We use cookies to provide you with a better experience. If you continue to use this site, we'll assume you're happy with this. Alternatively, click here to find out how to manage these cookies

80,259 News Articles

# A Fly in the Ointment: Pricing Algorithms Run Amuck

I'm starting a large research project on pricing. So naturally Ron Kleinman's April 24th post on the ARTS's (Association for Retail Technology Standards) Linkedin Group, "When pricing algorithms collide...it isn't pretty", caught my eye. It reminded me of the value of business rules in pricing, the common sense these rules should instantiate, and the danger of running algorithmic price optimization in a "lights out" fashion.

In this instance, the lack of business rules and a black box approach lead to a new copy of a book on Amazon being priced at \$1.7 million, plus \$3.9 shipping, when used copies were available from \$35.52.

?The \$1,730,045.91 (+\$3.99 shipping) Book

Ron's post referenced a post by an evolutionary biologist named Michael Eisen. He needed a copy of a classic but out-of-print work in developmental biology, The Making of a Fly. He found 17 copies available on Amazon; 15 used from \$35.54 and two new from \$1,730,045.91 (+\$3.99 shipping). The good doctor did some scientific snooping-time series observations and simple modeling, and found two warring algorithms, one from each of the two booksellers selling the new copies. One vendor's algorithm set its price at 99.8% of the second seller's price. The second seller set its price at 127.1% of the first's price. Each vendor scraped its competitor's site daily, evidently for a very long time, retrieved that day's price, and let its algorithm fly. Eventually someone at the first seller took notice and reset its price to \$106.23. The next day the other seller's price was \$134.97, exactly 27.1% higher.

What Happened

Dr Eisen reasoned that the second seller didn't own a copy of the book and planned to fulfill an order for the book by buying it from its competitor, a transaction that seller could absorb in its 27.1% gross margin -- a reasonable hypothesis we'll take as read.

We can draw additional conclusions from Dr Eisen's snooping and his hypothesis:

• Neither vendor's algorithm was very sophisticated, each operating with a single business rule -- one pricing at a discount against the competition, the other pricing at a premium. Neither rule understood product substitutions, here buying a used copy when the price premium of a new copy gets too high, as a consumer's response to pricing.
• The second vendor felt it didn't need to compete on pricing and was defending its premium price position based on a better reputation evidenced by having a higher buyer rating than the first vendor.
• Both pricing rules implemented a price image strategy -- an "under" pricing strategy at a narrow discount and a premium pricing strategy at a wide margin.
• Both pricing strategies reflected the company's inventory management and fulfillment strategy -- one holding inventory and absorbing financial risk, the other using drop-ship fulfillment and absorbing a brand risk (either a shipping label likely not carrying its brand or a longer, two-step fulfillment cycle).

Implications and Lessons to be Learned

The extreme -- even absurd consequences of the price war waged by the two booksellers and the simplicity of pricing strategies behind the absurdity of it all make this example a good object lesson in pricing. I can easily think of four points to keep in mind. They're calls to action to assess whether your pricing technology needs a refresh or as key early steps in any such refresh:

• First and foremost, understand your own pricing rules -- what you think they are, what they in fact are, and the business tactics and strategies driving them and being driven by them. From my experience, the implementation of pricing technologies, those employing algorithms of any sort and especially optimization one, can help an organization discover its inherent pricing strategies and their interplay with business strategies.
• Second, understand your competitors' pricing strategies and tactics. Once you understand how your own pricing rules interact to drive your pricing actions, ferret out the inherent business rules driving theirs.
• Third, with your own as-is pricing rules and your competitors (as best you can infer them) in view, lay out what you think your pricing rules should be, from the operational and tactical to strategic, and simulate or stress test them under various competitive and vendor strategy scenarios.
• Finally, look at optimization technologies once you've worked through the first two bullet points and concurrently with the third. Don't jump to optimization any sooner. If you do, you stand the risk of optimizing for infeasible solutions or absurd outcomes.

Let Me Finish

In reality rules governing pricing actions are a complex. They include operational rules (e.g., round ending numbers to \$0.99), tactical rules (e.g., private label product prices set at 95 percent of the national brand price or collapse markdowns to one of three percent-off marks), and strategic rules (e.g., create a value price image for known value items in a destination category). All in it wouldn't be uncommon for a double-digit number of rules to interplay in governing constraints on any single item's price.

Moreover, pricing rules instantiate an important, and in some cases the most important, part of a retailer's business strategy. They govern the flywheel power of price to drive sales, earn margin, maintain differentiation, burnish a brand image, and increase customer loyalty.

Pricing rule maintenance is a complex task. Rules need to be maintained in easily constructed natural language statements with logical operators, valid field names, valid specified field values, and restrictions which reference any dimension of a category or assortment management planning cube. Absent tools which afford these types of specifications, rule maintenance is onerous and complicated, the implications of the interplay among rules aren't easily understood, and rules maintenance is unlikely to actualize intended business strategies meant to be carried by pricing rules.

There's no excuse for laboring with onerous pricing maintenance or carrying the handicap of pricing rules which don't execute business strategies or tactics. Leading application vendors offer these capabilities in their pricing systems -- obviously with differing capabilities to meet different levels of sophistication.

Android M Developer Preview announced at Google I/O: Android M UK release date and new features. Wh?......

Why I think the Apple Watch sucks and you'd be mad to buy it

Ben & Holly's Game of Thrones titles spoof is delightfully silly

Mac OS X 10.11 release date rumours: all the new features expected in Yosemite successor