Little Result Precision Helps Add Value

Jul 03, 2012 by David Loshin in Data Management, Master Data Management

The performance measure (the one we have been using is “average check per restaurant customer”) that is computed from big results provides a benchmark. But if I wanted to use that performance measure for improvement, I would prefer to consider what pins that average amount to its current location and how I might improve that number.

• Understand what motivates sales with the good customers;

• Determine what types of customers are desired and what are their characteristics and behaviors;

• Consider ways to “promote” second- and third-tier customers into better customer segments;

• Explore ways that new “good customers” can be acquired

• Determine which types of customers are not desired and what can be done to change behaviors;

• Modify the “intended experience” so that undesired customers will be de-incentivized to visit the restaurants.

One idea is to increase the amount of restaurant sales for each customer. That would make sense, although it might imply some amount of effort across the management chain: improving the menu by adding new items, modify the marketing and advertising effort, training store staff to promote additional items, modify the sales processes, introduce bundling and upsells, or increase the cost of the menu items.

Another approach might be to better understand who your customers are, why they do what they do, and use that information to adjust the center point of their normal distribution. Some ideas might be:

However, this approach might also motivate a different set of ideas: to understand the causal behavior behind customer purchases and alternate measures that can help in planning improvements. If an undesired customer is one whose check is below a certain value, introduce measures that identify why those checks are so low. This provides greater precision, accuracy, and subsequently insight into the little result.

One example of an undesired customer might turn out to be the one that just buys coffee and occupies space for hours at a time. In turn, you can de-incentivize that behavior by modifying the rules: if you are sitting at a table, set a minimum charge, or state that customers only buying coffee must sit at the counter. Both of these approaches address the core measure of average check per guest –the table might be made available for other guests, the check amount is increased to at least the minimum, and the average amount per guest will become greater. Or the undesired guest decides to go elsewhere, in which case that low check is no longer added to the average.

My conclusion is that when there is a real intent to use the performance measure for improving performance, the accuracy in the determination of the little result can add value.

Tags: Data Management