“Customers think the routine is too complicated” is a tempting conclusion from a handful of support messages. It may be right. It may also describe a particular product bundle, a confusing email or a temporary problem after an instruction changed.
AI can collect similar comments in seconds. The growth team's job is to preserve enough context to decide which explanation is worth testing.
Keep the comment attached to the situation
Imagine a hypothetical consumer brand reviewing support tickets, product reviews and post-purchase surveys. A customer asking how to use a product has bought it. A visitor abandoning a basket has not. A review on a retailer's site may concern an older variant.
Combining these statements into one theme can help discover a question. It cannot establish how common the problem is among all potential buyers. The article on review selection explores that distinction in more detail.
The OECD's review guidance discusses issues including incentives, authenticity and moderation. Those features affect what enters a feedback dataset. Record them where they are known instead of treating every comment as an equivalent observation.
Write the competing explanations
For the routine example, the problem could be too many steps, unclear instructions or uncertainty about which product comes first. Each explanation suggests a different change. A shorter routine, a sequence diagram and clearer bundle naming are different interventions.
Ask AI to retain contradictory comments and examples that do not fit the dominant theme. A positive review explaining why the multi-step process feels useful may identify a segment the proposed simplification would disappoint.
The resulting brief can be short enough to discuss in one meeting. Here is an illustrative version:
First-time bundle buyers appear to misunderstand sequence. We do not yet know whether this affects visitors before purchase. Test a simple explanation near the bundle selector, then examine purchase behaviour alongside subsequent clarification requests.
That is more actionable than “improve messaging.” It specifies a group, an information gap and a result the team can examine. It also avoids claiming that an internal theme is already a population estimate.
Check the explanation with customers
Asking “Would clearer instructions make this easier?” steers the conversation towards the proposed solution. Ask customers to describe what they expected, where they hesitated and how they resolved it. When using structured questions, test whether respondents understand them as intended.
Pew Research Center's questionnaire guidance explains how wording and context can affect responses. The implication for growth research is to examine the problem before making customers evaluate the team's preferred answer.
After the change, review the results and return to the original statements. If confusion persists, the team should be able to reconsider the explanation. AI is useful when it reduces the effort of that loop, rather than turning the first theme into a permanent truth.
Frontier's AI Growth approach connects customer signals to prioritised work. The final asset is a better question and an accountable test, with the customer's meaning still intact.