Skip to content
Tutorials Center
My learning

Topics

Lesson 10 of 50 ยท Requirements Engineering

Gathering Requirements

Learn how to find stakeholders and elicit requirements with interviews, surveys, observation and workshops, and how to avoid the classic user-research pitfalls.

A requirement is a statement of something the software must do or be. Before you can design or build anything, someone has to find out what those statements are. This activity is called requirements elicitation (or gathering), and it is harder than it sounds: people often do not know what they want, cannot describe what they do every day, disagree with each other, or ask for a solution when they have a problem. A large share of project failures traces back to requirements that were missing, wrong or misunderstood, which is why this chapter of the tutorial exists.

Elicitation is a human activity, so the real techniques here are conversations, observations and workshops. The runnable examples are JavaScript stand-ins that show how collected input can be organized and checked; they cannot replace talking to people.

Who are the stakeholders?

A stakeholder is anyone affected by the system or able to affect it. Beginners list only "the users". A better list includes:

  • End users, who work with the software daily. They know the real process, including its workarounds.
  • Customers or sponsors, who pay for it and care about cost, deadlines and business results.
  • Operators and support staff, who run, monitor and repair it.
  • Domain experts, who know the rules (an accountant, a pharmacist, a pilot).
  • Regulators and legal teams, whose rules are not negotiable.
  • Integrators, whose systems must exchange data with yours.
  • Developers and testers, whose constraints shape what is feasible.
  • Opponents, people who lose something when the system arrives (a department whose manual work is automated). Ignoring them causes quiet sabotage.

A stakeholder map with two axes, influence (how much they can change the project) and interest (how much they care), helps you decide how to involve each person: high on both means working closely; high influence but low interest means keeping them informed; low influence but high interest means listening and updating them.

Beyond "the users"

End users They know the real workflow.
Sponsors They fund it and want results.
Operators They keep it running.
Regulators Their rules are fixed.

Missing one group is how surprises reach production.

1 / 2

Elicitation techniques

Interviews. One conversation with one stakeholder is the most flexible technique. Prepare open questions ("Walk me through the last time you processed a refund"), listen more than you talk, ask "why" a few times to move from a requested feature to the underlying need, and write down exact words. Interview several people in the same role, because individuals differ.

Surveys and questionnaires. Useful when you need numbers from many people (how often do you use feature X?). They are cheap to distribute but shallow: you only get answers to the questions you thought of, and response rates are often low. Use them to confirm what interviews suggested, not to discover it.

Observation (ethnography). Watch people do their real work, ideally in their real setting. People forget steps that have become automatic, and they often describe the official process rather than the actual one. Observation finds the sticky notes on the monitor, the spreadsheet on the side, the phone call that "everybody just knows" to make.

Workshops. A facilitated meeting with several stakeholders in one room. Good for agreeing on priorities and resolving conflicts, because disagreements surface and are settled on the spot. A skilled facilitator keeps loud voices from dominating.

Document and system analysis. Read existing manuals, forms, tickets, logs and the old system itself. Support tickets show what annoys users right now.

Prototypes and storyboards. Show something concrete. People who cannot say what they want can usually say what is wrong with a sketch. See prototyping for the cost trade-off.

Analytics and user research. For a product with live users, usage data tells you what people do; interviews tell you why. You need both.

Interview questions that work and questions that do not

A good question is open, neutral and about specific past behavior. Compare:

Weak question Why it is weak Better question
"Wouldn't a dashboard be great?" Leading: it invites "yes" "How do you check progress today?"
"Do you like the current system?" Closed and vague "What was the most frustrating moment with it last week?"
"Would you use a mobile app?" Hypothetical, so people overstate "Where are you when you need this information?"
"What features do you want?" Asks for solutions "What takes you longest in your day?"

This stand-in is a rough checker for interview scripts. It flags questions that begin like closed questions, that lead the witness, or that ask about the future. It is only a keyword heuristic and will miss subtle problems, but it makes the patterns easy to see.

JS Check interview questions

Press Run (or Ctrl + Enter) to see the console output here.

No output.

Two of the six questions pass. Rewrite the others as open questions about real, recent events and run the check again.

Merging what people tell you

After a round of interviews you hold several lists, one per stakeholder, full of overlaps ("export to Excel" and "download spreadsheet"), unique items and outright conflicts (one person wants approvals required, another wants them removed). Before writing formal requirements, consolidate:

  1. Normalize wording so duplicates match.
  2. Count and attribute: who asked for it, and how many people in which roles.
  3. Flag conflicts explicitly and resolve them in a workshop or by asking the sponsor; never silently pick a side.
  4. Record the source of every item, so you can ask follow-up questions and later trace each requirement to its origin (traceability).

The stand-in below merges four invented input lists. A small table maps different phrasings to one canonical need, and a second table lists known conflicts.

JS Merge stakeholder input

Press Run (or Ctrl + Enter) to see the console output here.

No output.

One need was requested by three people and three more by two each, and the conflict line shows Manager and Auditor on one side and Support on the other. The code cannot decide who is right; that is a business decision for the sponsor, ideally made in a workshop and written down with the reason.

Pitfalls in user research

  • Solutions disguised as requirements. "We need a dropdown" is a design; "I must pick one of four shipping methods quickly" is the need. Dig for the need.
  • Hypothetical bias. People predict their own behavior badly. Ask what they did last time, not what they would do.
  • Only the loudest voice. A senior person or a confident talker can crowd out quieter users who do the real work.
  • Confirmation bias. If you already believe in a feature, you will hear support for it everywhere. Write your assumptions down and try to disprove them.
  • Sampling error. Interviewing five power users misses the casual ones; interviewing only existing customers misses the people who left.
  • Stale knowledge. Requirements change as the business changes. Eliciting once at the start and never again is a waterfall habit (waterfall model); agile teams keep talking to users every iteration.
  • Tacit knowledge. The most important rules are often unspoken. Observation and prototypes expose them.
  • Politics and privacy. Recording interviews and sharing quotes needs consent; never expose what an individual said to their manager.

Whatever the technique, close the loop: show people what you understood and ask them to correct it. A short summary sent after each interview ("here is what I heard") catches misunderstandings when they are cheap.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Which is the best interview question for discovering how people really work?

    It is open, neutral and about a specific past event. The other questions are leading, closed or hypothetical.
  2. 2. Why is observation valuable in addition to interviews?

    Watching real work reveals workarounds, unwritten rules and steps that people no longer notice they perform.
  3. 3. Two stakeholders give conflicting requirements. What should you do?

    Conflicts are business decisions. Resolve them openly and record who decided and why.
  4. 4. Which statement is a solution disguised as a requirement?

    A dropdown is a design choice. The underlying need is choosing a shipping method quickly, and several designs could satisfy it.

Summary

  • A stakeholder is anyone who affects or is affected by the system; list more than the end users, including operators, regulators and opponents.
  • Combine techniques: interviews for depth, surveys for breadth, observation for the real process, workshops for conflicts, prototypes for concrete feedback.
  • Ask open, neutral questions about specific past events; avoid leading, closed and hypothetical questions.
  • Merge input from all stakeholders, keep the source of each item, and resolve conflicts openly.
  • Beware solutions disguised as needs, loud voices, biased samples and tacit knowledge, and always confirm what you understood.

Next: Functional and Non-Functional Requirements.