First, the problem is somewhat complex, and you will probably need to read it a few times. But if your real-world problems were simple, you could solve them with SQL or other conventional means. So the complexity is necessary.

Second, the problem does not address MAST's slicing-and-dicing capabilities. For more information on slicing-and-dicing, see our Census examples.

A business manager at a telecom realizes that his competitors have targeted his high revenue people, and he has lost many of his customers. The customers that he has lost fall into two main categories: those who make card calls (using the telecom card) to places other than their own home, and those who do international calling on weekends.

The manager has to do something to stop losing customers, but he wants to go beyond stopping the flow, and turn the problem into an opportunity. Some of his customers actually cost him money, so he would like to shift those over to the competition, let the competition keep those non-profitable customers that it has already acquired, retain the best customers for himself, and regain the lost profitable customers. So he doesn't want to do blanket advertising, but target specific customers. He has to move fast before he loses many more.

He produces the following plan:

1. Identify, based on last month's billing, all people who spend over $4 calling places other than their own homes with a calling card. For those who did, recalculate their bill by giving them a 12% discount off all weekday card calling, regardless of where the call terminates.
2. Identify, based on last month's billing, all people who spend over 30 minutes making international calls on weekends. Recalculate their bill, giving them a 15% discount on all international calls which lasted over 8 minutes.
3. Target the customers who fall into one (or both) of the above categories and determine how much profit was made off of each, based on the recalculated bills, exclusive of fixed costs (e.g. printing and mailing a bill). Only select those who provide at least $3 per month profit, in order to cover billing and other fixed administrative costs, and still provide a reasonable profit.
4. For those customers who fall into the above two categories and provide the required profit, determine whether or not they have already left for another company.
5. If they have left for another company, send them a check to come back. Make the check out for 60% of the non-fixed cost profit of the recalculated bill, and require them to remain for 3 months if they sign the check. Enroll them in a plan which gives them a 15% discount on all international calls over 8 minutes and/or a plan which gives them a 12% discount off all weekday card calling, depending on which category(s) they were in.
6. If they have not yet left, credit their account for 35% of the non-fixed cost profit of the recalculated bill, and enroll them in the plans described in the previous paragraph. Send them a letter telling them that this is being done.
7. Get the letters and checks in the mail within 3 days.

1. Recalculate revenues based upon two new calling plans.
2. Determine the profit for every call, based upon recalculated revenues, in order to determine the profit per customer.
3. Categorize each customer based upon multiple criteria, including whether or not they're still our customer, how much they spend calling away from home with a card, how much they spend calling internationally on weekends and how much non-fixed cost profit we make from them (all of this based upon the recalculated revenues). Any 'whole customer analysis' can be difficult, and this kind prohibits us from using any sort of aggregate.
4. Create lists of customers depending on what criteria they meet.
5. Do all of the above in 3 days.

The IT contact offers this as a conclusion: Any one of the above requirements would take more than 3 days. Things like "whole customer analysis" and "recalculating revenues" are much more difficult than they appear. We can't do it with SQL, so custom code must be written. Because of the iterative nature of the request - the need to solve one problem before moving on to the next - there will be many passes through the database - in fact, once the software is written, it may take several days just to run it. We estimate several months to get the answers, mostly because of the need for custom software. And realize that in order to do this, we're going to have to pull some of our best people off of other projects. The short answer is that it can't be done.

But if the IT staff is using MAST, they should say: "No problem. Is there anything else you'd like?"

MAST, of course, does not send checks or letters, or enroll people in calling plans. But it can perform all the analyses, create the lists with all the necessary information, including customer ID number, how large of a check or credit the customer should get, which categorization(s) the customer belongs in (what plan(s) they should be enrolled in). If it was desirable to send a new copy of the discounted bill to the customer, MAST could write out each call with the new revenue applied. MAST will also provide a sliced-and-diced spreadsheet output that shows how much of an effect the company will experience by applying these discounts, and how much of an effect will be felt by each category of calling or customers.

If this scenario did occur, it is likely that the benefits that the telecom would reap from this single analysis would far outweigh many years of costs in running a MAST system.