1. Courses
  2. Business & Finance
  3. Business Analysis
  4. Stop 2

35 minBeginner

Stakeholders and elicitation

Before you write a single requirement, you need to know who cares about the change and what is really happening today. This lesson shows you how to map stakeholders, choose the right elicitation technique, run a workshop, and confirm what you heard, using the search for why 1,500 partner orders a week still come by phone.

By the end you can

  • Build a stakeholder list, place it on a power and interest grid, and turn it into an engagement plan with a RACI for decisions
  • Pick between interviews, observation, document analysis, workshops and surveys for a given question
  • Plan and facilitate a workshop that first collects ideas, then narrows them
  • Confirm findings by playing them back and triangulating what people say, what they do and what the data shows

The everyday picture

A good car mechanic does three things before touching a spanner. She asks the driver about the last time the noise happened: on the motorway, when braking, on a cold morning? She takes the car for a short drive and listens herself. And she plugs in the diagnostic reader to see the stored error codes. If the driver says "brakes", the drive says "front left wheel" and the codes say "wheel sensor", she trusts the point where the three agree.

She also knows the driver is not the only person with a stake. The owner pays the bill, the insurer may pay part of it, and the driver's partner uses the car every weekend and hears noises the driver never mentions. Elicitation is the same job. You ask, you watch, you read the records, you talk to everyone who uses or pays for the thing, and you believe the findings that hold up from more than one direction.

Who has a stake, and how much attention each needs

In What business analysts do you met stakeholders as one of the six core concepts. A is anyone who affects the change or is affected by it. Users, customers and stakeholders taught the basic map. As the business analyst you use that map for one more job: deciding who you elicit from, and how.

Start with a list, and find people with three questions. Who does the work today? Who decides, approves or pays? Who loses something if this succeeds? For "Phone orders into the app" the list includes shop owners like Ravi, the 40 sales reps, the head of partner sales (the sponsor), Imran as product manager, finance credit control, DC operations, the 120 drivers, the support team and Lagoon Digital, the agency that builds the app.

Then place each one on a . The corner tells you how much contact to plan: manage closely, keep satisfied, keep informed or monitor. Two cautions. The grid tells you how much to communicate, not whom to learn from; shop owners have little power one by one, yet their behaviour decides whether the project works, so you interview and observe them. And positions move. Credit control has little interest today, but the moment you propose an in-app credit increase, they own the decision.

Keep satisfied Finance director DC operations Credit control Manage closely Head of sales Imran (PM) 40 sales reps Lagoon Digital Credit control Monitor Drivers (120) Marketing Keep informed Shop owners like Ravi (1,800) Support team still elicit: users power interest
The grid sets how much contact each group needs. Credit control moves from keep satisfied to manage closely once the project proposes in-app credit increases. Shop owners sit in keep informed, yet they are the users you must interview and observe.

Turn the grid into an : for each person or group, what you need from them, the technique, how often, and who owns the contact. Then add a for the decisions you can already see coming. BABOK lists the RACI matrix with stakeholder maps. The rule that matters is one Accountable person per row; Stakeholders, negotiation and leading without authority shows what goes wrong with two.

DecisionResponsibleAccountableConsultedInformed
Which reasons for phoning we fix firstYou (BA)Head of partner salesImran, reps, shop owners, LagoonDC operations, support
Shops can ask for a one-off credit increase in the appYou and ImranCredit control managerHead of partner sales, LagoonReps, support
Rep bonus counts app orders from their shopsYou (the case and numbers)Head of partner salesFinance director, reps, ImranSupport

Quick check

The credit control manager skips your project meetings, but nothing about credit can go into the app without her yes. What should your engagement plan do?

Choosing the technique for the question

is more than asking. Each technique answers a different kind of question, so start from the question and pick the technique, never the other way round.

TechniqueBest forWatch out for
InterviewWhy people act as they do; stories of real eventsLeading questions; people's guesses about the future
Observation (job shadowing)What people actually do, how long it takes, the workaroundsPeople work "properly" while watched; one day may be unusual
Document analysisFacts at scale from records: logs, credit notes, rulesRecords show what was written down, not why
WorkshopShared understanding and agreement across groupsThe loudest or most senior voice wins
SurveyHow common something is across many peopleOnly some people answer; you can't ask "why" twice

Interviews. User interviews that don't lie taught the story-based interview: ask about the last time, not about opinions or the future. As a BA you add one thing: you come with a list of what you must learn for the requirements (pack sizes, credit, timing), and you check each item off without leading. "Tell me about your last order" is open. "Was the app too slow?" puts the answer in their mouth.

Observation. In Watching the work you ran a field visit. For a BA, is where you get times and steps you can model later. People can't tell you that re-keying takes ten minutes; they think it takes "a few".

