Analytics Blog

Optimizing Parts Of A Whole

Commentary on choice modeling is confined mostly to discussions of optimizing an entire product or service, its pricing, and perhaps even its inclusion in the broader portfolio.

There is considerably less information floating around for times when the focus of the work is not the whole product, but rather a part.

So what do we do with those products that are not wholes, but rather some individual component or ingredient? Often manufacturers want to optimize the features and the pricing for their product parts. The knee-jerk response is to include the individual part in a choice exercise and vary its features and pricing.

Imagine there is a company that makes engines that are purchased by manufacturers of outdoor power equipment, such as lawn mowers. The engine company likely wants to know how to design the most appealing and highest revenue- or margin-generating product. We could design a choice exercise that simply optimizes the features and the pricing of an engine. After all, the company that makes these engines isn’t really interested in the actual end product (i.e., the lawn mower). This sounds like a great solution, but people who need lawn equipment don’t buy just the engine. They buy the entire product.

We need to design the choice exercise to allow respondents to make purchase decisions as they would in the marketplace. People buy lawn mowers or laptops, not engines or processors. The correct purchase context promotes more real-world reactions that produce better data. We want to place the consumer in the mindset and framework of the actual purchase experience.

There are limitations. Choice modeling studies are conducted largely online. An online survey representation of store shelves filled with products is not really the same thing as a real shelf in a brick-and-mortar store, but it is good enough in most cases. Likewise it is better to ask respondents to purchase a complete product versus a piece or part, especially when pricing is involved. Consumers are much better at buying an entire product with a realistic price.

If the goal is to understand purchase likelihood for a part (or component or ingredient) plus pricing in a more realistic sense within a choice exercise (i.e., a feature ranking isn’t the sole objective of the research), consider the following:

• Context is important. When optimizing ingredients for sliced sandwich bread, for example, design the choice exercise as a shelf set of bread products (white, wheat, whole grain, ancient grain) that contain various ingredients and pricing. All other things equal, improving the realism of the purchase decision will yield better results.
• Keep it simple. It’s not necessary to boil the ocean to optimize a part and its pricing. Placing respondents into a semi-realistic purchase situation can work too. For example, ask respondents to assume they are shopping for their favorite brand/size of sliced sandwich bread. The ingredients (grams of protein, fat content, sweeteners) and price are the varied elements. Holding the product constant forces more focus on the ingredients, but within a somewhat realistic purchase decision (buying a loaf of bread, not just the protein or fat content).
• Leverage learning for the whole, not just the part. For instance, companies that make processors for computers have major skin in the computer sales game. Research conducted by parts manufacturers can also provide valuable intelligence about the entire product package among potential buyers. Testing other attributes, such as brand, screen resolution, graphics cards, and so forth, may yield additional information that the chip manufacturer could share with its customers (computer manufacturers) in the spirit of partnership.

Optimizing parts (or components or ingredients) doesn’t need to be overly complex or mysterious. Preserve the context of the purchase decision, simplify if possible, and consider using any broader insights gained to cement a relationship with a buyer. Companies following these guidelines will be well on their way to configuring the optimal part with respect to the whole.

Elizabeth Horn

Senior VP, Advanced Analytics

Beth has provided expertise and high-end analytics for Decision Analyst for over 25 years. She is responsible for design, analyses, and insights derived from discrete choice models; MaxDiff analysis; volumetric forecasting; predictive modeling; GIS analysis; and market segmentation. She regularly consults with clients regarding best practices in research methodology. Beth earned a Ph.D. and a Master of Science in Experimental Psychology with emphasis on psychological principles, research methods, and statistics from Texas Christian University in Fort Worth, TX.

• 
•