Elicitation When Nobody Has Time to Talk to You

A case built with zero stakeholder minutes: what public sources settled, and six open questions sorted by whether silence can answer them.

Numbers in this note carry source tags ([POLICY], [THREADS n=10], [MINE]). What the tags mean.

This piece stands on the returns case published on this site: damaged-delivery refunds at a mid-size online retailer, built only from public sources - a published returns policy, help centre articles, ten public complaint threads I selected myself, consumer law and a walk through the published customer journey. No client, no engagement, no interviews. Nobody inside that company gave the case a single minute, so it is the extreme version of the problem in the title: not a stakeholder who is busy, but one who does not exist. The sorting and the message template below are mine [MINE]. You can repeat both from the published case and disagree line by line.

Elicitation without stakeholder time is the practice of settling everything that can be settled from sources that do not need a person - published rules, records of what customers already said, and the process itself walked from the outside - and then turning what is left into written proposals with a stated default, so that the few minutes a stakeholder does give go only to the questions that silence cannot answer.

In short

• Most of what an interview would collect can be read first. The case built eight numbered observations from four kinds of source, with nobody's time spent on any of them.
• What reading cannot settle becomes an open question with an owner, a date and a fallback. The fallback is what happens if nobody answers, which makes it the default you can send instead of a meeting request.
• Not every default is safe to adopt by silence. Sort each question by what its fallback changes: nothing a customer can read, a document someone can send, or a public promise.
• In the case, six open questions and four asks from the one-page summary collapse into one conversation with one role about two questions, plus one person who has to be in the room before anything ships.

What can you learn without anybody's time?

More than the request for a workshop suggests. Wikipedia's article on requirements elicitation defines it as "the practice of researching and discovering the requirements of a system from users, customers, and other stakeholders", and lists the usual practices: "interviews, questionnaires, user observation, workshops, brainstorming, use cases, role playing and prototyping". Every item on that list needs somebody else's time. Reading what is already written does not, and it is not on the list.

Social science has a name for evidence gathered without asking anybody. The article on unobtrusive research credits the term to Webb, Campbell, Schwartz and Sechrest, whose 1966 book "described methodologies which do not involve direct elicitation of data from the research subjects". A complaint thread is that kind of evidence. It was written for other customers, not for an analyst, so nobody shaped it to please me. The selection was still mine, which is why the case keeps n=10 inside the tag and never turns the count into a rate.

Here is what the case settled with zero stakeholder minutes, by source.

Eight observations: two from the walk, three from the help centre, two from the threads, one from the policy. The source table in section 1.1 of the case adds consumer law as a fifth source, which supplies constraints rather than observations. The case sheet puts the whole case at about 13 hours of the author's time, spread over four evenings and one Saturday.

The right-hand column is the useful one. It is the list of things a person would have to tell you, and the case did not let it stay a vague feeling. Every line in it became an open question.

When nobody answers, what does your document do?

The case carries six open questions in section 8, each with an owner, a due date and a fallback. The register's own wording: "The fallback is the part people leave out, and without it an open question is a wish with a due date attached."

Read the fallback column again with the title of this piece in mind. A fallback is what the document does when the owner stays silent. That means it is also a proposal you can put in writing: "unless you tell me otherwise by this date, here is what happens". That message costs the reader a yes or a no, not a slot in their calendar.

But a default is only safe to adopt by silence when silence cannot do harm. So the sorting question for each open question is not "who should I interview?" but "what does the fallback change if nobody replies?" I used four bins [MINE]:

• Silence settles it. The fallback removes or freezes something, and nothing a customer can read or be paid changes.
• A document settles it. The answer is a record or a written opinion, not a conversation. You need a file, not a meeting.
• No addressee. Nobody can be asked yet, because the owner slot is empty. The fallback holds until a name exists.
• A person must say it. The fallback changes a public promise. Silence may not do that on anybody's behalf.

Six questions: one settled by silence, two by documents, one with no addressee, two that need a person. The two that need a person have the same owner. That is one conversation, not six.

The one-page summary in section 7 makes four asks of the Head of Customer Care by 30 April 2026. Three of them are OQ-02, OQ-05 and OQ-04. The fourth is different: "Put the goods-in lead in the room before anything ships." That one has no fallback, and on purpose. The case's to-be moves the condition dispute to after the payout, REQ-017 is the goods-in lead's way back, and the stakeholder map calls that role "the row nobody invites". A written default sent to the person who loses the most from the change would be the shortcut the whole case argues against.

So the live time the case needs is two items: one conversation about two questions, and one person who has to see the design before it ships. Everything else can travel as a message with a date on it.

Where does the time actually go?

Into the author's hours instead of the stakeholder's. The case sheet puts the largest single block of those 13 hours into the process model, "and most of that was one full redraw". And the most important line in the as-is, the parked branch, did not come from the reading at all. The first version had no parked branch; it arrived through an objection from a peer reviewer, logged as RL-01. The two reviewers were peers reading a document as a favour, not stakeholders, and the case says so.

That matters for anyone tempted by this method as a way to skip people altogether. Reading settles facts that are written down. It does not settle decisions nobody has made yet, and it does not catch what you failed to see. The first is what the "person must say it" bin is for. The second is what a reviewer is for.

The default-on-silence message, filled in

One message per question in the first two bins. This is the one for OQ-05, written as I would send it if the case had an addressee. It has not been sent, because there is nobody to send it to [MINE].

The last row is the one people skip. When silence applies a default, the register records that it was silence, and on which date. A default nobody confirmed and nobody recorded turns into "we agreed" within a month.

The default-on-silence message, blank

Before you send it. Ask whether the fallback changes anything a customer can read, anything that moves money, or anything a person with a veto would want to see first. If it does, it is not a message. It goes on the short list for the one conversation you will get.

Where does this method stop?

It depends on a process that leaves public traces. A returns policy, a help centre and complaint threads exist for this case. An internal approval chain with no customer-facing side leaves nothing to read, and the step "pick a process you can watch from the outside" in the case's own method exists because of that limit.

It also leans on people reading their messages. A default sent into an inbox nobody opens is not consent, which is why the "person must say it" bin exists and why the register records silence as silence.

What it reliably does is shrink the ask. A request for a workshop competes with everything else in a stakeholder's week. Two questions and one person in the room is a much smaller request. The questions you would bring to that conversation, and what each answer changes, are in the eight questions in a first stakeholder interview. If you do get a room, a requirements workshop that ends with a document builds the agenda from the cells that need two decision rights at once. The tags used above are explained on the method page, and checks 05 and 25 in the scorecard - open questions with an owner and a date, and a page that says what happens if nothing is decided - are the two this piece leans on. The full case is at /case.

All field notes · The portfolio guide · More in The Case File