Posted by: zyxo | February 12, 2010

Data Mining Models ? 10 reasons for not using them !


  1. Modeling is difficult :
    Wrong ! You are a marketer, so you do not have to do the modeling yourself.
    Find or hire a data-loving whizzkid with a solid analysis background. He/she will do the hard work for you
  2. Models are to expensive :
    Wrong ! Wrong for two reasons :
    1. The tools are actually free (download R or Weka from the internet)
    2. The model pays itself if you use it in one or multiple campaigns
  3. Models are black boxes :
    Right ! But that doesn’t mean they are not useful !
    Do you know exactly how your PC works ? Your TV set ? Your car ?
  4. I am not sure the models will work :
    Wrong ! Just ask the whizzkid to test the model with real-life data, that he did not use for modelling. He can make you a nice chart, showing that the top selected customers do much better than the others.
  5. It is impossible to explain the models to the sales people :
    Right ! Because the models are black boxes.
    Wong ! Ask your whizzkid to give you the three most significant variables of the model and communicate these to your sales people.
    (Not so black-box after all !)
  6. selections based on the model are to heterogeneous / targets do not match our campaign materials (banners, ads …) :
    Wrong ! OK, sometimes heterogeneous perhaps, but who said you have to communicate to all of them in the same way ? All you have to do is segment the selection with a communication goal in mind (talk to your whizzkid)
  7. the selection will become to complicated :
    Wrong ! The selection becomes simpler ! Only use the model score as the selection criterion (besides other “administrative criteria”).
  8. I did it for years without data mining models. Why change ? :
    So Sad ! You need to change, otherwise the competition will leapfrog you !
  9. The expected returns based on the model data are way to low. I want to get more out of my campaign! : Unrealistic ! Ask your whizzkid to compare the expected returns from the model with those from your own selection criteria. Then let the data decide !
    A decent data mining model allways outperforms man-made selection rules.
  10. My job is to spend the campaign budget. I do not care about the results (nobody else does anyway) :
    So sad ! You make life easy for your competitors !

You can also view this list as a free to download presentation

Reblog this post [with Zemanta]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Categories

%d bloggers like this: