I’m a big fan of automating pricing tasks. Companies spend way too much time on mundane issues like moving pricing information from one system (or spreadsheet) to another, and far too little time really thinking strategically about price and value. A few simple automation steps can free up time and money for those strategic activities, which are not only more valuable, they’re much more fun, too.
However, when you automate, you want to know what your trusty little helpers are doing. When you hand over your car keys to your teenager, you know that eventually, your teen will come home for gas, food, money, or sleep. When you hand over your pricing to your automated apprentice, you are giving your car keys to someone who never gets tired, runs out of gas, or needs food, and who may not know how to drive once they get out of the neighborhood.
Recently, just such a setup produced a book that listed for over $20 million on Amazon, as 2 automated pricing bots kept raising prices until finally somebody noticed. (Thanks to the many alert readers who sent me a link.)
While amusing, this situation didn’t really cause anyone any harm. (I’m going out on a limb here and assuming no one actually bought the book for $23M.) However, automated systems can cause real havoc, especially when they go outside the “reasonable” bounds for which they were designed. You can then up with heavily discounted items that the seller didn’t intend to discount. In one case, a manufacturer’s pricing system returned a price of -$1 to mean “no price found.” The idea was a good one. Put in something so blatantly wrong that someone would have to investigate it and correct it. The execution was flawed, however, as the approval process team (and code) was not aware of this potential situation. It happened once on a huge quote, so the people glancing at the totals never noticed. But the company was quite upset when the customer insisted that they honor the price instead of the $100,000+ it should have cost.