Sun Microsystems Inc. chairman Scott McNealy forecast that “With recent advances in wireless and information technology, even our cars could … call for bids whenever the fuel tank runs low, displaying a list of results from nearby gas stations right on the dashboard.”1
It sounds far-fetched. But dynamic pricing — where prices respond to supply and demand pressures in real time or near-real time — is making inroads in many different sectors, including apparel, automobiles, consumer electronics, personal services (such as haircuts), telecommunications and second-hand goods. The advent of the Internet led to cost transparency, decreased search costs and ease of price comparison. Some observers concluded that as a result, prices would decrease and equalize across different channels, and that fixed prices would continue to be the norm.2 However, price dispersion continues to be widespread and dynamic pricing is entering new sectors. EBay Inc. used auctions to sell more than $20 billion worth of goods in 2005. Ford Motor Co. sold more than $50 billion worth of automobiles in North America with demand-based DP in 2003, exceeding profit targets by $1 billion.
Fixed prices are, after all, a relatively recent phenomenon — a product of mass manufacturing that came about after the Industrial Revolution. Before that event, fixed prices were the exception; DP was the norm, with both buyer and seller able to benefit in many DP transactions.
However, apart from airlines and hotels, which employ DP routinely, most companies still use relatively simple strategies for determining prices: competitive pricing (pegging prices to competitors’ prices) or cost-plus pricing (calculating the cost of a good or service and adding profit). Now dynamic pricing optimization offers companies in many other sectors the alternative of raising average realized prices in the face of increased pricing pressures.
Four principal reasons are driving the increasing use of DP today:
- More companies can now access and deploy the technology for DP at affordable prices in new product and service categories.
- Recent research shows that with the right approach, consumers will accept DP even if they are currently buying using fixed prices.
- Increased pricing pressures and supply constraints in different industries are driving companies to look at new ways of extracting value and reallocating demand.
- In their continuing bid to increase efficiency, many companies (having already integrated upstream supply chains with their operations) are now turning to the downstream aspect, where DP is a natural consequence.
Dynamic pricing, managed well, offers a feasible and attractive path to increase revenues and profits. (See “Why Dynamic Pricing?”) Implementing DP can improve revenues and profits by up to 8% and 25%, respectively.3 But it’s not just that DP offers greater profits. For example, for personal services such as haircuts and public services such as road space in metropolitan areas, DP can be used to reallocate demand to more suitable times and manage a limited supply base. And where products have a limited shelf life with a salvage value (for example, apparel), DP can be used to improve realizations from fast-moving lines of goods by raising prices in real time or near-real time, and similarly to push slow-moving goods by lowering prices. Also, in formats such as Internet auctions (for example, eBay) and group buying, DP can aggregate bigger audiences than is possible in a physical setting. New, excess and reassigned inventory can be sold for higher realizations using an auction format such as eBay.
Forms of Dynamic Pricing: Posted Price Mechanisms and Price Discovery
Dynamic pricing differs fundamentally from fixed pricing because it allows prices for the same good or service to change by customer, time, aggregate demand and other situation-specific parameters. There are two broad categories of DP: (1) posted prices that customer can see; and (2) price discovery mechanisms, in which customers determine prices through their own actions during the transaction.
Why Dynamic Pricing?
Fixed prices are a form of posted prices, of course. When companies fix the posted price of a product or service for a relatively long period, this is mainly due to lack of accurate demand information, high transaction costs associated with changing prices continuously, and the huge investments required for the software and hardware necessary for implementing DP. But the increasing power of information technology now enables access to demand information in real time, permitting a matching of demand and supply and, therefore, intertemporal DP. Technology enables a typical retailer to make pricing and inventory decisions for hundreds of thousands of products. For example, there can be as many as 50,000 stock-keeping units in a grocery store or drugstore, with items possibly priced differently across stores.
Dynamic posted pricing includes systems such as revenue management for airlines — also called yield management (based on demand with a supply constraint) — and demand-based variable pricing (prices change according to the demand for a product), or a combination of these approaches. Revenue management and demand-based pricing use historical data and mathematical models to predict demand at future points in time. Different prices are then set for these different time points according to predicted demand, as well as adjusting prices for actual demand.
When the price discovery approach is used, prices are determined by the active participation of the customer in the transaction. In other words, the price changes during the transaction, whereas with dynamic posted pricing the price changes across transactions. Price discovery-based mechanisms include different forms of auctions, group buying and negotiations.
