by: Issues: July / August 2001. Tags: Strategy. Categories: Strategy.

Understanding and describing algorithms may be a formidable and unappealing task, but it is a challenge that managers today must meet, if they are to survive and win on the Net. The reason is that e-business today is driven by algorithms, step-by-step coded procedures that determine how tasks on the Internet are performed. Examples of such tasks include a potential customer browsing with a shopping cart or suppliers bidding to do business with a certain manufacturer in a Web auction. In clear and simple language, this article helps readers understand algorithms and the critical role they play in an e-business strategy. Using market examples to illustrate his points, he shows managers how they can manage – if not thoroughly understand – algorithms to their advantage.

E-business today is driven by algorithms, step-by-step coded procedures that determine how tasks on the Internet are performed. Examples include a customer browsing an e-store with a shopping cart, or a Web auction where suppliers bid to do business with a major manufacturer. Algorithms are the intellectual capital of the Internet, and not surprisingly, firms doing business on the Web live or die based on the quality of their algorithms.

Understanding algorithms, and what can and cannot be done “algorithmically,” is essential to a successful e-business strategy. Some of the finest minds anywhere are now working on algorithms for e-commerce, and an effective competitive strategy begins by understanding what these people are up to and how that is driving, and will continue to drive, the world of e-commerce.


If you wish to perform a task on the Internet, it has to be programmed, or broken down into its logical steps, which in turn must be translated into lines of code. For example, a WWW site can sometimes seem quite slow, even when it belongs to a major company and is at the end of a very fast connection and server. What’s going on? The answer lies in understanding the algorithms at work.

The host computer of a B2C site will first establish your identity by reading your computer’s Internet address or accessing a “cookie” that it left on your PC the last time you visited it. Once it knows your Internet “name,” the host will then try to learn something about you. These details may come from the information you supplied when you registered with the site, but it may also include an interpretation of data that you provided, though you were not aware that you had done so. The host may fire off a “Who is this?” request to a company such as Engage Technologies (, whose databases hold the names of millions of computer users together with many of the WWW sites that users have visited. This data is called click-stream data.

Engage also has best-of-breed, data-mining algorithms that analyze this data in many different ways. One set of algorithms “clusters” computer addresses based on the sites that their users have visited and any other data that these users have provided. Marketers call this profiling. The clusters that are produced group users according to different kinds of consumers. For example, a Harris Interactive study of 3,000 Internet shoppers last year identified 17 percent of the sample as Time Sensitive Materialists, a cluster most interested in convenience, and one that is not particularly price-sensitive. The study also identified 23 percent as Clicks and Mortar Shoppers, a cluster that browses on-line but prefers to buy off-line, and is more likely to be female with privacy and security concerns. Another 20 percent were identified as Hunter-Gatherers, typically 30- to 49-year-olds with two kids, who are most likely to visit sites that provide information and comparisons.

While you are waiting to see something on your PC screen, Engage will run some fast, real-time algorithms and send a reply back to your host, reporting what it has found out about you or your cluster. The host will then run some more algorithms, called optimizations. The host may check its advertising database and select a set of advertisements for you to view. It will check its products catalogue and determine which “value packages” (or products) to show you on the opening page and which menus or links to additional products it will display. Finally, it will use its knowledge of your consumer “cluster” to tailor the product/value packages displayed. This may include prices calculated just for you.

For example, a customer identified as a Time Sensitive Materialist may be shown high-end products with prices that include next-day delivery and a warranty. A Clicks and Mortar shopper, on the other hand, may be shown economy products with low prices, with a number of delivery and/or warranty options available at extra cost. B2B sites function differently, since the users are well known to the host. Participation in a B2B exchange, or dealing with an industrial buyer on-line, usually requires that the co-operating firms know each other well. Since the B2B host knows each visitor, the host’s algorithms can customize the screen displays for the maximum possible appeal for each individual log-in (called “1-to-1 marketing”), much like a salesperson who is personally familiar with each customer. When a visitor asks for a price/delivery quote on a part, algorithms on the host computer check the part’s availability, first from inventory and then from scheduled production. The algorithm may then interact with the production schedule to see when it might be possible to manufacture the order. From these queries, the algorithm will prepare a delivery date and price quote based on its assessment of the ease and timing of manufacture, the demand for the part in question, and the price sensitivity of the buyer.


State-of-the-art algorithms are a necessity to compete successfully in e-commerce. As we watch the development of the Worldwide Web, we can see how improved algorithms affect the competitive balance between firms. For example, consider the case of supply chain management. ERP software suppliers such as Oracle, SAP, PeopleSoft, Baan and J.D. Edwards have held a very strong position by using a human-driven, information-systems approach based on the philosophy that humans can manage a supply chain if they are provided with information that is current and accurate. However, a new group of companies, led by i2, Aspen Technologies and Manugistics, have developed supply chain optimization algorithms. A recent article in Business Week (Oct. 13, 2000) reports that while the ERP market is growing at five percent annually, the supply-chain optimization market is growing by 40 percent annually. Not surprisingly, the ERP companies are now scrambling to incorporate elements of optimization into their products.

