Professional Business Analysts / Finding out

People cannot tell you what they need, and that is not their fault

The central problem of the discipline is that requirements do not exist before somebody goes looking for them. They are constructed, and the construction is the job.

There is a persistent misunderstanding about what a business analyst does, and it is held as firmly inside the profession as outside it. The misunderstanding is that somewhere in the organisation there exists a set of requirements, held in the heads of the people who do the work, and the analyst's task is to extract them accurately and write them down. Under that model, a good analyst is a good stenographer with better questions, and failure means somebody was not listened to carefully enough.

This is not how it works. Ask a competent person to describe what they need from a new system and you will get a mixture of things: their current workaround described as a requirement, a feature they saw somewhere else, a complaint about the present arrangement, a constraint they believe is fixed but is not, and a genuine need buried among them and expressed obliquely. None of that is because they are unclear thinkers. It is because the question asks them to specify a thing that does not exist yet, in terms of a solution space they have not seen, while abstracting from a job they perform by habit rather than by rule.

The revealing experiment is one every analyst has run accidentally. Show somebody a rough version of the wrong thing and they will tell you, immediately, precisely and with confidence, exactly what is wrong with it and what it should do instead. The information was always there. It was not accessible to a question in the abstract and it became instantly accessible against a concrete artefact. That asymmetry is the single most useful fact in the discipline, and a great deal of good practice is organised around exploiting it.

It follows that requirements are not gathered, as though they were lying about waiting to be picked up. They are constructed, jointly, through a process of proposing something specific and being corrected. The analyst's contribution is not neutrality. It is the willingness to be concrete early and wrong publicly, to put a version in front of people that is definite enough to argue with, and to treat the argument as the deliverable rather than as a setback.

This reframing changes what good work looks like. An analyst who spends six weeks interviewing everybody and produces a comprehensive document has done a great deal of work and has not yet tested anything. An analyst who produces a crude model in the second week and takes it to the people who will be affected has less written down and knows considerably more, because every correction is a piece of information that would not have surfaced any other way.

It also changes what to be suspicious of. A stakeholder who agrees with everything in a requirements document is not confirming that the document is right. They are confirming that it is abstract enough to be agreeable, which is a different and much weaker signal. Agreement without friction usually means the document has not yet said anything specific enough to conflict with what somebody assumed.

The last consequence is about time. Because requirements are constructed rather than discovered, they can be constructed badly and quickly or well and quickly, but they cannot be finished in advance of contact with reality. An organisation that wants them signed off before anything is built is asking for a document, and it will receive one, and the document will be revised repeatedly under a different name once the building starts. That is not a failure of analysis. It is the predictable result of asking for certainty before the point at which it can exist.

One consequence deserves emphasis because it is where the role most often gets into difficulty. If requirements are constructed rather than discovered, then the analyst is not a neutral recorder and cannot claim to be. Every decision about which artefact to show, which stakeholder to show it to first, and which question to ask shapes what is produced. That influence is unavoidable and it is legitimate, provided it is exercised visibly. What is not legitimate is exercising it while presenting the output as though it had simply been gathered, because that removes the ability of anybody else to challenge choices they cannot see being made.