Some of the most difficult market research tasks are associated with the run-up to a product launch. Making predictions is challenging and when the number of unknown variables increases, so does the complexity of the task. Planning for a product launch taps the best that market research can provide, and it can mean turning to machine learning to get a handle on some of that complexity. Using Bayesian networks to create probabilistic models, market researchers can inform advertising, marketing, and distribution decisions associated with a new product release.
How Do We Define Market Share?
Let's make a brief sidebar comment or two on market share, that elusive measure that everyone wants to know. Market share is typically reported as a percentage and reflects the portion of a target market that is an actual customer base. A "market" is a universe of customers or clients who are interested in a particular product, service, or company. An increase in market share often makes economies of scale available that can contribute to earnings and profit margins so, naturally, businesses seek to increase their market share.
Market share may be reported as revenue or unit sales from a specific market divided by the total possible revenue or unit sales in that market. Because the figure for the possible / available revenue or unit sales in a specific market must be estimated or extrapolated, it is common for a firm to use market research to arrive at that figure.
Bayesian Networks for Market Share Simulations
BayesiaLab has created a powerful Bayesia Market Simulator that can be used to compute market share for different product offering scenarios.
A market researcher can start with a data base that describes several offer choices that are believed to have particular appeal. Then, using the learning algorithms provided by BayesiaLab, a probabilistic model can be automatically created (or "induced," if we want to use the Bayesian vernacular). The probabilistic model factors the offer choices and the known market characteristics. And, a market researcher can further modify the Bayesian network that is produced, ensuring that expert knowledge is also in play.
Using the Bayesia Market Simulator is easier than you might think
- Identify & associate a "type" for each variable: offer variable, market variable, choice variable.
- Review the files that describe different markets.
- Choose one market type to use as the simulation base.
- Determine which offer characteristics you want to launch on the market.
- Define the scenario using the offer characteristics you identified.
- Decide how to characterize the market segment to be explored.
- Compute the market shares that correspond to the scenario offers.
- For each line of your market file, save the probability distribution over the offers.
- You're done! Send a thank-you email to Bayesia S.A.S.
Repurpose and Reuse Primary Research
Syndicated research and primary research are expensive. Using the Bayesia Market Simulator, market researchers can repurpose or reuse this expensive resource to be used as "pre-introduction" data in market share simulation.
An estimation of market shares of products yet to be launched can be generated on the basis of research that has already been commissioned and produced. So -- in addition to being fast (its based on algorithms and machine learning), highly practical (it deals with variables identified by market researchers)-- the Bayesia Market Simulator is also economical.
Market researchers can perform market share simulations right on their own desktop computers. The basis of this type of simulation is the Bayesian networks framework. Bayesia S.A.S. has created software packages for the BayesiaLab and the Bayesia Market Simulator. These are substantive innovative tools that can readily be utilized to enhance the forecasting capabilities of market researchers.
Bayesia S.A.S. has produced another great webinar about using the Bayesia Market Simulator. The product being introduced -- you are going to like this! -- is the new Porsche Panamera in the U.S. market. Using the Bayesian Market Simulator, the market researchers were able to confirm that Bayesian networks can be used in small (read: niche) markets for which relatively few data points or observations are available.
An easy way to learn more about BayesiaLab resources is to visit the tutorial on the Bayesia.com website. Another idea would be to join the Bayesian Belief Networks Group on LinkedIn. Consider this: the book Bayesian Reasoning and Machine Learning by David Barber sells on Amazon for $85. It is bound to be a great resource, but hink how much of a jump start someone could get on the content by regularly exploring the Bayesia.com website.
For those who feel ready for more detail, an article on the methodology can also be accessed. The article describes how the Bayesia Market Simulator can be used to create market share forecasts for new products, changes to existing products, hypothetical scenarios of external factors like changes in energy prices or transportation costs, and consumer attitude changes. The article also explains why Bayesian networks provides superior consumer choice modelng than traditional choice modeling, such as discrete choice models.