A few hours after publishing yesterday’s post I stumbled upon two interesting articles that are relevant to Uber and the pitfalls of valuing an algorithm above everything.
First, this opinion piece in the New York Daily News on the Uber Sydney’s surge pricing during the hostage crisis and when so-called “market forces” cross the line into price gouging. Mr. Louis made an important distinction: what happened in Sydney was price gouging, not market forces driven by a neutral algorithm. It’s not about what the market will bear and using the higher prices to entice more drivers to the area. It’s about taking advantage of an emergency situation and the difficult situation riders are in to charge a higher price.
Uber has already gotten into pricing trouble in New York when during Hurricane Sandy surge pricing was eight times the normal rate. That motivated City Councilman David Greenfield to introduce a bill to prohibit surge pricing of more than 100%. He is quoted in the piece as saying: “People can make twice as much money, so if it’s raining or snowing or New Year’s Eve, you can have more cars on the road. I think that’s fair. But to actually charge people 900% more, I think everybody would agree — that’s price gouging.”
It is because of such potential regulation and the fact that New York law prohibits price gouging that Uber changed its pricing algorithm, but only in New York. It now bases surge pricing during emergencies on the average fare charged in the prior two nonemergency months.
All it needs is to do now is to decide what constitutes an emergency.
Second, I came across this troubling use of an algorithm to tackle an expensive problem for retailers: fraudulent returns. These include “wardrobing” which is returning an expensive item of clothing after it is worn once for a special event. Currently retailers combat wardrobing by placing tags in “inconvenient” locations and enforcing strict return policies. Another example is the purchase of high-end electronics for use during a special event, such as a returning a TV after the Super Bowl. American retailers claim that 1% of shoppers are repeat offenders and that costs add up: $16.3 billion in 2013.
So, how will an algorithm combat fraudulent returns? By tracking purchases and returns across retailers (a gross invasion of privacy, in my opinion) and analyzing them to find patterns. After defining a behavior pattern for serial returners, that pattern will be matched to individual shoppers and the algorithm will notify sales people to refuse specific returns. In the article, Tom Rittman of The Retail Equation, the company that created the system, points out that “a refused return is therefore an objective result made by a computer, allowing even the most subjective cashier to be removed from the decision making process.” Not only is he putting all trust in the algorithm, he is adding that it will specifically block human intervention. Sure, that will work when the algorithm works as expected, but when it mistakenly tags the best customers as returnaholics and they take their business elsewhere, retailers might be less than pleased with it.
As an aside, Rent The Runway, a company that rents out clothing and accessories for very short periods, noticed the “wardrobing” problem and found a completely different solution. Instead of penalizing shoppers, it gave them what they wanted: the use of an expensive dress for a single event at a fraction of the cost of purchasing it. That’s a much more creative and user-friendly solution than the one created by The Retail Equation.