It's counter to think that research should be designed to flex based on what is known at the moment, but it's actually a sign of effective communication and learning.
In business to business, companies see different customer groups in terms of power, geography, ambition and potential. They may also face groups of individuals within companies that have distinctly different needs, like managers, buyers and users.
So how to develop high confidence in research when you have more than one type of market or customer to understand? Should the research sample grow to reflect each group and subgroup?
Quantity of surveys doesn't make a subgroup either relevant or representative, quality of insight does. The best approach to confidence is a combination of economies of scale and skill. Both can be accomplished with a process that we call bucketing, a simple tool for interpreting qualitative results into easily digested bites of learning.
Bucketing creates economies of scale by quickly testing your segmentation assumptions. Any research project should start with a basic map of the groups that you think you know.
For example, say you make elevators, but business has been flat. You might have potential customers in markets all over the world. But wherever they are, they are similar in that they all need lifting and someone to help them when they can't. (For many companies, their products act as a virtual meeting point based on common behaviors seen across disparate groups.) So while even though the market includes landlords, developers or plant managers in many countries (you probably have them grouped this way), your weak service levels might be sending some towards other suppliers and into a new bucket that we might call Disgruntled Defectors (DD's). Once identified, DD's will feel like an elephant in the room.
On the other hand, you also have customers that aren't leaving yet, perhaps because your products run longer between service calls. Call them the the Loyal Lifters (LL's).
Now your segmentation is oriented towards behaviors.
Strong qualitative researchers (the skill part) will use bucketing to guide the research effort towards groups that require "deeper dives", and away from groups that are either already well-known, or easy to manage or unattractive.
If a few DD's are found in early stages of a customer satisfaction project, the research should first be re-oriented just long enough to confirm the trend. Then the research should be re-oriented again to find desired outcomes and test response to corrective actions (for both DD's and LL's). Then the buckets should be hardened into action plans based on their responses.
The hard part is this: most of us have been taught that polling shouldn't stop until
... a solid majority can be found, or
... bias can be removed with a flood of data.
This basically means that we're collecting data for data's sake, when we should be collecting just enough to prove or disprove hypotheses.
The ability of a research team to make changes along the way depends on whether they see the trends early enough. Since business research isn't about elections, the most important factor determining research success is not sample size -- but research transparency and the amount of critical thinking that is done and shared through the project. To do this, many of the questions asked should be designed to find new answers, not confirmation of what was known before. Flexibility and progress are natural outcomes of the bucketing process.
In the end, your map should be redrawn to reflect the simpler, more actionable buckets, err groups, identified by the effort and confidence factors and margin of error reapplied. If done right, a highly efficient research project will create very high confidence in future decisions, because there will be fewer, clearer groups to manage and know.
