Using Decision Management to Make Your ATMs into a Sales Channel

(Posted by guest blogger, James Taylor)

My friends over at the Analytical Engine had an interesting post this week - ATM Machines As A Sales Channel. Their focus was on pre-approved offers and the possibility of loading them onto the ATM servers so that they could be displayed to customers, allowing a simple "Yes" for acceptance. They also point out that there are a number of challenges including not being able to upload all pre-approved offers, problems with limited timing especially at busy ATMs and not being able to close the sale through the ATM, requiring follow-up.

Like them I believe that ATMs will, indeed must, become better sales channels. Further I think that banks should be looking hard at the ATM experience and making sure that it meshes with the overall experience they wish to deliver to their customers. I thought it would be helpful to layout some of the steps in making ATMs work as sales channels as applying enterprise decision management or EDM to this process is a key element.

1. Identify and describe the decisions involved. These include, but are not limited to:
   1. What to display for a specific customer when they insert their card
   2. What offer(s) to make a specific customer
   3. When to make or not make an offer
   4. How to offer to fulfill the offer
2. Make the return of information to the ATM dependent on these decisions
   The process of displaying information on the ATM must be integrated with these decisions so that fixed returns (always display the same screen) are replaced with dynamic ones.
3. Specify the rules for these decisions so they can be automated at a basic level. Examples include:
   1. Usage rules such as time since last customer that focus on speed of transactions when the ATM is busy and sales when it is not
   2. Customer preference rules both explicit (never call me, customer only has checking account) and implicit (customer only seems to use the ATM to get cash)
   3. Rules that are location dependent such as not offering the option to enter the branch to complete a transaction unless the ATM is at a branch
   4. Regulatory rules based on state
4. Use analytics to improve these decisions. Consider, for instance:
   1. Segmenting customers based on behavior and lifetime value
   2. Predict offer interest for specific customers and specific offers
   3. Predict ATM usage and likelihood of acceptance at ATM to make sure it is worth targeting this customer at the ATM
   4. Predicting traffic at an ATM to avoid delaying transactions when likely to be busy
   5. Predicting most successful approach to closing offer for a given customer
5. Use adaptive control to continually try different approaches such as varying whether an offer is closed by a call back, email with a link to an online application or a request to come into a branch.
6. Optimization to manage trade-offs between profitability, cost of offers, cost of fulfillment, risk of annoying customers etc.

The best thing about this whole process is that once the decisions are identified and externalized they can easily be updated, improved and innovated. The ATM moves from a dumb machine that always asks the same questions to a personalized portal to the bank and its products.