In auctions, customers bid (up) to buy the product of choice. EBay is the most famous example of the straight auction where the highest bidder wins. In a reverse auction, suppliers bid (down) to sell goods to a buyer. Many large buyers employ this technique, sometimes to squeeze upstream costs. Vendors such as FreeMarkets Inc. have provided the platform to implement reverse auctions for clients including United Technologies Corp.
When Google Inc. went public in 2004, it employed a modified Dutch auction to price its shares. Would-be investors bid a price at which they were willing to buy a minimum of five shares. Google revealed the maximum number of shares being sold and a potential price range that was adjusted over time according to market conditions. Investors stated the number of shares they wanted and at what price. Once the minimum clearing price was determined, investors who bid at least that price were awarded shares. When there were more bids than shares available for a particular size of share lot, allotment was on a pro-rata basis — awarding a percentage of actual shares available based on the percentage bid for. (A similar type of auction, the Yankee auction, differs in that winning bidders pay exactly the prices they have bid.)
In group buying situations, customers, who may be known to one another or be complete strangers, aggregate to bid (down) the price — a form of volume discount obtained dynamically through the actions of the customers. For example, LetsBuyIt.com sells excess inventory for companies using the group buying process. Finally, negotiations are a well-known way of arriving at a mutually agreeable price. Software can now be used to automate negotiations in some contexts. (See “The Seven Types of Dynamic Pricing.”)
The Seven Types of Dynamic Pricing
The conventional view is that DP is limited to perishable products, such as hotel rooms and airline seats, whose usefulness and functionality last a finite length of time; in the event that they are not used prior to expiration, their value diminishes to zero. There is, however, another category of products and services whose value depreciates with time but has a salvage value, including apparel, automobiles, cell phones, computers, consumer electronics items and some packaged goods. As new models are introduced in the market, the older models become less desirable and potentially obsolete. After a limited period — between one and six months, depending on the product category — unsold products have a residual value that may be 20% to 75% less than the initial asking price.
Consider clothing lines that arrive on store shelves in autumn and have a shelf life until the end of the season. In a fixed-price regime the apparel lines will be marked at a constant price for three months, regardless of whether a line is moving fast or slow, and then will be marked down in end-of-season sales by up to 75%. In recent years, consumer electronics and computers have begun to behave like apparel: fixed prices for a season and then large markdowns. New cell phone models get marked down by as much as 30% after three months and by another 30% after six months. They then move to other channels and get marked down further. New automobiles have a shelf life of only a few weeks; if they have not been sold from showrooms by then, they get marked down.
There is an alternative for these products. Instead of setting a fixed price for two to four months and then holding “fire sales,” and instead of depending on ad hoc and fragmented data to decide on rebates, companies can use DP approaches to both raise and lower prices right from the beginning. Managers can calculate these price changes by examining the latitude of price acceptance.
How to Price Dynamically: Latitude of Price Acceptance
For the different categories of products that customers buy, they have a latitude of price acceptance, which is a range of possible prices within which price changes have little or no impact on their purchase decisions. A McKinsey & Co. study shows that LPAs can range widely: from 17% for branded consumer health and beauty products to 10% for engineered industrial components and apparel to only 2% for some financial products.4
LPA can be discovered through three approaches. The first is to observe the range of prices for a particular product that customers find in different channels. For instance, a Sony DVD player will be found at one price in a Best Buy, at another price in a Sony shop and at a third price at a supermarket such as Kroger. Customers in effect learn an LPA from this observed price range. The second approach to discovering LPA is based on surveys that test consumers’ willingness to pay. The third approach uses analysis of actual demand elasticities in geographies, products, sales channels and customer segments.
Primary data that I have collected with consumers of DVD players, mobile phones and apparel (See “About the Research”) shows that when a company uses DP, there is no difference between larger and smaller consumer LPAs in either the relative impact on purchase intentions or their perceptions of price fairness compared to a fixed-pricing regime.5 Companies can create LPA ranges in consumers’ minds through the variation of product prices across different channels and geographies and through appropriate messages. For instance, Northwest Airlines Corp. sells tickets on the Internet, by telephone and at airline counters. Like other airlines, Northwest sets different prices for the three channels and changes prices more frequently on its Web site than on the other channels.6
Corporations that move to the higher end of an LPA band can substantially increase profits. The McKinsey study suggests, for example, that a financial services company moving from the middle to the top of a 2% LPA band for personal loan products would generate an 11% increase in operating profits for those products. By implementing the appropriate version of DP and by remaining cognizant of the limits prescribed by LPA, companies may very well increase their overall revenue realization — instead of the guaranteed decrease in sales inherent in fixed prices.
