Research Integrity in Stakeholder Intelligence Work

Research integrity has become the most important and most ignored question in stakeholder intelligence. Boards and executives are receiving more research than ever, produced faster, formatted more persuasively, and grounded in less verifiable work than at any point in the last twenty years. The presentation has improved. The substance has not.

We see this directly in financial services. A chair preparing for a contested AGM, a CEO weighing a strategic shift, a board considering a sensitive appointment: each is now routinely handed material that reads as authoritative but cannot withstand a second question.

• Where did this view come from?
• Who actually said it?
• Was the person spoken to, or was their public position summarised by a model?
• Was the sample real?
• Did the analyst understand the sector well enough to know what they were hearing?

In a worrying share of cases, the honest answer is no, and the document does not say so.

The pressure points are familiar to anyone who has commissioned this kind of work. Synthetic interviews dressed up as primary research. AI summaries of analyst notes presented as independent analysis. Quotes that are real in spirit but not in attribution. Sentiment scores generated from scraped text with no understanding of who was speaking or why. Much of it is the natural consequence of cost pressure meeting capable tools meeting clients who do not ask hard questions about method. The result is that the floor of acceptable research has dropped, and the ceiling, the work that genuinely informs a difficult decision, has become harder to identify from the outside.

The implication for senior leaders is this. Use stakeholder intelligence - i.e. speaking to real people about real, current topics of interest - at exactly the moments when being wrong is most expensive: before a capital raise, before a board change, before a public position on a contested issue. A misread of how three large shareholders actually feel, as opposed to how a model thinks they feel based on their last public letter, is the difference between a managed conversation and a public reversal.

What good research integrity looks like in 2026 is not complicated, but it is uncomfortable, because it slows things down and costs more. It means naming who was spoken to, in what capacity, and when. It means distinguishing between what a stakeholder said, what they implied, and what the analyst inferred. It means disclosing where models and platforms were used and where they were not, and being specific about it. It means an analyst willing to write that the evidence is thinner than the client would like, rather than producing a confident chart to fill the gap. It means a firm willing to lose work by declining to make claims it cannot defend. These are not academic standards. They are the basic conditions under which research is worth paying for.

The practical challenge for boards and executives is to stop treating research integrity as a procurement question and start treating it as a governance one. The next time a piece of stakeholder research lands on your desk, ask three questions before you read the executive summary. Who was actually spoken to? For how long? Where in this document is the author telling me what they do not know? If the answers are unclear, the document needs to be read with caution.

We have set out our own position on this in more detail at polarinsight.com/research-integrity. It is the standard we hold ourselves to, and the one we think the people relying on this work are entitled to expect.