One of the top marketing analytics goals for retailers has, for years, been the single view of the customer. That means being able to identify shoppers as they move from desktop to mobile to in-store and when they phone a call center. A mobile device that typically stays with a person at almost all times has been seen by many as the answer—and it is.
But Google has been pushing an alternative theory, which is to trust it blindly. It doesn’t quite phrase it that way, but that’s where it ends up. On Monday (July 31), a major privacy organization, the Electronic Privacy Information Center, filed a legal complaint with the U.S. Federal Trade Commission asking federal officials to look into a Google program that extracts data from payment card records.
Here’s where things get dicey. The idea is that Google, through an undisclosed third party, has access to “70 percent of transactions for credit and debit cards in the United States,” The Washington Post said.
Its story quotes Jerry Dischler, vice president of product management for AdWords, Google's online advertising service, saying, “Through a mathematical property, we can do double-blind matching between their data and our data. Neither gets to the see the encrypted data that the other side brings.”
The approach is apparently based on a 2011 research paper from three MIT scientists, a project funded by Google and Citigroup, Dischler said.
No purchase data revealed?
In an attempt to buttress its privacy argument, Google argues that the data is aggregated and that it reveals nothing about an individual shopper’s purchases.
And that’s where the logic of all of this falls apart. If the data is truly aggregated throughout the process, then it would show nothing. If would say that tens of millions of people looked at ads online and that millions of them made in-store purchases. If aggregated throughout, it couldn’t know whether they were the same people or what kind of a time lag existed or much else useful.
What would make more sense—slightly—is if the data were linked initially, by this third party, and then made anonymous by the time it’s shared with Google.
You know what’s fun about double-blind? It replaces verifiable data with numbers that the end user must take on faith. And what possible incentive would Google have for indicating that its expensive services are more effective than they are? I can’t think of any, he said, trying to keep a straight face.
The truth is mobile is the only viable way to connect the dots between online and in-store activity. Well, not just mobile, but mobile is the link. The eventual answer lies in mobile (to flag when this mobile or online shopper materializes in-store), digital video analytics (Amazon Go’s efforts of using in-store video cameras to interpret what actions people are taking in the aisles is a good example) and then looking at the time stamps from POS data.
Before we get into the how, we need to think about the why. What’s the point of connecting those dots? If you’re willing to assume that Google’s data is accurate, the only possible action is to buy more Google ads (what a coincidence). Looking solely at POS data and matching it against known online/mobile users is good for figuring out online effectiveness, which is nice. In other words, it would theoretically prove that online ads were more effective than they initially appeared. If 5 percent of people clicked on the link and purchased, that’s a 5 percent ROI. But if you later learned that another 20 percent of those clickers bought that item in-store, your ROI is now adjusted to 25 percent.
That, however, is not where the single-view-of-the-customer money truly lies. Why merely observe and note shopper behavior when you have a shot at changing/influencing it? Yet again, this brings us back to mobile.
Mobile as a communication device
Mobile is not merely a geolocation-tracking device, one that uses Wi-Fi triangulation to pinpoint a shopper’s precise location (more or less). It’s also a two-way communication device. Not only will that Android or iOS device flag you when a customer is in your store’s shirt aisle, but it also gives you a way to talk with them.
Consider this scenario: You know that shopper 663492 visited a specific page for a high-priced tailored shirt at 10:10 a.m. today, a Saturday morning. At 12:05 p.m., this same shopper’s phone is detected in your store closest to his home. Your Wi-Fi triangulates that he’s in the shirt aisle. Your video analytics show him examining the same shirt he was looking at on his screen. But when he appears to peer at the price tag, he puts it back and walks away.
What if you could text him right then and there, with an offer of an extra 25 percent off if he purchases in the next 30 minutes. Think it could save a sale?
Scenario 2: What if he took that shirt into the changing room and then put it back? In that case, he probably knew the price before he tried it on, so a discount might not overcome the purchase hesitation. Perhaps your text could offer to overnight a different size for free instead.
Texting the customer is good, but demographics play a role. If the shopper is in her early 20s, a text could be effective. But if the customer is in his 60s, it might be better to text a store associate who can approach the customer with the offer. With mobile and video analytics, you have choices. With POS analysis alone, you have almost none.
Want additional verification beyond mobile? Perhaps that shopper has her phone turned off or chooses to not ride your Wi-Fi. Some stores have toyed with license plate recognition. This works in reverse. The customer is identified at the POS using a payment card. Digital analytics can match the face to the identity on the card.
When that person is tracked leaving the store, external security cameras can record her license plate number. Link established. The next time that license plate is detected by license plate-recognition cameras in the parking lot, your system knows who is coming and which entrance she is likely to use. And license plate tracking can sometimes be better than car tracking, since many consumers will retain a license plate when they buy a new car.
Does much of this raise privacy questions? Absolutely. But this is merely the automation of age-old retail tactics. Did anyone cry privacy when a veteran sales person noticed the car of a frequent shopper driving in? Or recognized her as she walked into the store? Why is that OK, but it’s a sin when done via software?
Connecting the dots between offline and in-store can indeed be done without mobile. The question is: Why would you want to?