So when should companies employ dynamic pricing? In the following sections, I outline rules for when to use DP, what form of DP to use in which situation and issues in implementing DP.
Eight Situations For Using Dynamic Pricing
1. The bigger the market, the larger the number of customers and the greater the number of transactions, the greater the opportunity for DP.
The best example of this DP situation is a stock exchange like the New York Stock Exchange Inc., where prices vary in real time continuously. The large numbers of transactions and lots (products) enable numerous buyers and sellers to bid for the products (company stocks), and the price changes according to the demand.
Another example comes from the highly competitive American automobile market, which is the world’s largest, with more than 15 million new vehicles sold. Over the past few years, the Big Three (Ford, General Motors and DaimlerChrysler) have tried to maintain market share by resorting to aggressive price decreases through incentives and rebates. Ford used DP in 2003 to preserve its margins, however. Instead of providing a single offer price with uniform incentives and rebates for a car model across all its markets in the United States, Ford used detailed market data about how a model was selling in different regions and raised and lowered its final prices to the customer (including rebates and incentives) according to near-real-time demand. Using daily sales data from dealers, Ford has changed rebate levels as frequently as once a week.7
Pricing based on evolving customer demand for different product lines has allowed Ford to hold average prices across all models steady in certain years; prices for some models such as Explorer decreased, but those of other models such as Taurus increased. While net vehicle prices at General Motors fell 2% during the first quarter of 2003, a 0.2% average price gain across all models allowed Ford to raise its revenue per vehicle during the same period.
In this example, Ford employed a variant of demand-based DP that depends on a near-real-time understanding of demand for each model category by geography and price. The large number of transactions provides the base on which judicious use of computing power, based on real-time sales information, leads to profitable DP.
2. The more the customer is involved in the process, and the greater the heterogeneity in valuation that customers put on the same service, the greater the opportunity for DP to reallocate and manage demand.
Maria, a hairdresser in a London suburb, was turning away customers on Saturdays because she did not have, or want to add, the required capacity. During the week, however, her premises were frequently half full. (She closed the salon on Sundays.) Maria had four principal groups of customers: busy professionals living in the area, pensioners, homemakers and mothers with children. She observed that the professionals tended to come only on Saturdays. Maria considered raising her prices on Saturdays and decreasing them on Tuesdays and Wednesdays as an incentive for some of the mothers, pensioners and homemakers to switch to midweek visits, so that she could cater to more of the professionals who might be willing to pay higher prices on Saturdays.
The economics of the idea were sound given her customer base, but Maria was not sure how customers would react. What would they think about paying different prices for the same basic service, a haircut? Maria informally floated the idea among her customers. To her surprise, she found that many were not averse to the idea; the pensioners, homemakers and mothers would benefit from lower prices and had schedules that were flexible enough for them to come midweek. The busy professionals were happy to pay a higher price and not have to wait the usual 30 to 45 minutes (in spite of an appointment) on Saturdays. Customers placed different valuations and had different constraints on their time.
Getting the customers involved in the pricing process (and pricing by the relative demand for her perishable service) enabled Maria to implement a form of DP, which at first blush did not appear to be something that customers would find fair and equitable. Maria also directed demand toward available capacity and segregated her customers by the ability to pay. Some form of customer heterogeneity is present in every market, and Maria was able to leverage the heterogeneity in the valuation for her services. DP increased Maria’s number of customers served, revenues (by 10%) and profits (by 25%). Customer involvement in the DP process obviated possible negative backlash.
In the personal computer industry, Dell Inc. is able to predict its near-term demand in the United States and Western Europe and adjust prices to moderate demand. By lowering prices for those product configurations that are in stock or already in the pipeline and raising prices for configurations that may cause undue ripples in its supply chain, Dell is able to optimize customer demand. Low-cost airline easyJet plc also follows a yield management form of pricing that allows it to moderate and direct demand.
3. Products and services that have a well-defined shelf life (that is, they eventually become obsolescent) are amenable to the use of demand-based DP, even if they are not perishable in the conventional sense but nonetheless have a salvage value.
