What is the most important part of a conjoint project? Well-balanced design and representative sample aside, it would be the results! You may ask yourself, “what’s the best way to understand my conjoint data?” or “how do I share the most important information with my team, executive board, or client?” One of the best tools to present the findings from your research is to use a market simulator.
Conjoint Review
When you conduct a conjoint experiment, you estimate a utility score that quantifies respondents’ preferences for the levels of each attribute. There are many ways you can analyze utilities: you may create charts, graphs, or summarize the average utilities. However, the average utilities may omit valuable data. Averages hide a wider range of behavior that you might see at the segment or individual level, especially in a heterogeneous sample. Depending on your situation, you could make costly mistakes following only the average utilities. Let me share an example of one way only looking at average utilities turns sour.
Let’s say you conducted a conjoint on different sodas at varying prices and pack sizes. And say the soda flavors were Coke, Pepsi, and Sprite. You may have two kinds of respondents, those who prefer Pepsi and those who prefer Coke. Their utilities might look like this:
| Respondent1 | Respondent 2 | |
| Coke | 90 | -100 |
| Pepsi | -100 | 90 |
| Sprite | 10 | 10 |
If you had 100 of each type of respondent, the average utilities across all 200 respondents will look like this:
Average Utilities
| Coke | -5 |
| Pepsi | -5 |
| Sprite | 10 |
On average, there is no difference between Coke and Pepsi, and Sprite may look like it is most preferred. This is opposite from what we saw in the individual respondent data: the respondents had the highest utility for either Coke or Pepsi. We could imagine if Respondent 1 went to the grocery store for soda, they would purchase a Coke, not Sprite. The table below shows which soda the respondents would choose, based on their individual preferences.
Which Soda would you choose?
| Coke | Pepsi | Sprite | |
| Respondent 1 | X | ||
| Respondent 2 | X |
Though this is a simple example, it helps illustrate how average utilities can sometimes hide information when you have lots of different people in your data and how simulations based on individual- or segment-level data can be very helpful. We highly recommend estimating individual utilities through Hierarchical Bayes, which will give you respondent-level data. However, looking over every individual respondent can be overwhelming, especially considering a typical Choice-Based Conjoint (CBC) has 200 respondents or more. You can go beyond just the average utilities and consider other things like standard errors, boxplots, and histograms to understand how much variation there is in the data. To read more on how utilities and attribute importances are created and how to avoid misinterpreting them, take a look at this article.
In summary, utilities quantify respondent preferences, but they may not be the best deliverable to share with your client or team. Utilities are often difficult for non-researchers to understand and are frequently misinterpreted as ratio scaled when they’re actually interval scaled. If you present only the average utilities, you will possibly miss out on important research findings and may spend precious time explaining how to interpret utilities rather than the actual results of your research.
So, what do I do?
Simulators are helpful because they can help bypass a lot of the problems with reporting average utilities. There is no need to worry about incorrectly interpreting utilities or missing out on individual data when you use a market simulator.
In a market simulator, you create different product configurations with your conjoint attributes. The simulator then takes the preference data (utilities) from your respondents and computes what percent of the respondents would prefer each product. You are left with a simulated preference share – a percentage (like in the image below). You can easily say “Product A has a 75% share of preference vs Product B that has a 25% share of preference.” No need to worry about misinterpreting the results.
Simulators also look at individual respondents. From our Coke vs Pepsi example above, the simulator would recognize that not all respondents prefer Sprite. If you simulated both a Coke and a Pepsi soda, the shares of preference would be around 50% for each product (all else equal). You can also include demographic data in your simulator, like country of residence or household income, to help you get a better picture of potential segment preferences.
Simulators are a “choice laboratory” for testing alternative market strategies. They allow you to play different “what if” games with your conjoint data and allow you to search for ways to improve your product or service. You can answer questions such as:
- At which price would respondents switch to a competitor ?
- Can we modify our product to reduce cost while maintaining share?
- Should we launch a high-end product or a budget model (or both)?
- Will the new product cannibalize our own sales?
Simulators are not a crystal ball, but they are one of the best ways to understand and interpret your conjoint results and allow you to gain important insights from your data.
You've convinced me, what are my options?
Sign up for a free tour of our market simulators and to get more information about how simulators work, what simulators we offer, or how to get access to a simulator.
Sawtooth Software has a few simulator options as well as an easy way to export an Excel-based simulator from our desktop applications (Lighthouse Studio and Choice Simulator). Let’s dive in:
Choice Simulator
The Choice Simulator is a windows-based program available within Lighthouse Studio or as a standalone software. The Choice Simulator allows for segmentation, sensitivity analysis, as well as advanced settings like search algorithms, distribution and product awareness assumptions, cost and revenue, etc. You can create multiple market scenarios and change the simulation method in the Choice Simulator. There is a free Client Simulator export as a Choice Simulator file or an Excel file.
Online Simulator
The Online Simulator is a web-based simulator tool available in Discover or as a standalone product (app.sawtoothsoftware.com). In the Online Simulator, you can create numerous market scenarios, use multiple simulation methods, add segmentation variables, run a sensitivity analysis, and get automatic simulator insights.
Excel Simulator
Excel-based simulators are another option to simulate your results. Excel Simulators are completely customizable, and you have complete format control. You can add dropdown menus, add variables for segmentation or weighting, or use your company’s color scheme. You can use a Share of Preference simulation method or we have a Randomized First Choice plugin (for Windows-based Excel) as well. Below is one example of a user-built Excel simulator that has a customized interface with dropdown menus, segmentation variables, and colors.
(Above is on example of a user-built Excel Simulator)
More information on Market Simulators:
Introduction to Market Simulators for Conjoint Analysis (2019) Technical Paper
Avoid These Common Mistakes When Conducting Market Simulations to Derive SKU Demand Curves
Or checkout some webinars on Market Simulators