Lead generation introduction

Let’s take a first example. TIMi framework is currently used on a day-to-day basis by the Benelux Business & Decision company to create “Propensity to buy” models for their MMS service (Managed Marketing Service). Business & Decision creates (with TIMi) “Propensity to buy” models for famous companies like: Audi, DDB, Renault, etc. A more complete list is available here: http://www.businessdecision.be/2277-about-us.htm.

Before selecting TIMi as their one and only datamining tool, Business&Decision made an extensive review and an extensive benchmarking of many other very well known datamining tools. The benchmark was the following: each software vendor receives an anonymized (incomplete) extraction out of the dataset used for the MMS service and they must create a predictive model as accurate as possible (the lift on the TEST set must be as high as possible) in a few days. The results of the benchmark were the following:

• Analytical Tool 1 (AT1): The Analytical tool AT1 is a well known commercial and very expensive statistical suite, supossedly leader in its field. After 2 days of hard work the internal team of consultants from AT1 finally manages to import the text-file containing the dataset into AT1. Many difficulties (due to the large size of the dataset: 3000 columns) were encountered. AT1 was not able to analyze directly the dataset (once again because of the large size of the dataset): some “manual” and approximate decisions had to be made to reduce manually (based on the “intuition” of the statistician) the size of the dataset. After one additional workday, the dataset was “small” enough to be injected into the AT1 analytical engine. After 42 hours of computing-time, AT1 delivered a predictive model with ZERO accuracy. The team of consultants from AT1 then came back to see the Business&Decision company to ask if they were “making stupid joke with a dataset containing only random numbers?“
• Analytical Tool 1 (AT2): The Analytical tool AT2 is a well known commercial, very expensive and automated datamining suite. The AT2 software managed to create a normal-accuracy predictive model (lift=40% at 10%) after 3 hours of computing time. Some short manual manipulations were required to reduce the dataset set size before the analysis (sampling) (this manual procedure took less than 1 additionnal hour).
• TIMi: The TIMi software reliably created a highly accurate predictive model (lift=50% at 10%) within 4 minutes of computing time. The workload for the analyst was short and easy: a few mouse clicks and that’s all!

These benchmarking results were informally revealed by the head of the Business &Decision Benelux predictive datamining team.

These benchmarks once again demonstrates the superiority of the TIMi predictive engine on “real-live” industry-level-datasets. Our outstanding results at various world-level datamining competition also demonstrate the outstanding accuracy of the TIMi analytical engine.