French fries and the law of unintended consequences

Editor’s note: Brad Barash is vice president of Decision Insight, a Kansas City, Mo., research firm.

There was a time when many believed unsaturated trans fats were healthier than saturated fats. That belief (along with cost implications - let’s not kid ourselves) led to the overwhelming use of partially-hydrogenated oil in the preparation of American staples like french fries. What the health advocates who supported the use of partially-hydrogenated oil did not realize at the time was that unsaturated trans fats raise the bad cholesterol and lower the good cholesterol.

This is an example of the law of unintended consequences. Every cause has more than one effect, and sometimes these side effects can be more significant than the intended effect. It is nearly impossible to prepare for every potential implication of a specific action, but there are approaches that can bridge that gap.

Fast-food restaurants can face their own battle with unintended consequences when they look to optimize their menus. Menu optimization requires a complete understanding of how appealing an item is and how an item will affect other menu-item sales. Avoiding unintended consequences and instead generating a positive outcome means going beyond asking consumers how they might feel about a potential new item and delving into how a potential new item may or may not change purchase behaviors.

The trouble is, the marketing research techniques that most restaurants use today fall drastically short in considering the consequences, intended and unintended, of new strategies and tactics. Let’s use french fries to illustrate this. Health trends in the U.S. have created a bizarre segment of the population that refuses to eat fries (I shouldn’t judge, but come on … How could you not eat french fries?!). The implication is that many restaurants have seen a drop in side-item incidence. These restaurants, therefore, have searched for healthier alternatives to add to the menu in an attempt to restore side-item orders.

The natural tendency, from a research perspective, would be to test these healthier items among non-fry users (Are the items appealing?, How likely are non-fry users to purchase them?, etc.). Makes sense, right? Wrong.

Here’s the rub: French fries are among the most profitable items on any menu. So, while it may be a benefit to increase side-item incidence among non-fry users, it is critical to consider the law of unintended consequences. That is, if current fry users switch to the new item, overall side-item incidence may still increase (given the reach of non-fry users), but profitability may decrease with the loss of fry orders among current users.

Gauge the interaction

It is critically important to understand not only the performance of one potential menu item but also the interaction of that item with existing items. It is common for restaurants to use marketing research to optimize a menu, solving for which existing items can be safely eliminated from the menu and which of a large volume of potential new items should be added to the menu. Simply getting an appeal score for every item is not enough. Again, it is critical to gauge the interaction of items against each other.

Given this, many research companies will recommend TURF analysis (total unduplicated reach and frequency). TURF can provide some insight into identifying items that reach customers who would not otherwise order an existing item (this is the definition of unduplicated reach).

This approach can add value but has some flaws to be aware of. Our experience is that most restaurants have already maxed out reach with just a small portion of their current menu. And, it is possible to increase unduplicated reach but still end up with a situation similar to the french fry example noted earlier.

Other measures beyond unduplicated reach can provide substantial value. There are a few simple exercises that provide several beneficial measures. The first is a simplified order exercise. Expose respondents to all menu items (even simple text descriptions of each item), existing and proposed, and ask them to indicate which they would order over a series of visits. They could order the same item(s) multiple times or order something different each visit.

This exercise provides two key measures for every menu item, existing and proposed: breadth of appeal (What percentage of customers order the item at least once?) and depth of appeal (How often is the item ordered?). There may be items that reach a smaller proportion of customers but those customers order that item more frequently.

Substitutability is another important measure. Respondents can be queried on what (if anything) they would order instead if the item they select is not available. This provides a measure of which items are substitutable with each other, and importantly, which items are less likely to be substitutable with any other item. There is also a loyalty measure that can be obtained by asking customers which items must be on the menu for them to continue visiting the restaurant as often as they do now.

Not an easy task

The bottom line: optimizing a menu is not an easy task, particularly when there is a large volume of proposed items. One perfect solution simply does not exist. As with any research, several factors should be considered when deciding which existing items should be eliminated and which new items should be added. The research measures described above should be evaluated in unison with other measures, such as ingredients needed (Does the item use unique ingredients or does it leverage ingredients also available for other items?), operational difficulties and profitability.

Each of the measures previously described can help to narrow down a large list of tested items to a more manageable size. Ultimately, before any menu changes are made, there are two critical areas to address from a research standpoint.

The first area is one that most current research techniques address: attitudinal measures. How appealing is the menu? What is the impact on customer satisfaction? What is the impact on intent to visit in the future? What is the impact on various restaurant perceptions?

The next area is equally as important but seldom addressed in research: behavioral measures. What is the impact on product mix (proportion of orders across menu items)? What is the impact on average spend?

Behavioral measures are not often addressed in research because they have been difficult to obtain. Historically, short of actually rolling out the menu, only an in-market test could provide that output. In other words, implement the new menu in a subset of stores and then compare sales of those test stores against the rest of the system.

Virtual representation

Today, online research techniques have made it possible to project those key data points in a survey - without the risk of actually implementing a strategy in a live market. The output is obtained by exposing respondents to a virtual representation of a menu or a menu board and asking them to order just as they would in real life. The result is simulated sales data, including product mix (proportion of orders across items and categories) and average ticket price. Following a menu-order exercise, respondents can be probed on satisfaction with the menu, intent to visit in the future, perceptions of menu categories, value perceptions, etc.

Back to french fries: This approach would allow a restaurant to understand if a new, healthier side item could accomplish all of the following: increase total side-item incidence without stealing from french fries, increase average spend (due to the increase in side-item incidence), and improve perceptions of the restaurant offering healthier items.

Without a robust menu-order exercise, only the last measure could be obtained - restaurant perceptions. Again, it is very conceivable that test items could improve restaurant perceptions and even increase side-item incidence, but they may cannibalize french fries, resulting in average ticket prices that are flat - and a decrease in profitability. That side effect would not be discovered until after the new product is actually available in the restaurants.

Tremendous impact

Over years of restaurant research, we have seen the influence of several factors on how and what customers order. Everything from the flow and placement of items on a menu or menu board, the imagery used, promotions and, of course, pricing, can have a tremendous impact on orders. That impact is manifested in product mix and average spend. Yet, again, most restaurants depend on attitudinal measures when testing these solutions. So, they expose respondents to a menu or menu board and ask them how appealing it is. Attitudinal measures are, of course, important, but they should never be used in isolation - not when it is so easy to obtain the behavioral measures as well.

Short of operational efficiencies (i.e. improving profitability of existing business), there are three primary approaches to growing sales in the restaurant world: attracting new customers, getting current customers to visit more often and getting current customers to spend more each visit.

For the first two approaches, both focused on generating additional traffic, traditional research techniques only tell part of the story (intent to visit). It is critical to also understand how traffic-building initiatives impact what people order. For example, adding value-priced items to drive traffic may accomplish the goal of increasing visits, but does the increase in traffic overcome the potential drop in average spend if a high proportion of visitors migrate to the lower-priced items? A thorough menu-order exercise can provide that behavioral output and can also gauge the effectiveness of strategies used to increase average ticket price, allowing you to avoid getting trapped by your own set of unintended consequences.