Document analysis. Before you meet anyone, read what already exists: the reps' call logs and chat threads, correction credit notes, the bonus scheme, the app's error log. It costs no one's time and gives you numbers to test interview claims against. Tomás can pull the data, but you decide which questions it must answer.

Surveys give breadth once you know what to ask. Use them to count, never to discover: a survey can tell you how many shops lose mobile data every day, but only if an interview already told you data is a problem.

You also need the right people. Picking only the friendly shops near the office gives you a tidy and wrong picture. Use data to choose: the shops that phone most, a few mixed shops and a few app-only shops, spread over the four DCs. App-only shops tell you what already works. The practice task builds this list with SQL.

Quick check

You need to know how long a rep spends turning one phone order into a back-office order. Which technique gives the most reliable answer?

Running a workshop that does not just confirm the boss

A brings different groups into one room to build a shared picture and agree what matters. It is powerful because reps hear shop owners and finance hears both. It is risky because a group follows its most senior or loudest member.

Prepare four things. A goal written as an output: "a ranked list of the reasons shops phone, with evidence". An agenda with times, sent with a short pre-read a few days before. Ground rules you read out at the start: stories before opinions, one conversation at a time, no solutions until the last part. And two roles: the (you) runs the process and stays neutral; the records what people say in their own words.

Then run it in two phases. First diverge: get many ideas out before anyone judges them. Ask everyone to write one reason per card, in silence, for a few minutes. Silent writing stops the first speaker from setting the agenda, and it gives the quiet rep and the nervous shop owner equal space. Then converge: put the cards up, cluster duplicates, name each cluster, and let people vote with dots for the clusters that matter most. Facilitation writers such as Sam Kaner describe the uncomfortable middle between the two as normal. Don't rush it.

Ask the sponsor to join for the last fifteen minutes, when the group presents its result, rather than sitting in from the start. Reps will not say "my bonus counts orders I key in" with the head of sales watching.

Think first: dot voting puts "credit" top, but the reps in the room outnumber the shop owners two to one. Is the vote the answer?

Not on its own. A vote shows what this room believes, and this room is weighted toward reps. Treat the ranking as a hypothesis and check it: how many phone orders really go over the limit, and what did the shop owners you interviewed say? A workshop result is one source, to triangulate like any other.

Confirming what you heard

BABOK's Elicitation and Collaboration knowledge area has four steps you will repeat on every project: prepare, conduct the elicitation, confirm the results, and communicate them. Working with stakeholders runs alongside all four. The step people skip is the third.

1. Prepare goal, who, technique plan, guide 2. Elicit ask, watch, read raw notes 3. Confirm play back, compare gaps or conflicts confirmed 4. Communicate summary to all ongoing: collaboration
Elicitation is a loop. Confirming the results finds gaps and conflicts, which send you back for another round before anything is communicated as a finding.

Confirm in two ways. Play back: within a day or two, send each person a short summary of what you heard and ask what you got wrong. People correct a written summary far more readily than they correct you in a meeting. Triangulate: means checking each finding against at least two kinds of source: what people say (interviews, workshop), what they do (observation), and what the records show (data and documents). A finding seen in all three goes into the requirements. A finding from one source is a hypothesis to test.

Three biases do most of the damage. The loudest voice: the most senior or confident person's view becomes "what stakeholders want". Leading questions: you get your own idea back, politely confirmed. Solution-jumping: someone says "we need a WhatsApp ordering bot" and the session turns into designing it. Note the idea, then ask "what would that fix for you?" and go back to the problem. Watch for a fourth in your sample: a survey answered mostly by app users says little about phone users.

Quick check

In interviews, reps say most corrections are DC picking mistakes. The credit notes Tomás pulled say about half are case-versus-unit mix-ups. What do you do next?

Do it

Plan and run one round of elicitation for a change in your own work, using the Mangrove templates.

  1. List and map the stakeholders

    Use the three questions (who does the work, who decides or pays, who loses). Place everyone on the grid, then write the engagement plan and a RACI for the decisions you can see.

    Markdown
    | Stakeholder        | Grid           | What we need from them        | Technique           | How often   | Owner |
|--------------------|----------------|-------------------------------|---------------------|-------------|-------|
| Head of partner sales | Manage closely | Goal, bonus decision, sign-off | 1:1 meeting       | Every 2 wks | BA    |
| Sales reps (40)    | Manage closely | How phone orders work today   | Shadow 2, interview 6, workshop | Weeks 1-3 | BA |
| Shop owners        | Keep informed  | Why they phone; what works in the app | Interview 12, survey | Weeks 1-3 | BA |
| Credit control     | Manage closely | Credit rules; in-app increase | 1:1 meeting         | Weeks 2 and 4 | BA + Imran |

  2. Read the records first

    Ask Tomás for four weeks of credit notes and phone orders by shop. Write down the claims the data must test, and pick interviewees with the query in the practice task.

