Category Archives: Data Analytics

Data Analytics – The Myopia of Estimates – Part 2

Curiosity Wasn’t Priced In

Curiosity is a good thing. It’s vital when it comes to data. It uncovers the “why” behind the “what”. Failing to price in curiosity can be the difference between “servicable” and “wow”. Good analysts will follow their curiosity anyway.

Timelines Allowing No Mistakes or Delays

Someone will screw up and code will need fixing. Someone will come to work with the flu and infect 3 people. Machines will break. Approvals will be delayed. People will change their minds. Optimism isn’t your friend when it comes to estimating.

Death by Meeting

How long will it take to code this? About 8 hours.
Estimate: 8 hours
8 hours of coding plus 2 hours of meetings = 10 hours
Boom! You’re 25% over on the job.

The Myopia of Estimates Part 1

Data Analytics – Define Channel

“A third of our sales are coming in through the web channel – let’s move budget from direct mail into display ads!”

Whoa – display ads are in the digital channel – but is display actually driving those purchases? Would more display ads drive in more purchases?

What do you mean by channel exactly? What if I respond to a direct mail piece by calling your Inbound Sales Center, then go onto your website to make a purchase? Which channel do you attribute me to – Direct Mail, Inbound Call or Web?

The answer is that every response has two at least two types of channel attribution.

One is the marketing stimulus channel – in this example, Direct Mail. The other is the response channel – in this example, Inbound Call. In many cases, a third channel is the purchase channel – in this example, it’s Web.

Moving money from Direct Mail to Display might be the right move – or it might cut off the main pipeline into your purchase funnel. I don’t want to make your head explode – but there may be a combination of market stimuli that constitute the actual Market Channel. It’s another facet of multi-channel attribution.

So you’re not measuring all this precisely? You’re not alone – many firms, even some really big ones you’ve heard of, aren’t doing it all that well either. Getting attribution right is a commitment – time and money – and is an iterative process. It should ultimately answer the question of where to spend your marketing money, gaining more precision with time.

Data Analytics – Why The C-Suite Ignores Data

The data is too math-intensive to hold their interest.

Eyes glaze over. Gazes wander. Executives shift in seats. Expensive pens tap on conference tables. Math does that to people. You remember high school math class when you had to turn word problems into equations? Well, now you have to do that in reverse.

They really believe in their hunches.

C-Level executives are paid a lot of money on the assumption that they know stuff mere mortals don’t. And they got swagger. Who the hell are you to tell them their intuition doesn’t reflect reality? Well, it’s actually your job to tell them, isn’t it? Break the ice by asking if they saw Moneyball. Then slip the name Nate Silver into the conversation at some time in the presentation. You’re doing them a huge favor correcting assumptions that are untrue. But telling the truth isn’t enough – you have to sell the truth. Get your own swagger on.

They get a conflicting story from their own tribe.

Department heads like to say “Don’t bring me problems, bring me solutions”. So, unfavorable data within a department will often be put through the “spin cycle” before they are presented to the C-level person. Or, the data may be buried altogether so as not to make the middle managers look bad. For instance, maybe the IT team only presents the CTO cumulative data on users of the new app – the arrow on the graph is a never-ending march upward. It’s no wonder, then, that the CTO may not embrace Marketing telling her that customers don’t like the new app, and never come back after their first visit. It may be the first time she’s heard it.

How Data-Driven Marketing Contributes to the Class Divide

Once on a calibration call, I heard a customer service rep actually tell a customer who was doggedly asking for a better rate something like “You’re one of our low-segment customers. No matter how many times you ask, none of you guys can get that rate.”

Clearly, that rep was seriously off-script. But the segmentation score on her CRM screen told her that the company didn’t value this customer very much, so she didn’t either. I built the segmentation model that told her that. For me, it brought into stark relief how marketing segmentation can affect dynamics far down the road.

In many years in the field, I’ve seen the good that data-driven marketing can do. It makes online life more relevant. It helps businesses stock products their customers want to buy. But the intrinsic power of data analytics to segment a population can also be wielded to divide it.

The last twenty years in which database marketing has hit the big-time coincides with a period of increasing polarization between rich and poor. I’m not suggesting that segmentation causes this polarization. Rather, it’s that it helps drive the wedge, already in place due to a complex mix of social, economic and political factors, deeper.

The objective of segmentation is to enable businesses to target their marketing capital toward the acquisition and retention of those customers yielding the greatest profit. There isn’t anything wrong with this, per se. Making money is what businesses are supposed to do, and it is the responsibility of their marketing organizations to help make that happen.

Customers and prospects are identified by their potential to enhance the bottom line, and strategies are crafted to reward the more desirable segments for doing business with them and not reward less desirable groups for it (or even subtly discourage them from it). The most profitable customers are not always the wealthiest – but let’s face it, it’s often the way predictive models will tell you to bet.

Predictive and yield models tell builders how to market and build most profitably. A prospect who can only afford a $195K house is courted by no one and can’t find a new house to buy. A prospect who can afford a $950K house is courted by everyone and has plenty of choices.

Profiling will tell businesses which customers are likely to have the wherewithal to pay on time and upgrade to more profitable products. This insight will be incorporated into the firms’ CRM systems. Those segments will receive the best offers, the special concierge customer service phone lines, the waived fees. There might also be “aspirational” or “elite-in-training” groups that get slightly better treatment in hopes that they will start behaving like the elite groups. And the other segments?

It costs them more to do business. They pay more for products. They have to wait in a longer phone queue for customer service. As for the service they do get when the phone is answered, there is no scripting in the CRM system that explicitly says “you don’t have to go the extra mile to treat this customer well”. But it’s pretty much guaranteed that some harried customer service reps will (perhaps in a rush to minimize the Average-Handle-Time metrics they’re bonused on) interpret it that way.

Before analytics, businesses often had policies that every customer should be treated like they’re the best customer – because absent the data, the assumption was that every customer had that potential. But in the data age, there is no more benefit of the doubt. When people complain that customer service doesn’t exist anymore, they’re wrong. It’s still alive and well – it’s just heavily up-market.

To reiterate, marketing segmentation and analytics did not cause the class divide. That has existed for millennia. Let’s at least be aware about how marketing analytics contributes to it in present day.