Timing, Counting, & Choice. “Most real-world business problems are just some combination of those building blocks jammed together” – Peter Fader
Over at CIO Insight we have this very practical article on Data Mining by Fader. What it’s good for, what it’s not good for. If you have wondered how you might use this tool, especially if you are a Marketer, you should read this article.
I say the article is practical because even though there are many ways to create mathematical models of customer data, if the end result is not something a Marketer can use to actually increase Marketing Productivity, then you really cannot do much with the output. The models have to create leverage of some kind that can be used to take real world action. In other words, a model can be “technically correct” but completely useless to a Marketer.
For example, just because you can identify a segment doesn’t mean it is practical or viable to address that segment with a unique marketing treatment. And just because the segment has unique characteristics doesn’t mean those characteristics create any real marketing opportunity.
Key takeaways for Marketers from this article should be:
1. Too much data tends to mess up a model. This is especially true if you try jamming all kinds of demographic crap into a model that is trying to predict behavior. If you want behavior as an output, use behavioral variables in your models.
2. Data mining is a great classification tool; it is good at telling you why segments are different. But in order for this to be useful, you need actionable segments to begin with. For example, data mining can tell you the demographic differences between people likely to respond versus people not likely to respond – if there is a demographic difference. But you have to know this “likely to respond” element first. While we’re on this topic, the same idea holds true for surveys. If you want the survey output to be actionable, get to known behavioral segments first, then do your surveys of each segment.
Often, people use technical tools for the wrong Marketing reasons. I see this problem coming down the tracks in web analytics, people are getting so wrapped up in the minutia and the automation of testing they are missing out on the basic stuff. Just like the data mining wave got people off track and into the bushes with “collecting all the data so we can mine it”. But it doesn’t matter how much data you have, the tool does what it does and doesn’t do what it doesn’t do.
Check out the article What Data Mining Can and Can’t Do here.
Any thoughts from the Data Miners out there on this?Follow:
5 thoughts on “***** What Data Mining Can and Can’t Do”
Both this and the other piece are very highly simplistic views of this space. AND, I’m agreeing what is fundamentally being said, but there are other considerations. My take is, keep it on the radar. Where it is absolutely the right mechanism for loss of money, it will raise itself (MCI used it extensively to detect fraudulent use of accounts, esp. calling cards).
And it has great marketing potential (actually, more about relationship potential, than marketing potential…but then, who owns relationships is the big fundamental disconnects for most companies). The problem is drawing conclusions from the findings. Data mining’s real potential is to call attention to things for further investigation. It helps to identify real possibilities (for significance of possibilities, see the 5P’s of Design & Development.
But it takes a good head to add more filters to the considerations. For example, in 1996 I was at a conference where a vendor was displaying a 3D rendering of the data and was pointing out the significance of the red area — people who had called the call center excessive times. They proposed that this be further investigated to cut down costs for the company…perhaps these were people you didn’t want to do business with. To the contrary, I would have been IN that data. At that time I had changed residences and had moved my phone number…all arranged weeks in advance. I spent hours getting hung up on and being given random answers to account for why my phone was not in service and/or when I might expect to have service.
As Avinash says, the tool should be 10% of the solution, the other 90% is raw brains. In the MCI scenario, that was the case…the ‘model’ was constantly tweaked by the guy who designed the algorithm — the tweaking was done based on how accurate the results were.
It is this and this alone that is CRITICAL to the use of any metric.
Jim, the article is interesting and I have taken your concern seriously. But I have some reservations against both the articles.
I have written a blog entry to take forward the discussion. We will need to address this issue quite seriously.
Paul and Bhupendra, thanks for the comments. Based on your input, I’m thinking I did not express myself clearly. So I’ll try again, since there is a lot of confusion around data mining in the Marketing world. Rather than run off on a long comment, I will write a new post…
I think the article is more an attempt to push his approach than anything else. The problem with data mining (or rules or probability models or anything really) is that the technology becomes the focus not the problem. Companies need to decide what DECISION they are trying to influence and then figure out what technology (business rules, optimization, analytics, data mining…) will help them do a better job of making that decision.
Author, with Neil Raden, of Smart (Enough) Systems, a book about this.
I’m not following you James. He’s a professor at the Wharton school, what benefit does he get from pushing an approach? On the other hand…you work for Fair Issac, right?
Do you dispute the basic point that data mining is good at classification but not good at probability? Just askin’.