytics to improve its delivery processes. This approach allows dispatchers to cut the average response time for changed orders from three hours to twenty minutes.²⁴ Not only did this system increase truck productivity by 35 percent, it also wedded customers firmly to the brand.²⁵

CEMEX’s strategy was powerful because the company changed its focus from the sale of a commodity to the sale of something customers really cared about. In short, the unit of business shifted from cubic yards to the delivery window. This was a simple change in one sense. But CEMEX then oriented its information, logistics, and delivery infrastructure around the delivery window concept, creating far-reaching changes in the company and eventually throughout the industry.²⁶

Conclusion

Analytical competitors have recognized that the lines between supply and demand have blurred. As a result, they are using sophisticated analytics in their supply chain and customer-facing processes to create distinctive capabilities that help them serve their customers better and work with their suppliers more effectively.

The discipline of supply chain management has deep roots in analytical mastery; companies that have excelled in this area have a decades-long history of using quantitative analysis to optimize logistics. Companies getting a later start, however, have clear opportunities to embrace an analytical approach to customer relationship management and other demand processes.

In part I of this book, we have described the nature of analytical competition. In part II, we lay out the steps companies need to take and the key technical and human resources needed for analytical competition.