Farm machinery manufacturer John Deere & Co. has recognized the new competitive reality: “A customer asks when a tractor will be delivered,” says Jay Fortenberry, worldwide logistics director at Deere. “We have to respond in seconds, or they may go somewhere else. It’s all driven by order fulfillment.” Competitive survival requires that processes be algorithmically optimized. “A human might be able to figure out a single plant,” says Fortenberry, “but it’s just not possible for him or her to try to optimize the whole process, customer to plant to suppliers and back.”

For the moment, the supply chain optimizers are win winning this skirmish. Can the ERP companies respond? The outcome may well result in the formation of strategic alliances between the ERP houses, which have extensive client lists, and the optimizers, which own the improved algorithms. Just recently, ERP supplier SAP hooked up with ILOG, a leading-edge supplier of optimization software, to add optimization to its popular MySAP supply chain management software.

Algorithm wars do not occur exclusively in cyberspace. High-flying People Express airline met a sudden death in mid-1985, when it flew into new pricing algorithms developed by American Airlines’ Decision Technologies group (now Sabre). Donald Burr, founder and CEO of People Express, believes that the major carrier’s use of sophisticated computer programs to immediately match or undercut his prices ultimately killed People Express. For “sophisticated computer programs” read “algorithms.”

National Car Rental transformed itself from the big loser in the rental car business to one of the winners by implementing state-of-the art pricing and inventory control algorithms. Intersil Corporation (formerly Harris Semiconductor) turned an annual loss of $75 million into an annual profit of $25 million by implementing algorithms that optimized production at its 30 worldwide manufacturing facilities (all currency in U.S. dollars).


E-commerce has emerged as a major competitive arena for the algorithm developers. On the Internet, everything has to be programmed, so every company operating on the Internet is using algorithms. Firms with an e-commerce presence don’t have to be sold algorithms; they cannot exist without them. There are good algorithms and bad ones, simple and complex ones, effective and ineffective ones. If my algorithms are better than yours, I will have an advantage that you will have to somehow overcome. If you implement a bad algorithm, say one that computes the price of your product as $6.99 instead of $699, the damage can be swift and severe.


Not all algorithms are designed to handle procurement, supply chain optimization, or technical functions such as aircrew scheduling or vehicle routing.

An innovative group of algorithms are the Worldwide Web “salespeople” (for examples, visit “Emily” at or “Eve” at While these algorithms appear like “Help” features, you should not be misled. Just as Frank at the used-car lot will help you understand cars, Emily’s help is chiefly an effort to make the sale. Visit and try designing your own kitchen and cabinetry. You can lay out the basic cabinets, counters and flooring, and change finishes with just a mouse click. When you are happy with how your design looks, you must head to Home Depot for a final consultation with a designer before ordering your cabinets. Mill’s Pride’s algorithm allows you to be creative and to have fun, but it also sells cabinets!


While many algorithms merely automate existing business activities (such as “making the sale”) and are, therefore, not revolutionary, one group of algorithms is fundamentally changing the way firms are managed. These are the pricing, or “revenue management,” algorithms—pioneered by Sabre beginning in the mid-1980s—that are driving the move to individualized pricing.

New pricing algorithms can compute a price for a product and display it on a page that is customized to each individual customer’s profile. This price might depend on who the customer is, the market segment they belong to, the time of day, the day of the week or the point in the sales season or other factors. Developing these pricing algorithms is a specialized and very competitive business. Firms with skills in this area are in high demand, in part because most firms that switch to algorithmic pricing can increase their revenues by five to 10 percent with no additional cost, other than the cost of the algorithms and their implementation. Such incremental revenue has a disproportionate effect on profits: a five to 10 percent increase in revenues, with no increase in costs, can often double a firm’s profits.

Pricing algorithms usually take the amount of product available for sale as a given. Once American Airlines decides that a Boeing 767 will fly from Chicago to London at 9 a.m. on March 1, the amount of product available for sale is determined. Sabre’s pricing algorithms then take over and sell that volume of product while attempting to maximize the revenue received. The same approach works for goods: If you can decide how much merchandise to put into the marketplace, the pricing algorithms will sell that volume while attempting to maximize revenue.

As the airlines have noticed, these pricing algorithms render the concept of market share pretty meaningless, since firms can choose whatever volume they wish to sell, and hence the market share they will claim. Profitable market share has become more important. This recognizes that there is no point in having 50 percent of the market if you lose money on every unit sold. Successful pricing means having a higher “profit per market share point” than the competition.

The view that pricing starts with an inventory of product is now old-fashioned, since it reflects a view that manufacturers produce to inventory and then try to sell the inventory. Today, supply-chain management is increasingly about optimizing the way that the entire supply chain—from raw material suppliers through inbound shippers, the manufacturing plants, and outbound shippers and warehousing—responds to customer demand. The adage “Make it and they will buy” made sense when it took years to design, build and distribute a product. However, those days are gone, and gone they will remain. Retailers such as Sears, Roebuck and Wal-Mart Stores Inc. now tell the manufacturers what to make, when to make it, how to ship it, where to send it, and which time and day it must arrive.

