This morning, while driving into work, I heard an advertisement on the radio from a major bank. The premise of the commercial was that a father instructed his daughter to use his bank account wisely. Either he gave her a credit card, or he gave her his PIN number to his ATM card… or perhaps he was a joint owner on his daughter’s account. The point is, he seemed to be responsible for funding the account and his daughter had access.
The bank was advertising some new service – every time his daughter spent money on the account, he would be notified. Well, he got one of these notifications and it looked fishy (phishy?). At the end of the commercial, he asked his daughter “Why did you spend xx dollars on this particular item?” She replied “Oh, don’t worry, I got a good deal on it and it was totally worth it.” Ha, ha, ha!
Anyhow, the feature the bank is pushing is that you can now receive notifications for every transaction that is made on the account. This is supposed to be a good feature because you are having information brought to you and you can detect anomalous behavior.
But is this such a good feature?
It is in theory. But as I have been writing about sporadically over the past few months, sometimes technology’s benefits are overstated, and I think that this is no different.
I say this because it leads to information overload. If people get so many notifications with information that is always benign, eventually alerts will be ignored. In other words, even though these alerts are being brought to the end user, almost all of them contain data that is not worthwhile looking at. Yes, I bought some groceries. Yup, I paid my insurance. Yeesh, why did I pay money for that Vi@gkra spam last month? And if there’s nothing ever interesting, complacency sets in. It becomes disinteresting because there is nothing interesting.
I see this all the time at work. I get tons of email notifications for all sorts of stuff. I ignore pretty much all of them because I get them every day and most of the time there is nothing special. But once in a while there is something to pay attention to, but because my attention is divided amongst many other tasks, these potential alerts go unheeded (at least by me).
And these alerts from the bank are no different. Instead, what really is needed is anomaly detection. Credit cards have the right idea; some credit cards have security built in where if it detects spending patterns outside of your normal routine, it alerts you to potential fraud. For example, if you live in Seattle and all of a sudden, charges start showing up in Paris, France, they either (a) refuse to process the charges, or (b) give you a phone call asking what the deal is. Unlike notifications for every transaction, these types of messages are interesting because it detects patterns that deviate from the norm and ignore the benign.
I’m not sure whether or not a bank will be able to figure out spending patterns of families and whether or not a purchase from a store represents a legitimate expense. It’s a nice theory, but whether or not it works in practice is something else.
I think that the key to pattern detection is detecting when something becomes abnormal from an established baseline of behavior. Without this, the information overload makes a feature much less useful.