You Know Me Too Well….

I had an interesting experience this morning.  Every morning on my way into the office I stop for breakfast but recently I’ve found a new bagel shop — I’ve been going there for about two weeks.  Most days I order a bagel with butter, but some days I’m in the mood for an egg sandwich.  This morning was one of those mornings so I strode in ready to order my sandwich.  But, to my surprise, as I approached the counter they offered me my regular “everything bagel with butter today sir?”.  “Yes, please”, I replied not getting my egg sandwich.  What happened there set off a fire in my brain.  Why did this happen?  Am I doing this to my customers?

Just as this happened I was coming to the end of a book, The Numerati, a fascinating walk through the world of behavioral modeling through statistical probabilities.  What physics does to model our physical world and the quants to do model the stock market, the numerati do to model our behavioral world.  I had just read about smart shopping carts being designed for use in supermarkets and had found it promising; but something didn’t sit right.  This was it.

Picture this:  you’re walking through the isles of the local grocery store loading up your smart cart.  There’s a screen customized to your behavior based on a swipe of your loyalty card and the computers behind the scenes know everything about your shopping history.  Marketers would have us believe this is a perfect model to deliver enticing offers to our customers and, until today, I would have believed them.  But now the world looks different to me.

To put this in perspective, friend on Facebook shared the following status: “standing behind a guy in line with $640 worth of groceries.  After a pile of coupons he got it down to $24.50!”  You or I wouldn’t clip so many coupons — that takes dedication — but you would click the button to view sale items only.  Can the enticing models of recommended fully priced splurges with high margins win out over the convenience of “show me all the sale items”.  I’m not so sure.

As my wife puts it, “It’s nice to know that when I walk in to a store I can buy something at full price without feeling like a chump.”  Swarovski learned this lesson with their annual Christmas tree ornaments.  Sales for the ornament were down, but they sold all remaining stock at the outlets the following year.  People wanted the ornament, no demand problem, but they didn’t want to pay full price so they were okay with staying one year behind at the sale price.  Imagine what sales would look if you walked into a Swarovski store and the associate told you where and when to get it cheaper.  This is what smart carts do.  Not good.

Another fundamental flaw with this kind of behavioral modeling akin to the Heisenberg Uncertainty Principle which states, in plain language, you cannot observe one variable without affecting another.  This is what happened with my egg sandwich.  The associate identified me as a bagel and butter person after just a few visits and began offering me a bagel and butter.  Convenient, yes.  What I want, no.  Revenue lost and utility lowered, affirmative.

This is what we, as marketers, do to our customers: use history to predict the future.  Accurate models help us live within the expectations we set for ourselves by resetting them for us next time with a margin of error.  Last week’s budget is this week’s budget range.  This isn’t modeling, this is reiteration.  Can the computer know that today I feel like I deserve the extra calories and cost of an egg sandwich?  Probably not, so it won’t recommend it.