Clothing is one of the largest categories of perishable products. Although people do not typically think of clothing as perishable, fashion changes quickly, so retailers need to sell clothing before it turns unfashionable — usually before the end of the season, at which point an unsold line of apparel will have a salvage value that is a fraction of its original price. (In comparison, the value of an unsold airline seat becomes zero as soon as the plane takes off.) Like airlines, apparel retailers can employ statistical models that estimate demand and then track actual demand in different lines and adjust prices. In addition, a clothing retailer can obtain more supplies of a fast-selling line of clothing (unlike airlines, which cannot change the number or type of seats on a scheduled flight) — provided that information is available early in the season.
The difficulty is that a big retail store may stock tens of thousands of different clothing items. To perform DP, a retailer would have to work up an elaborate mathematical model of how each product performs in the market and then estimate the trade-off between different pricing levels and the chances of purchase as the product proceeds toward “perishability.”
As computing power and storage have become cheaper, however, demand-based DP has become feasible for clothing retailers. Today, a customer shopping at Macy’s, Gymboree, Ann Taylor or any of several other retail chains may well be paying a price that changes on a daily basis according to variations in aggregate demand for different lines of clothing. Macy’s does not reprice merchandise in its stores more than once a day, but on its Web site every posted price can be reset as frequently as desired. A statistical model can be used to update prices continually to maximize gross margins.
CXB, a pseudonym for a department store that sells apparel to mid- to upmarket clientele, gets new lines of menswear, women’s wear and children’s wear every autumn that it sells for the marked price until around mid-November. At that point, promotions (10% off, 20% off and so on) start and continue until an end-of-season clearance sale just after Christmas, when prices are usually marked down between 50% and 75% from the original price.
My research suggested that if CXB were able to implement demand-based DP successfully, it could raise profitability by up to 15%. As a result, management launched a demand-based pricing system. Newly installed inventory and transaction recording systems enabled near-real-time tracking of sales momentum by different lines of apparel in each store. When demand for a line increased, the price would also increase. Prices of slow-moving items would be cut. Thus the price of the Roca line of menswear, which came in on September 1, rose by 2% overall on September 8 and by another 3% on September 15, since the Roca line was moving very quickly. On the other hand, the price of the Canyon line of menswear was lowered by 7% on September 8 since it was not selling at all. In the first season that DP was implemented, prices were changed on a weekly basis in response to the variations in demand. At the end of that season, the clearance sale covered only 30% of the usual volume of apparel that went on sale. In the following season, prices were changed twice weekly, and the end-of-season clearance sale was left with only 20% of the baseline volume. CXB changed its prices within a range of 10%, which was determined as the latitude of price acceptance for its clothing lines. In both seasons, profitability increased as a consequence of higher sales.
4. The more that a company needs to sell excess or reassigned inventory, the greater the potential role for DP.
“Investment recovery” from sales of assets that are no longer needed (such as surplus inventory), at significant discounts, is a standard practice at almost all companies. The use of DP can increase the recovery price and therefore decrease the loss due to write-downs. Online sites such as LetsBuyIt, eBay and uBid.com, among others, have persuaded different customers to pay different prices using various DP mechanisms for new, excess, mature or reassigned inventory. EBay in effect acts as an inventory clearinghouse for companies such as Hewlett-Packard Co. and Sun Microsystems. Use of these channels has increased both the pool of potential buyers and the price realization for companies.
LetsBuyIt, an online B2C Web site based in Gateshead, United Kingdom, uses a price-discovery type of DP to sell a limited assortment of mainly white goods. Prices depend on the number of people who have signed up to buy a product. If, for example, between one and five people sign up to buy a Palm V PDA, the price decreases from £300 (the original retail price) to £259. If six to nine people register, the price goes down to £245. If 10 people or more sign up, then the price falls to £225, the lowest price offered. The consumer can sign up to buy at only the lowest price (that is, only if 10 or more people sign up) or can opt to buy at the prevailing price when the “co-buy” closes. The potential number of people from whom the “group” can emerge to accomplish the co-buy is orders of magnitude higher than the number in a brick-and-mortar situation. The Internet provides an aggregation and access level that otherwise would not be possible.
5. The greater the possibility of using one-off transactions to obtain inputs for production, the greater the potential use of DP (specifically, reverse auctions).
FreeMarkets Inc. (now part of Ariba Inc., which is headquartered in Sunnyvale, California) creates markets called competitive bidding events where suppliers (sellers) bid for the business of manufacturers (buyers). The FreeMarkets approach is to provide the total package of services necessary for a reverse auction process. The staff works with the suppliers to prequalify them for the event.
