You will get some of the best insights about your customers and market by performing primary research. That’s because you can tailor the research specifically to your needs and circumstances. But performing primary research takes a lot more time and effort than relying on something off the shelf, and it’s not uncommon to go through the entire process of preparing, fielding, and analyzing a research study only to arrive at a set of “fun facts.” These are interesting tidbits that are fun to tell others but ultimately have no business impact.

The best way that I’ve found to avoid this trap is to work backwards, starting with the objective, or better yet, the actual statement that you want to be able to make at the end of the research.

For example, at my last company, we wanted to understand the apparel buying behavior for people that work in office environments with different dress codes: casual, business casual, etc. If we had gone into the research with only that sentiment, we might or we might not have come away with some useful information. Instead, we spent more time defining our research objectives and broke it down into more granular statements.

The initial sentiment was refined to “We want to determine how much employees in these different office environments spend each year on clothes so we can prioritize who we target for customer acquisition.” That led to a series of individual statements / questions to be resolved through research:

  • How do we accurately define the different office dress codes? Can we say that people that wore this set of items to work over some period of time were more likely to be in one dress code than another? I described how we did this in this post on cluster analysis.
  • Approximately how much did people in these different office environments spend on work clothes over the last year? Can we say that employees in these environments spent $X on average versus employees in these other environments, and we know that is an actual difference (not just noise in our data)? What does this look like for people that have been in their job for over a year versus people that just started their role in the last year (assumption being that a new job/role is a catalyst to buy new work clothes).
  • Where do people go to shop for different types of clothes in these office environments? This helps you understand price sensitivity, competitors, etc.
  • What industries / job types are the most common for each office dress code?
  • So on and so forth

The next step was to prioritize and rank the list of questions, select the ones we thought were absolutely necessary, and then build out the best interview and survey questions to answer that set of questions.

By getting as specific as possible with the questions that we hoped to answer through our research, we were able to design it so that it would be almost impossible not to come away with answers to those questions. It also ensured that we remained focused on what we were trying to learn.

I first learned about the idea of working backwards from the desired output from marketing research professors at Kellogg, and it was an approach that we used at BCG often (and sometimes took to the next level by creating the actual output with placeholder data before starting the research). I’ve since become a true believer that you should almost always clearly define your desired outcomes before starting any research to get the best results.