Home Depot Privacy Pratfall: Spotting Web Shoppers In-Store
Written by Evan SchumanHome Depot has been using a CRM practice that uses payment-card numbers to match in-store customers with their online purchases. It’s a move that, although likely passable for PCI, is rather unnerving to privacy advocates.
Home Depot officials stress that they only use the technique with shoppers who opt in, an argument that is somewhat tempered by how often consumers don’t even notice privacy opt-in and opt-out Web site declarations. The chain has been using this technique for various purposes, including E-mailing in-store customers to ask them to review their recent purchases.
Home Depot’s use of the card-matching procedure is not that unusual among major chains, but the norm is for the effort to be kept internal, to help improve general marketing. It was Home Depot’s reaching out to customers that made some of them realize what was going on. And therein lies the problem.
One such shopper reached out to The Consumerist, which published his comments. That shopper said he had only visited HomeDepot.com as a guest and then made the in-store purchase of a drill months later.
“I can only conclude that Home Depot held my credit card information in one of their internal systems without my authorization, linked it to the e-mail address I used on the one-time purchase without my authorization, and is now watching what I buy in their store using the card so that I can build up their rating database and presumably so they can target their advertising to me,” the story quoted the anonymous shopper as saying.
To be precise, Home Depot does not, of course, store payment-card data. It converts that data to a character string and then looks for a match with that string when it processes a card in-store.
Home Depot wouldn’t say exactly how it converts the numbers, but there are only a few likely ways. PCI Columnist—and QSA Extraordinaire—Walter Conway said Home Depot could be using strong encryption or tokenization, or it might simply be comparing a non-reversible one-way hash. All approaches, he said, would enable the chain to track shoppers across channels. And, depending on the implementation, that is perfectly acceptable for PCI purposes.
Steve Sommers, a senior VP for applications development at security vendor Shift4, said his guess is that this is likely being done with a multi-use token, something Sommers argues is less than ideal.
Multi-use tokens came into being “when PCI SSC bastardized the tokenization definition and allowed for tokens to be mathematically derived from the PAN, either via encryption or some hash algorithm. Now with this method, the same card number would normally generate the same token. This would greatly reduce the effectiveness of the tokens to protect the data, but would give the merchant the tracking ability described. This is probably the method the merchant is relying on. While Shift4 adopted multi-use tokens, we do not believe that a token mathematically based on the PAN is secure, because it could be susceptible to rainbow attacks if the hash salt or key is ever compromised.”
January 17th, 2013 at 12:53 pm
One slight clarification:- Multi-use tokens and what I would call repeatable tokens are two different aspects. Multi-use token simply means that a token can be used multiple times for multiple transactions, like card-on-file or express check-out. Multi-use tokes, provided they are not mathematically derived from PAN are very secure. On the other hand, repeatable token or a token mathematically derived from the PAN (hash or encryption for example), are not nearly as secure as their non-mathematically derived counterpart and if improperly implemented, can actually be fairly insecure. Multi-use and repeatable represent different aspects of tokenization.
January 18th, 2013 at 2:59 pm
If companies want to offer customized marketing while navigating around a backlash they need to understand where the lines is drawn for the consumer. There is a clear difference between learning about your customer and stalking them.
The fact that they are linking data gathered at both their online store and real-world locations is also no surprise. They already run their brick and mortar locations like they were an e-commerce site. Try checking out at Home Depot and you will likely find that there are somewhere between 1 and zero humans to scan your purchases. Retailers: Make this rule # 1: If you want my money, be willing to put out a human hand to take it. Rule # 2: spend less time trying to understand what I MIGHT want to buy and more time helping me find what I DO want to buy. If I walk into Home Depot, I probably already know what I came for. If I need your help, I’ll ask. Yet if you’ve ever tried to find a sales person at one of their stores, you already know that it is an exercise in frustration. There will be somewhere between 1 and zero humans there to help you and good luck finding them.
January 18th, 2013 at 5:34 pm
People who opt in sometimes aren’t aware of it. Not everyone reads and scans everything in front of them. Sad, but that’s how many businesses operate.
January 22nd, 2013 at 12:01 pm
My own experience is that people are much less concerned about this type of privacy concern than they once were. More likely people expect that a store can find their old purchases, and like the extra catering to their desires and needs. Shoppers like it when they can come into the store and want a new line feeder for their weed whacker. And if they don’t remember which one they need, the associate can look it up. Shoppers seems to expect that level of service. Perhaps THD should examine the use of guest shopping accounts for the same purpose though.
January 22nd, 2013 at 2:43 pm
Apple does this as well. I bought an iPad at an Apple Store over the holidays using the payment card that I have on file with them and by the time I got home I had an email thanking me for my purchase and describing how to attach it to my existing Apple ID. I did not provide my email address or name or anything. Just the card. No idea if this is covered in the iTunes TOS or not.