Pretty much every e-commerce site now allows customers to express their opinion about products. Such reviews can certainly be helpful to other customers, but how do firms use the feedback? According to the Wall Street Journal, some firms are now using the reviews to monitor for quality problems (Firms Take Online Reviews to Heart, Jul 29).
L.L. Bean Inc. noticed earlier this year that one of its top-selling products, Supima Cotton Fitted Sheets, was being slammed in online customer reviews.
The company, which pulled the sheets from its website, found that a wrinkle-resistance treatment mistakenly added by a contractor was causing the cotton fabric to unravel. It offered new sheets to the 6,300 customers who had purchased the set and destroyed the rest of the faulty batch.
“Before, it would have taken us months and months to figure out if something was wrong with the product through returns, if we ever would have known at all,” said Steve Fuller, L.L. Bean’s chief marketing officer.
This is an intriguing way of managing quality. It obviously has some comparisons with standard statistical process control. Bean has some idea of what the usual distribution of reviews is. That long history allows them to identify when the process “goes out of control” and the number of negative reviews increases. That is, in any given set of reviews, there will be some nattering nabobs who whine about a product that the vast majority of customers love. The question is whether one can discern when the level of negativism deviates from historical norms.
One could also presumably track the number of submitted as a percent of sales for signs of trouble. To the extent that the desire to write a review is driven by extreme opinions, an uptick in review volume could be a sign of trouble. Note that this method would be particularly helpful when reviews don’t have a specific numeric scale. Bean’s customer rate products on a five-star scale so there is a clear measure of worsening reviews. If there is no explicit scale, somebody (or some algorithm) has to pour over reviews to determine how whether they are positive or negative. Variations in volume may flag when that is worth the effort.
There is, of course, an important difference between this and your usual implementation of process control: These sheets had already reached customers. That raises a question about what to do with those posted reviews after the problem is fixed.
L.L. Bean also said it hasn’t settled on what to do once it fixes a product. If it puts the altered item back online, the negative ratings remain, but if it advertises the revamped product under a different name, that might mislead customers. This dilemma has kept it from putting the revamped sheets back online.
“It’s a challenge we haven’t quite figured out yet,” said Mr. Fuller.