>> What We Do: Predictive Modeling
Predictive Modeling Services:
Over the past 6 years alone, we have produced over 200 predictive models. Now on
our fifth generation of predictive modeling technology, we are able to produce models
that are superior in terms of accuracy and speed of development, as well as up to
50% lower in cost than most other firms. Just like other aspects of direct marketing,
predictive models can be tested against one another. Datalab has a flawless record
in testing against other models.
Increased Model Accuracy
Our latest predictive modeling methodology, Generation 5 technology, offers higher
accuracy than other methodologies such as RFM, linear or logistical regression,
neural or genetic networks and simple decision trees. At the same time, our models
remain remarkably robust - even in the presence of common data problems such as
dirty data. They are also resistant to over-training and approximately 100 times
faster than a neural net.
Faster Model Development
Our predictive models are normally developed in 24 – 48 hours once data preparation
is completed. This speed of development is partially because of our proprietary
automated software and also because our implementation of the scoring algorithm
in is SQL – the native environment for most of our marketing databases...No time
is lost in transferring, converting or deconstructing data.
Lower Modeling Prices
Because our models are faster to develop and less labor intensive, the cost to our
clients is reduced.
Types of Predictive Models:
We have experience in the following:
Response Models
Which customers / prospects are most likely to respond to our offer? What are their
component characteristics (and combinations) that most influence their response?
How can we optimize our targeting strategy to produce the greatest response?
Profitability Models
Most savvy marketers prefer their models to be based on profitability, not just
response. Based on the customer / prospect data, all of its component variables,
our offer & pricing and the targeting strategy chosen, how much actual profit
will the company make when we market to certain people? How deep into the marketing
universe can we go while still achieving a break-even ROI? What is the predicted
profitability per piece mailed / per prospect reached? How can we optimize all this?
Risk/Delinquency Models
How will our prospects behave once they respond and become customers? What are the
component characteristics that predict payment delinquency or risk? What prospects
should not be solicited due to delinquency likelihood? What are the observable “cues”
that can predict likely delinquency or risk?
Retention Models
How long will they stay customers with the company? What characteristics make them
attrite early? Are there observable cues that can help predict upcoming attrition?
Multi-Product Optimization/Cross-Sell and Up-Sell
For multi-product companies, choosing which products to market to which customers
and in what order can mean difficult questions. Which product should be offered
to which customer to make the most money? What is the best combination of products
and customers for profit as well as other goals such as market share and long term
customer persistency? How does achieving these goals conflict with budgets and product
eligibility rules?
>> Read Less