One of the most common questions asked by clients when designing a B2B research project is “What is the sample size I need to ensure the data is ‘statistically valid’?” These sample size questions typically include common statistical terms such as margin of error, randomness, confidence level, and statistical validity. In most cases we find that the client possesses limited knowledge of what these terms actually mean, or the role they play in determining a sample size. What is evident is that the foundation for these questions and concerns are products of having some knowledge of the issues determining B2C research sample. But B2B markets offer unique challenges not seen in B2C, and force a very different thought process in determining the proper sample. For example, B2B sample size can be effected by:
- a smaller target population. As an example there are a limited number of Plastic Resin R&D Vice Presidents in the US. This can make it difficult to collect a large number of responses.
- a very complex decision process, with economic buyers, user buyers, technical buyers, supply chain influencers, etc. This needs to be taken into account in sample determination, as there is often value in segmenting the data to reflect each constituent in the decision.
- all customers are not created equal; businesses often operate on the 80/20 rule, where 20% of their customer represent 80% of their revenue. It is critical that important customers are part of the research (this trumps randomness of the sample). For example, if your target market is US grocery retail a random list may not include some of the market leaders (Walmart, Kroger, Safeway, HEB, etc.). Therefore, it is typical to construct the sample to include many of the top chains and fewer smaller chains.
So, how do you determine B2B qualitative research sample size? In B2B research, sample size is primarily driven by answering the following question; “how many interviews are needed to provide a reasonable degree of reliability?” The answer is often a function of the population size and the expected variation in the answers. In most B2B studies the expected variation in the answer is addressed by segmentation, as the potential for a different perspective is often the primary reason to segment. The next issue is the population. Using the grocery retail example, a logical approach would be to break the industry into two groups by size, the top 50 chains and all other. Then overlay the segmentation scheme as follows:
- Top 50 economic buyers (title = purchasing)
- Top 50 technical buyers (title = R&D, tech support)
- Top 50 user buyers (title = operations)
- Top 50 influencers(title = customer service, sustainability)
The same segments would also be applied to “all other” grocery retailers, creating 8 unique groups. The final step is to take each group, determine the population for each group, and estimate the number of interviews needed to provide a “reasonable degree of reliability.” As a rule of thumb a minimum of 50 quantitative interviews are required to provide a reasonable degree of reliability.(Figure 1)
For example, as a rule of thumb a degree of reliability in B2B research can be obtained with a minimum of 50 quantitative interviews, but in consumer research this would be far too low (several hundred would be more appropriate given the population size).
At this point in the process, determining sample size is more an art than a science. However, there are some variables that provide direction in assigning a target interview number (sample).
- The Top 50 are more important than the rest of the market. Target a greater number of interviews in this tier.
- The importance of the segment is also a consideration. In this case, less emphasis was placed on the Influencer than other segments
- The overall comfort level of the targeted number of interviews versus the estimated population. In the above scenario, there are at least 50 completed interviews within each of the primary segments (buying influences) as well as each of the tiers (Top 50 and all others). This should provide a reasonable degree of reliability in the overall data as well as for the key data segmentation cuts.
The common B2C statistical metric of a 95% confidence level with a margin of error of 3% for the estimated population of 1475 in the B2B example above would require a sample of 620. Compare that to the 205 completed interviews that we would recommend. Even more challenging, using the same criteria the economic buyer segment, one would need 235 completed interviews out of a population of only 300 (50 in the Top 50 and 250 in all others). Needless to say, completing this large number of interviews would not only be nearly impossible given a population, but the expense could be prohibitive.