While Choice-Based Conjoint (CBC) works great for many research problems, its models are limited to making an independent choice: one phone compared to other phones, for example. If you want to model dependent choices, such as a phone and then a cellular plan, or a meal (appetizer, entrée, dessert), you’ll need Menu-Based Conjoint (MBC). At the Quirks Event in London, Aaron Hill of Sawtooth and Chris Moore of Ipsos MORI discussed the use of MBC.
MBC can handle these types of problems:
- Cell phone plan/phone purchase
- Bundled cable offerings
- Restaurant menus
- Feature configurations to add to a new car
- Multi-subscription bundles
- Business software suites
- Cruise inclusions vs. add-ons
- Insurance policy riders
- Transportation options.
MBC exercises allow you to simultaneously measure correlated decisions where consumers create their own product through their choices. In addition to CBC’s benefits of price elasticity and cross-effects, you can now model the real-world complexity of actual decision-making processes. For instance, raising the price of a chicken appetizer from £2.99 to £5.99 might mean people pick a different appetizer but switch the main course to chicken.
Chris then shared a case study of MBC in action. On behalf of TGI Fridays, Ipsos Mori conducted an MBC study that led to 31% year-over-year net increases in profit, compared to 12% in the restaurants in the control sample!
The initial goal was to optimize the pricing of key dishes on their menu in order to maximize profit while cannibalizing key competitors. The survey collected 1,490 responses from UK residents aged 16 to 35 who eat out at least monthly in competing restaurants (branded chains with table service and alcoholic beverages, with a TGI Fridays in their area). The questionnaire included a screener, a CBC exercise of TGI Fridays vs. a preferred competitor, and the MBC exercise.
After fielding completed, Chris built a simulator to estimate how many people would go into the restaurant on an average month, then created a model to predict volume of dishes. Since the goal was to increase net profits, the client provided all fixed and variable costs so that Chris could estimate profit per configuration.
Chris provided the following tips for conducting Menu-Based Conjoint:
- Be pragmatic. MBC projects can become complex and expensive very quickly.
- Restrict the choices to make analysis easier. For TGI Fridays, Chris modeled only the high-volume dishes, responsible for 70% of purchases: 5 starters, 2 bundled offerings, 10 main courses, 6 desserts, and 5 drinks.
- Simpler models with fewer cross-effects work better. Only include significant effects. For instance, since only 4% of restaurant patrons choose two starters instead of an entree, respondents were limited to selecting a single starter and had to select an entrée.
- MBC is very data hungry. In order to model cross-effects, N=1000 is a good starting point.
- Include holdout tasks to check the validity of your model, simulating actual tasks.
- Don’t use the None option in choice exercises if optimizing for revenue or profit, where it can lead to misoptimization. “Respondents are trying to help, and most want to make choices; you don’t get that level of people leaving a restaurant without ordering.”
- Context is extremely important. “What is the occasion? Who is buying? Are there different menus by time?” Ground everything in the particular context. Respondents were asked about their last dining occasion at a branded restaurant and then everything was framed with this most recent occurrence in mind; the choices even changed by weekday reported in order to exclude items not available Friday to Sunday.
- Don’t underestimate the time needed in the set-up phase; unlike CBC, the MBC method has been around commercially for a decade but is not yet used very often.
As TGI Fridays discovered, Menu-Based Conjoint can drive bottom-live results. Make sure this technique is on your menu of approaches.
Author Notes:
Jeffrey Henning
