 Frequently Asked Questions
Why should I use Cato Consulting Group? If you ask 10 analysts about modeling, you will likely get 10 different approaches. We believe that our proven track record of implementing quantitative and related management techniques makes us unique in the industry. We will ensure that we are there with you for every step of the process. This includes negotiating with list owners for the test data, developing the model, verifying that the scored universe is identical to the test universe, and ensuring that the scoring is applied to the universe that contains the correct customer history. And, finally, we will be there when you evaluate the roll-out results and compare them to the projections. How much will my mailings improve? For models that utilize the list mailer’s purchase data, plus promotion history and our unique “deliverability” indicators, it is reasonable to expect a 25% improvement for three-quarters of the mailing universe; 50% improvement for half of the universe and 75% improvement for the top 25% of the universe. How long does it take to prepare a model? From the time the test data is available, a model can be completed in about 3 weeks. How long will the models hold up? Technically, for a long time. Practically, each model should be renewed yearly, to ensure that it compensates for changes in the target list (such as what happens when the mailer changes product offers or if the mix of sources for new customers changes.) Successful users revise models more frequently because the results warrant updates. Is there a way to distinguish “good” from “bad” internet orders? If “good” means profitable and “bad” means unprofitable, definitely. For example, for a major continuity marketer, we developed on-line and real-time models that determine which offers to extend to new orders. How do I know where to find my best customers? In profitable lists, modeling will allow you to find names that you should not be mailing and in unprofitable lists, you will find names that you can mail. A major difference with modeling is that you can use much more data to discriminate among prospects and increase both response and conversion/back-end performance. What is the value of each customer? The model will determine the expected profitability of each name in the entire list. It is important, therefore, to project a realistic future value calculation in the modeling process, so as to ensure prudent cut-off decisions. How should marginal customers be treated? With sound modeling, you will be able to identify significant quantities of marginal names who can be promoted profitably. However, you should probably mail them less frequently than you mail your best names. It is a good practice to mail these names with your best offer at the best time of the year, and no more. What data is pertinent? In modeling, data is the ultimate leverage. The more there is the stronger the model will be. When testing, obtain all the data that the list owner will release. We have successfully negotiated with most list owners so that they provide their entire history files, including as many as 400 data sets. Models to these lists provide the strongest gains. What benefit does regression offer with outside lists? Modeling will allow you to promote significant chunks of large outside lists in situations where you could only promote smaller segments, manually selected. For example, our initial tests are usually mailed to universes of names with purchase activity in the last 12 months and we target, where practical, universes of one million names. If you had a good mailing experience with segments of that list previously, you can expect to mail from 50-60% of that list, if the models had access to all of the customer information. |