In one case, 35 vendors spread around the world bid to supply a global manufacturer with printed circuit boards in 12 lots. The manufacturer planned to award the business to the lowest bidder. In this reverse auction, bidding started at a reserve price level (corresponding to the price at which the manufacturer had bought the components the previous year) and prices then decreased. This auction saved the buyer 43% from the previous year’s price.8 On average, buyers using FreeMarkets have experienced savings of between 2% and 25%.
Smaller organizations can experience similar benefits if they band together to exercise buying power though FreeMarkets or industry vertical portals. The volume of purchases the buyers collectively offer can be sufficient to induce several suppliers to enter into a reverse bidding war. In essence, by increasing the number of possible sources of goods, the buyer is using a price discovery mechanism to get a discount.
6. Where the final price has little relation to cost and the product can be viewed and evaluated at a distance, DP methods such as auctions can be used to determine a price range.
Los Angeles-based Sun Jewelry has used eBay as a test bed for the fixed prices it offers in its physical stores. Sun lists an item on eBay, observes the bidding process and the final prices offered, and then uses the information gathered to set prices for similar products in the store. Sun saves itself the cost of conventional market research to determine customer willingness to pay; packaged goods companies and automakers, for example, have traditionally spent large sums to determine appropriate price levels through extensive contacts with potential customers. Sun is able to set prices that are representative of what customers are willing to pay, potentially increasing the profits it can make from its lines of jewelry through higher volumes.
7. Where there is a need to recover money quickly for improved cash flow, a DP method such as negotiations can be very useful.
On many occasions, an insurer and a doctor will haggle about payments over claims that the doctor has made. The insurer ordinarily has 45 days after processing the claim to send payment to the doctor. The doctor would like to get the money sooner, while the insurer has an incentive to pay out less money.
Automated software now exists that allows insurers and doctors to negotiate. Some insurers and doctors have agreed to automate their negotiations because it is cheaper than having outside negotiators (working for the insurance companies) call doctors, who are often too busy to talk. Insurance payments, even at the low end, tend to be high enough to make the negotiation software worthwhile for both the doctor and the insurer.
As the negotiations proceed, the software adjusts itself. It has been found that 98% of doctors reject the first offer and come back with a counteroffer. However, 75% of the negotiations end after the second offer. If a deal cannot be made, then the insurer pays the full claim under the original 45-day time frame. In more than 20,000 transactions executed during an 18-month period in 2002 and 2003, the average negotiated settlement using the software was $2,000, with insurers saving an average of $400 per settlement.9 The insurer improves its bottom line and the doctor is able to improve his or her cash flow.
The traditional Dutch auction for a sale lot begins with a high asking price, which is lowered until some participant is willing to accept the auctioneer’s price. That participant pays the last announced price. Google’s intention in using a modified form of Dutch auction for its initial public offering was to obtain the highest price for itself, to prevent or minimize the typical first-day jump in prices of its shares, and to make the IPO accessible to a high number of potential buyers.
Google’s IPO sold 19.6 million shares at $85 each to raise $1.67 billion. Because of the low market price for other Internet companies at that time, Google shares were initially priced near the low end of that range, but they closed the first day just above the middle of that range. There were no huge swings in price, up or down, in the immediate aftermarket. Even in a fairly adverse market, Google’s Dutch auction IPO undercut investment bankers — their average commission on the Google deal was 2.8% compared with 4% to 7% in conventional IPOs.
Four Keys to Implementing Dynamic Pricing
This article has so far viewed dynamic pricing from the perspective of the company. In implementing DP, it is important that customers accept the practice and not view it as iniquitous. Four factors must be carefully considered to ensure a successful implementation.
First, DP cannot be perceived to be inequitable. That can be deadly for a company when consumers can communicate and compare experiences with other consumers — as was the case with Amazon.com Inc.’s attempt to price DVDs differentially in 1999. Some Amazon customers discovered that they had been charged different prices for the same DVDs. Inferring that the price discrimination was based on their past purchasing behavior, customers complained vociferously in chat rooms and on Internet bulletin boards. The snowballing controversy led to Amazon’s retraction of the practice.10
As this experience demonstrates, consumers are resistant to DP that is based on their past individual behavior. They also resist dynamic prices derived from their individual capacity to pay. However, customers are much more accepting of DP mechanisms where they are involved in the pricing process. DP that uses price discovery mechanisms such as auctions and group buys always has a high degree of acceptance from buyers; their participation represents an acceptance of the practice. In posted price mechanisms, conversely, extra care needs to be exercised.