It is now recognized that real-time, individualized product pricing can help with supply-chain optimization. The availability of a product from the supply chain, either from inventory or scheduled production, can be taken into account when providing price quotes to customers. A customer who requests a product that has to be specially scheduled or is difficult to slot in to the existing manufacturing schedule may be quoted a high price. The same customer looking for the same product on a different day may be offered it at a low price; for example, if the part is available in inventory or if the loading on the plant makes for easy manufacture. Introducing variability into product prices can enhance revenues and also reduce the variability that the supply chain must cope with. In either case, supply-chain costs are cut, enabling the firm to win in both ways.

The merger of supply-chain optimization and algorithmic pricing is well under way. Talus Solutions was a very successful revenue management specialist with an extensive client list featuring many major corporations. Manugistics is a leading supply chain optimizer with a similarly impressive client list. In December, Manugistics completed the acquisition of Talus Solutions for $366 million. The new Manugistics is marketing Enterprise Profit Optimization, the merger of supply chain optimization and price and revenue optimization.


The importance of algorithms is behind the many attempts to patent basic business logic and calculations, and to trademark catchy, commonly used phrases. If you come up with what you think is a new way to make a forecast or calculate a price, you might find that it has already been patented. You may also find that common words and phrases that you have been using for years now appear on the Internet with ™ posted alongside.

Filing a patent on algorithm logic appears to be more of a publicity stunt for vendors than an effective way of preventing their replication. There are good business reasons why firms hold their algorithms confidential. It is often very difficult to reverse-engineer a competitor’s logic. Consequently, short of raiding employees, it is difficult to determine whether a competitor has stolen your logic. The great majority of patents on algorithms have not been defended, perhaps in part the result of the many competing claims. There does, however, remain the threat that if you come up with some really good logical way of doing something on the Internet, or if you come up with a catchy phrase to describe that something, you may receive a letter from a lawyer claiming “that’s ours.”


Should you build algorithms yourself or buy them? Conventional wisdom says that a firm should not outsource its mission-critical activities. But algorithm development is a very specialized business with very long development curves. Where and how to acquire algorithms is now a critical, strategic decision for many firms.

The leading-edge algorithm supplier is Sabre Inc., which started life as American Airlines Decision Technologies (AADT) and was responsible for the initial pricing, crewing and routing algorithms that have transformed air-carrier management. In the early ’80s, AADT employed less than one dozen algorithm developers. Fifteen years later, that number had grown to 500 developers as the division expanded its external consultancy practice. The market success of this unprecedented group of 500-plus people with advanced degrees in management science, mathematics, operations research or industrial engineering precipitated a corporate transformation. AADT first became Sabre Decision Technologies, a division of AA, and was later spun out of American Airlines to become independent as Sabre Inc. Sabre was first with many algorithm innovations and, consequently, the list of companies that use Sabre’s algorithms includes many industry leaders. Sabre’s 2000 revenues were $2.6 billion. In a recent move that will change the way firms access algorithms, Sabre decided to make 30 of its most important algorithms available on an ASP basis. There is good news and bad news in this decision: off-the shelf algorithms are better than none, but they are likely no match for custom algorithms developed by a leading-edge supplier.

Dallas-based i2 Technologies Inc. is one of a new kind of company whose credibility hinges on its algorithms. Executives say the company has saved its clients $7.5 billion over the past five years and will save them $75 billion by 2005. i2’s reported revenues of $1.1 billion for 2000, with an annual growth rate of just about 100 percent.

Many specialists are trying to construct algorithms that compete with those supplied by industry leaders such as Sabre and i2. There are a host of suppliers that serve various market segments, and many of the new firms are spinoffs of Sabre or other firms spun off from Sabre. Penetrating this secondary supplier maze is not easy, but it may be necessary, since the major players may be locked in strategic alliances with your competitors.

Should you develop your own algorithms? The good news is that some of the required components can be purchased. An ERP package might provide the basic information systems’ functionality; there are also several options for data warehousing and data management. An optimization engine, such as ILOG CPLEX or Dash XPRESS-MP, can perform some of the heavy number crunching. You will also need to assemble and invest in a technical group to develop, maintain and nurture your databases, knowledge bases and algorithms. All this is going to be expensive, and success is not guaranteed. What’s more, if you are successful, there is no rest. Continuous improvement is a must, since today’s successful algorithms can quickly turn to e-junk when someone else has a better idea.

Once, a shareholder of a chain of small-town hardware stores had a nightmare when he or she heard that Wal-Mart was coming to town. Perhaps today’s shareholder’s nightmare occurs when he or she hears that a major competitor has struck a deal to buy its algorithms from Sabre, i2 or Manugistics.

Leave a Reply

Please submit respectful comments only, including full name, professional title, and contact information (only name and title will be posted). Required fields are marked *