Talking to people and interviewing is all good and fine in the beginning of your new business. Asking open-ended questions is an exploration. The answers can surprise you, and you can learn something you might not have considered before. But what if you already know what to do? You already know your direction, you’ve explored a lot. Now you need detailed data you can exploit. The unstructured data from your interviews isn’t enough to satisfy you. You want structured details.
Then do surveys.
In recent years, the field of business analytics for marketing has boomed and radically changed the way businesses approach their marketing investments and campaigns. Hiring managers now require employees in marketing positions to be able to speak the language of analytics and metrics, and corporate-level executives increasingly demand ROI accountability from marketing staff. However, it is not uncommon for marketers to be unfamiliar with analytics. They place themselves at a disadvantage because of their lack of knowledge. When marketers conduct surveys, usually they want to get exact numbers or exact customer preferences. Marketers are masters of business surveys. Nevertheless, for efficient surveys, you have to use analytics, and use it smartly. I personally think all surveys need to be smart.
For example, you know what business you’re going to be in (since your previous idea about starting a new oil company failed). You decide to open a restaurant, or to produce new coffee machines. In any case, there are many things you’d like to know about your customers so you can match your product to their needs. I can tell you in advance that asking too many questions in one survey will not work. People will balk at answering more than 15 questions on average. Think of 15 questions as your limit. However, how can you create a survey with less than 15 questions and still collect enough data about the customer’s preferences? One of the most advanced techniques for analyzing surveys is conjoint analysis, but the data collection has to thoughtful.
There are three commonly used techniques for data collection in conjoint analysis: pairwise comparison, rank ordering, and rating scale. With each of these techniques, you will have to form bundles, which represent candidate ‘products’ for you to test. Marketing researchers often refer to the bundles as ‘cards’, because research in the past used cards to represent individual bundles.
In the pairwise comparison, respondents compare two different cards with different sets of product attributes, and tell you, which attribute set they prefer. Some responders find pairwise comparison easier than rank ordering.
In rank ordering, you provide all the cards at once to the responder and ask them to rank the cards in order of preference, from their first to last choice. The advantage of rank ordering is speed, but the complexity or ordering process that the responders face is a disadvantage. Generally, responders quickly establish their most- and least-preferred choices, but find it difficult to rank the choices in the middle.
The third data collection technique is the rating scale. With this technique, you ask respondents to rate each choice on an absolute scale. The rating scale can be easier for respondents than ranking, but some of them might find it difficult to assign ratings for fine rating scales. To counter this disadvantage and to improve the consistency of results, I recommend providing guidance to respondents on how they should assign ratings to their choices, explaining in a short sentence what the particular rating means. For example, accompany a five-star rating ‘outstanding’ with a clarifying description like ‘I will definitely buy it!’.
Conjoint analysis is a market research technique used to examine the trade-offs consumers make between product attributes. By examining the trade-offs, you can infer the value that consumers place on individual attributes. Conjoint analysis is appropriate for situations where you need to quantify customer preferences for certain attributes. The conjoint analysis process reduces preferences for certain goods and services into values, called ‘part-worth’, that particular attributes hold for an individual. Part-worth shows the willingness of a customer to pay for a certain attribute. Companies often use conjoint analysis in product and service development and for market segmentation. A famous example is a feature selection for a casual student bag described by Hauser, a professor at the MIT Sloan School of Management in his ‘Note on Conjoint Analysis’.
A typical analysis with many attributes and many levels could result in hundreds or even thousands of combinations. To reduce the number of cards, researchers apply so-called fractional factorial techniques (for example, Taguchi orthogonal arrays). As I said before, the rule of thumb is to work with a manageable amount of cards (say, less than 15 cards).
In essence, conducting conjoint analysis has the following steps: defining the product attributes and form cards, or bundles; ask consumers to state their preferences for each bundle; code the data in a special form and conduct a regression analysis that links attributes with customer preferences; calculate the part-worths (the customer’s willingness to pay for a particular attribute). The part-worths will be simply the coefficients in the regression equations, different for each respondent. However, you can combine all the regression equations into one objective function, run an optimization algorithm, and come up with an attribute set that will drive the highest revenue from the product sales.