Second, an astute use of the latitude of price acceptance within a product category makes it more likely that consumers will accept DP. As previously discussed, my primary research shows that as long as posted prices stay within an LPA, consumers’ purchase intentions are not affected, nor do they consider DP unfair compared to a fixed-pricing regime. Fear of customer backlash has been an important factor holding back companies from implementing DP. A corollary is that involving customers in the pricing process makes them more accepting of a price format that clearly discriminates.
Third, despite increasing price transparency brought about by the Internet and the rapid spread of search bots, customers remain willing to pay different prices for the same product for several reasons: prior experience in the category, their personal tastes, situational exigencies and different levels of price consciousness. Individual customers show different reactions to prices of the same product in different situations, channels and occasions of use. For example, research on bookstore prices suggests that price dispersion is greater among online stores than among offline stores.11 Between 1997 and 1999, market leaders Amazon and BarnesandNoble.com raised online book prices by 8% and 7%, respectively, while discount competitor BooksaMillion.com lowered prices by 30%. Even with the proliferation of shopping bots, Amazon’s market share increased from 64% to 72% and BarnesandNoble.com’s share from 12% to 15%. Such “stickiness” augurs well for DP.12
Finally, for retailers implementing DP, there are logistical issues of changing prices on the price tags, apart from integrating the supply and demand information. Retailers that have implemented DP usually have replaced paper tags with electronic tags. Once the mechanisms are in place, large store chains can save up to $100,000 per store per year that they previously spent on physically implementing price change announcements.13
Price to the Beat of Your Customers’ Shopping Habits
Successful implementation of dynamic pricing requires that the seller manage the context in which product demand varies. The more the seller understands the buying cycles and habits of the customer, the more he or she is able to manage price margins to the rhythm of the customer’s shopping, to segment customers and to develop price discrimination. Customer participation in the pricing process decreases the chances of a consumer backlash. Customers also tend to embrace DP where the price reflects intensity of demand for the product, there is communication between the seller and the consumer, and the price difference is explained by a difference in perceived value across channels through which the transaction occurred. A company that masters dynamic pricing has a potential new source of competitive advantage.
1. S. McNealy, “Welcome to the Bazaar,” Harvard Business Review 79 (March 2001): 18-19.
2. R. Kuttner, “The Net: A Market Too Perfect for Profits,” BusinessWeek, May 11, 1998, 20.
3. M.J. Ashworth, “Revenue Management Builds Higher Profits,” Electric Light & Power 75 (November 1997): 4; and W. Zhao and Y.S. Zheng, “Optimal Dynamic Pricing for Perishable Assets with Nonhomogeneous Demand,” Management Science 46, no. 3 (March 2000): 375-388.
4. W. Baker, M. Marn and C. Zawada, “Price Smarter on the Net,” Harvard Business Review 79 (February 2001): 122-127.
5. A. Sahay, “Consumer Reactions to Dynamic Pricing,” working paper, Indian Institute of Management, Ahmedabad, India, 2005.
6. M. Maynard, “Will This Idea Fly? Charge Some Travelers $10 for Showing Up,” New York Times, August 25, 2004.
7. D. Welch, “Ford Tames the Rebate Monster,” BusinessWeek, May 5, 2003, 5-6.
8. I. Merson, “Reverse Auctions: An Overview,” Acquisition Directions Advisory, July 2000.
9. K. Belson, “Digital Dealmakers Meet in the Middle,” New York Times, Sept. 11, 2003, sec. E, p 1.
10. “Amazon.com Varies Prices of Identical Items For Test,” Wall Street Journal, Sept. 7, 2000, sec. B, p. 19.
11. E. Brynjolfsson and M.D. Smith, “Frictionless Commerce? A Comparison of Internet and Conventional Retailers,” Management Science 46, no. 4 (April 2000): 563-585.
12. M. Marn, C. Zawada, D. Swinford and W. Baker, “Internet Pricing: A Creator of Value — Not a Destroyer,” McKinsey Marketing Practice, September 2000, 2.
13. D. Levy, M. Bergen, S. Dutta and R. Venable, “The Magnitude of Menu Costs: Direct Evidence From Large U.S. Supermarket Chains,” The Quarterly Journal of Economics 112, no. 3 (August 1997): 791-825.