Guest post from Kurt Peckman, Program Director, Predictive Analytics, IBM Business Analytics
Before fully embracing the holiday season, I have to take a step back and tell a terrifying story from Halloween demonstrating how even children think about making better decisions.
Last Thursday, Nov. 1, the committee started their strategic planning for 2013.
The “committee” is made up of my three children. The “strategic plan” pertains to analytical decision making for Halloween Trick or Treating in 2013.
Before I provide key insights into the 2013 strategic planning process a few ground rules:
1. No judging this committee’s participation in a pseudo-holiday such as Halloween. Rather, appreciate the decision to participate through the eyes of a typical 10-year-old:
“Wait! I get to dress up like a ninja or stormtrooper, which I would do anyway. But, now I get to run from house to house and ring door bells and people give me free candy? And all of this is encouraged on Oct. 31 from 3:00 to 6:00 p.m.? I’m in!”
2. The discussions that occur during the strategic planning sessions rival that of the cut-throat office banter from Glengarry Glen Ross (“Put the candy down. Candy is for closers. You call yourself a closer?”).
It’s important to recognize that even kids actively participate in a Decision Management process. They start by linking the overall strategic plan (gather the best and most candy) and the operational decisions: Should we Trick or Treat at this house?
To guide their execution in 2013, this committee built a decision framework using (elements from) IBM Analytical Decision Management. In, particular, they used:
Business rules to govern decisions:
· Car(s) in driveway = somebody was/is home.
· Porch light is on = they have/had candy.
· Decorations on house = pro kid/pro candy.
· Someone is already out front with a bucket of candy = **DECISION POINT** >> Person with candy bucket is a mom-type wearing a pumpkin sweatshirt = GO; thing with candy bucket is a zombie who might grab me = NOGO.
**Some of these rules were likely based on predictive models and/or statistical analysis from existing data.
Models to drive predictive and probable states:
· Car type “A” in driveway = great candy (e.g., “…it’s the house with the car wrapped to look like candy – I think the lady works for a distributor…”).
· Age bracket = potential for jive candy (e.g., “who gave you a cough drop?”).
· Certain occupation = no candy (e.g. “she’s a dentist and gives out toothbrushes”).
· High costume cuteness = probability for double candy.
· Other kids already at a door = quick access to candy.
· Dad knows that dad = potential for severe route delay due to chat.
Optimization to allocate resources:
· What size candy bags should we use relative to our carrying capacity?
· What is the optimized route?
· Should we base it on the “traveling salesman” algorithm again?
· How many layers should we wear? It’s warm now but will be cold three hours from now.
· What kind of snack/meal should we have before Trick or Treating to maintain enough energy to finish the mission? (Extra credit: Create a differential equation to model the calories consumed from the bag of candy relative/with regards to the calories burned while running from door to door. Assume standard conditions and constraints, and test your model using Pixie Sticks and a 4th grader with above-average speed.)
Segmentation, data warehousing and reporting for the feedback loop:
· Segments that emerge after all candy has been collected and dumped on the floor include: pieces we won’t be allowed to eat (e.g., taffy due to retainer, candy that has come unwrapped); inedible items (e.g., who gave us pennies?); chocolate and cocoa related sub-segments; hard candy; things to trade with neighbors; gum and gum-like pieces; and so on.
Hopefully this simple glimpse of easily tying strategy to operations will help your own local committee drive decision-making during Halloween 2013 – or more importantly across your business (reducing churn, identifying fraud, predicting maintenance issues).
As a final tip, ask your committee to develop KPIs now: What is most important to them: Having an extended inventory of supply? Average “yumminess” per piece (YPP)? Variety of stock?
Remember, failing to plan is planning to fail. Take it from a chaperone that carried 47 pounds of clothing, costumes, lame “trade” candy, and some kid I didn’t even know.
For more information:
· Review the four principles for adopting decision management systems
· Watch the demo and see how business rules, predictive analytics and optimization create IBM Analytical Decision Management
*Glengarry Glen Ross is copyright of GGR, New Line Cinema & Zupnik Cinema Group.