Who’s Your Scapegoat?

I find it interesting that machines and procedures often become scapegoats for “human” errors. Remember the time when the word “mainframe” was a dirty word? As if those pieces of hardware were contaminated by some failure-inducing agents. Yeah, sure. All your worries will disappear along with those darn mainframes. Or did they?

I find it interesting that machines and procedures often become scapegoats for “human” errors. Remember the time when the word “mainframe” was a dirty word? As if those pieces of hardware were contaminated by some failure-inducing agents. Yeah, sure. All your worries will disappear along with those darn mainframes. Or did they? I don’t know what specific hardware is running behind those intangible “clouds” nowadays, but in the age when anyone can run any operating system on any type of hardware, the fact that such distinctions made so much mayhem in organizations is just ridiculous. I mean really, when most of computing and storage are taken care of in the big cloud, how is the screen that you’re looking at any different than a dummy terminal from the old days? Well, of course they are in (or near) retina display now, but I mean conceptually. The machines were just doing the work that they were designed to do. Someone started blaming the hardware for their own shortcomings, and soon, another dirty word was created.

In some circles of marketers, you don’t want to utter “CRM” either. I wasn’t a big fan of that word even when it was indeed popular. For a while, everything was CRM this or CRM that. Companies spent seven-figure sums on some automated CRM solution packages, or hired a whole bunch of specialists whose titles included the word CRM. Evidently, not every company broke even on that investment, and the very concept “CRM” became the scapegoat in many places. When the procedure itself is the bad guy, I guess fewer heads will roll—unless, of course, one’s title includes that dirty word. But really, how is that “Customer Relationship Management” could be all that bad? Delivering the right products and offers to the right person through the right channel can’t be that wrong, can it? Isn’t that the whole premise of one-to-one marketing, after all?

Now, if someone overinvested on some it-can-walk-on-the-water automated system, or just poorly managed the whole thing, let’s get the record straight. Someone just messed it all up. But the concept of taking care of customers with data-based marketing and sales programs was never the problem. If an unqualified driver creates a major car accident, is that the car’s fault? It would be easier to blame the internal combustion engine for human errors, but it just ain’t fair. Fair or not, however, over-investment or blind investment on anything will inevitably call for a scapegoat. If not now, in the near future. My prediction? The next scapegoat will be “Big Data” if that concept doesn’t create steady revenue streams for investors soon. But more on that later.

I’ve seen some folks who think “analytics” is bad, too. That one is tricky, as the word “analytics” doesn’t mean just one thing. It could be about knowing what is going on around us (like having a dashboard in a car). Or it could be about describing the target (where are the customers and what do they look like?). Or it could be about predicting the future (who is going to buy what and where?). So, when I hear that “analytics” didn’t work out for them, I am immediately thinking someone screwed things up dearly after overspending on that thing called “analytics,” and then started blaming everything else but themselves. But come on, if you bought a $30,000 grand piano for your kids to play chopsticks on it, is that the piano’s fault?

In the field of predictive analytics for marketing, the main goals come down to these two:

  1. To whom should you be talking, and
  2. If you decided to talk to someone, what are you going to offer? (Please don’t tell me “the same thing for everyone”.)

And that’s really it. Sure, we can talk about products and channels too, but those are all part of No. 2.

No. 1 is relatively simple. Let’s say you have an opportunity to talk to 1 million people, and let’s assume it will cost about $1 to talk to each of them. Now, if you can figure out who is more likely to respond to your offer “before” you start talking to them, you can obviously save a lot of money. Even with a rudimentary model with some clunky data, we can safely cut that list down to 1/10 without giving up much opportunity and save you $900,000. Even if your cost is a fraction of that figure, there still is a thing called “opportunity cost,” and you really don’t want to annoy people by over-communicating (as in “You’re spamming me!”). This has been the No. 1 reason why marketers have been employing predictive models, going back to the punchcard age of the ’60s. Of course, there have been carpet-bombers like AOL, but we can agree that such a practice calls for a really deep pocket.

No. 2 gets more interesting. In the age of ubiquitous data and communication channels, it must become the center of attention. Analytics are no longer about marketers deciding to whom to talk, as marketers are no longer the sole dictators of the communication. Now that it is driven by the person behind the screen in real-time, marketers don’t even get to decide whether they should talk to them or not. Yes, in traditional direct marketing or email channels, “selection” may still matter, but the age of “marketers ranking the list of prospects” is being rapidly replaced by “marketers having to match the right product and offer to the person behind the screen in real-time.” If someone is giving you about half a second for you to respond, then you’d best find the most suitable offer in that time, too. It’s all about the buyers now, not the marketers or the channels. And analytics drive such personalization. Without the analytics, everyone who lands on some website or passes by some screen will get the same offer. That is so “1984,” isn’t it?

Furthermore, the analytics that truly drive personalization at this level are not some simple segmentation techniques either. By design, segmentation techniques put millions of people in the same bucket, if a few commonalities are found among them. And such common variables could be as basic as age, income, region and number of children—hardly the whole picture of a person. The trouble with that type of simplistic approach is also very simple: Nobody is one-dimensional. Just because a few million other people in the same segment to which I happen to be assigned are more “likely” to be into outdoor sports, should I be getting camping equipment offers whenever I go to ESPN.com? No siree. Someone can be a green product user, avid golfer, gun owner, children’s product buyer, foreign traveler, frequent family restaurant visitor and conservative investor, all at the same time. And no, he may not even have multiple personalities; and no, don’t label him with this “one” segment name, no matter how cute that name may be.

To deal with this reality, marketers must embrace analytics even more. Yes, we can estimate the likelihood measures of all these human characteristics, and start customizing our products and offers accordingly. Once complex data variables are summarized into the form of “personas” based on model scores, one doesn’t have to be a math genius to know this particular guy would appreciate the discount offer for cruise tickets more than a 10 percent-off coupon for home theater systems.

Often people are afraid of the unknowns. But that’s OK. We all watch TV without really understanding how HD quality pictures show up on it. Let’s embrace the analytics that way, too. Let’s not worry about all the complex techniques and mystiques behind it. Making it easy for the users should be the job of analysts and data scientists, anyway. The only thing that the technical folks would want from the marketers is asking the right questions. That still is the human element in all this, and no one can provide a right answer to a wrong question. Then again, is that how analytics became a dirty word?