Understand what the research can establish

Permission to access a source does not by itself settle every use of the material in a later product. I ask the team to identify the proposed output, audience and distribution route so that qualified legal colleagues can assess the relevant arrangements. The data process then needs to carry those decisions into release controls. A source register that is disconnected from the product catalogue is difficult to use when a commercial proposal changes. I also distinguish missing permission from missing data: a fact may be available to an internal analyst without being available for the proposed external output. The practical response may be a narrower release, another permitted source or a different product. My role is to make the consequences visible and enforce the approved scope, not to invent a legal interpretation that preserves a preferred coverage number.

When experienced researchers classify the same text differently, I first ask whether the category itself contains several questions. One reviewer may be identifying a statement made by the company, while another is reacting to an allegation quoted in the document. A single label can conceal that difference. I work with research owners to separate the intended distinctions, provide examples and preserve an unresolved route where the evidence is insufficient. Agreement is useful, but forcing every reviewer to choose a category can make the number look better while reducing the honesty of the method. I want evaluation samples to include difficult cases and to explain how disagreements were resolved. The result should help users understand what a label establishes, without turning a research classification into a credit decision or a claim about facts the source does not support.

A comparison based only on organisations that remain active may describe survivors rather than the experience of the original group. I ask which businesses entered the study, which later disappeared and why evidence is missing. Closure, acquisition, changed reporting and an unavailable source can have different meanings; they should not be silently combined or assigned invented financial values. Research owners decide which comparison answers the question, supported by documented methods. My team makes the population record and exclusions available so that the published claim can be checked. Sometimes the right result is a narrower statement about continuing firms. Sometimes the original question needs a different study. I would rather explain that choice than retain an attractive headline whose population has changed without the reader being told what happened.

A careful research team can publish a responsible dataset that later becomes an overconfident slide or sales description. I work with product and commercial colleagues to identify the limitations that materially affect interpretation and the places where they must remain visible. Not every technical detail belongs in a headline, but the population, relevant source restrictions and meaning of a classification cannot disappear if their absence changes the claim. I use representative outputs to review the handover, including exports and sample reports. The aim is a usable product record, not an ever-longer disclaimer that nobody reads. I also ask who owns a later correction and how clients are informed when a material issue changes a result. Data responsibility continues after a file has passed its internal checks and moved into the commercial environment.