Imagine a household brand whose new refill pack sells more each month. The commercial team is ready to expand it. Then a round of customer interviews comes back, and the people interviewed describe the refill as awkward and disappointing. One source says the product is working. The other says its customers are unhappy with it.
When sales data and customer interviews disagree, the first step is to establish what each one describes: which customers, over which period, and against which idea of success. Often the two sources are answering different questions. Sometimes the disagreement is real, and it deserves investigation before either source settles the decision.
Two records of different things
Sales data records what was bought, where, when and at what price. It is a strong account of behavior under the conditions that applied: the range on the shelf, the promotions running and the alternatives available. It says much less about why people bought, what they expected or whether they will buy again.
Interviews give an account of experience: what people hoped for, what happened when they used the product, what they compared it with and what they intend to do next. That account depends on who was interviewed and what they remember, and on its own it can’t say how common an experience is.
Neither source is automatically the objective one. A sales figure may count shipments to retailers as well as purchases by shoppers. An interview can be vivid and unrepresentative. Each needs to be read for what it measures before the two are compared, and either can show that the other was measuring the wrong thing.
Where the difference might lie
In the hypothetical refill example, several explanations are possible, and none should be assumed.
The populations may differ. Growth could come from new buyers, while the interviews were recruited from long-standing customers of the original pack, or from people who had contacted the brand. The periods may differ too. Sales may reflect first purchases during an introductory offer, while interviews capture weeks of use, whose effect on repeat purchase may not show yet.
The alternatives matter as well. If the original pack has become harder to find, people may buy the refill because it is what the shelf offers. A purchase records a choice among the options available, and that choice may or may not be an enthusiastic one.
The two sources may also define success differently. A sale counts as success the moment it happens. A customer judges the product in use, against what the old pack did or what the refill promised. Both accounts can be accurate at the same time.
It may also be a real contradiction, such as a measurement problem in one source or an effect the other has missed. Either way, the disagreement shows where the evidence is thinner than it looked.
Evidence that would tell the explanations apart
Each explanation predicts something different, which makes the next step more focused than a large new program.
Splitting sales between new and repeat buyers would show whether growth depends on people buying for the first time. Repeat purchase over the following months would show whether disappointment turns into lost sales, and a look at how the interviewees were recruited would show whether they reached an unusual group.
Often the most useful next evidence already exists in the business, or can come from a small piece of work aimed at one question. Choosing it well depends on knowing which explanation would change the decision.
One team of health services researchers treated the same problem as a reason to look harder at both sources. In a pilot trial of welfare rights advice for people aged 60 and over, the measured outcomes showed little difference of practical importance between groups, while interviews with participants suggested wide-ranging effects.
The authors explored the discrepancy, including whether the datasets were comparable and whether their outcomes matched, and their approach showed “how such discrepancies can be harnessed to interrogate each dataset more fully.” Their setting was far from a supermarket shelf, and the discipline applies to a commercial team all the same.
A recommendation that keeps the uncertainty visible
Once the likely explanations have been examined, the recommendation should say what each source shows, which explanations the evidence has ruled out and which remain open.
Each remaining explanation may point to a different decision. If the interviews reached an unusual group, the expansion plan may need little change. If growth depends on first purchases that don’t repeat, it needs rethinking.
Forcing agreement helps nobody. A neat story that explains one source away can leave the business confident about the wrong thing. It is more useful to say what is known, what is still uncertain and what would change the recommendation, so the team can decide how much to commit while the question is open.
Sometimes the explanation stays unresolved. That is still a usable result when the team knows which decision it affects, what would settle it and what it would cost to wait.