  3. Write the interview guide

    Thirty minutes, one or two real stories, your must-learn list at the bottom to tick off without leading.

    Markdown
    # Interview guide: partner shop owner (30 min)
Intro (2 min): who I am, why, notes OK? Nothing you say is shared with your name.
Story (20 min)
- Tell me about the last time you ordered stock. How did it start?
- What happened next? Who did you talk to? How long did it take?
- Was anything wrong when it arrived? What did you do about it?
- Tell me about a time you used the app, or tried to. What happened?
Close (5 min): anything I should have asked? May I visit during an order?
Must learn (tick, do not ask directly):
[ ] pack size (case or unit)  [ ] credit limit hit  [ ] items not in app
[ ] who starts the order      [ ] mobile data at the shop

  4. Shadow before the workshop

    A day with each of two reps from different DCs. Time each order from first call to keyed in the back office, and note every paper, chat and phone call it touches.

  5. Run the workshop

    Eight to ten people: reps from all four DCs, shop owners from both the heavy-phone and app-only lists, one person from credit control. Tomás scribes and brings the data.

    Markdown
    # Workshop: why partner shops still phone (90 min)
Goal: a ranked list of reasons, each with its evidence
0:00 Welcome, goal, ground rules, roles (facilitator, scribe)   10
0:10 Walk through one real phone order, end to end, on the wall  15
0:25 Silent writing: one reason per card, from real orders        10
0:35 Share and cluster; name each cluster                         20
0:55 Break                                                         5
1:00 Dot vote (3 dots each); compare with the data                15
1:15 Top 3: what would have to be true for you to stop phoning?   10
1:25 Sponsor joins: group presents result; next steps, playback    5
Ground rules: stories before opinions; one conversation at a time;
no solutions before 1:15; bonus comments stay anonymous in notes.

  6. Confirm and communicate

    Play back each interview within two days. Put every finding in a said / seen / records table like the one above. Send a one-page summary to everyone you elicited from, with what happens next.

Bring it to your team

Ask these before the first interview is booked.

  • Who loses something if this works, and have we talked to them?
  • For each decision we can see coming, who is the one Accountable person?
  • How did we pick interviewees, and does the list include people who already use the app?
  • Which of our findings come from only one kind of source?
  • Who will see the summary before it goes to the sponsor, and have they had a chance to correct it?

Red flags

  • Every interviewee was suggested by the sponsor.
  • The workshop opened with a slide of the planned solution.
  • Requirements cite "the reps said" with no data or observation behind them.
  • Nobody has spoken to finance, yet credit appears in the findings.

Practice

Pick the heavy phone users to interview

The tables are partner_shop(id, name, dc) and partner_order(id, shop_id, channel, placed_on), where channel is 'app' or 'phone'. Look only at the last four weeks: placed_on from 2026-09-21 to 2026-10-18, both days included. Return shops with at least 3 phone orders in that window, with shop_id, name, dc, phone_orders, total_orders and phone_pct (phone share of their orders, rounded to a whole percent). Sort by phone_pct, highest first, then by phone_orders, highest first.

SQL exerciseRuns in a real PostgreSQL inside your browser
-- Shops with 3+ phone orders in the last 4 weeks, highest phone share first
SELECT s.id AS shop_id, s.name, s.dc, count(*) AS total_orders
FROM partner_shop s
JOIN partner_order o ON o.shop_id = s.id
GROUP BY s.id, s.name, s.dc
ORDER BY s.id;
Ctrl + Enter runs
Show a solution
SQL
SELECT s.id AS shop_id, s.name, s.dc,
       count(*) FILTER (WHERE o.channel = 'phone') AS phone_orders,
       count(*) AS total_orders,
       round(100.0 * count(*) FILTER (WHERE o.channel = 'phone') / count(*)) AS phone_pct
FROM partner_shop s
JOIN partner_order o ON o.shop_id = s.id
WHERE o.placed_on >= DATE '2026-09-21'
  AND o.placed_on < DATE '2026-10-19'          -- includes all of 18 October
GROUP BY s.id, s.name, s.dc
HAVING count(*) FILTER (WHERE o.channel = 'phone') >= 3
ORDER BY phone_pct DESC, phone_orders DESC;

Hilltop Provisions has 4 phone orders in all, but only 2 in the window, so the date filter must come before the count. For the app-only shops to interview as well, change the HAVING line to keep shops with 0 phone orders: here that returns Green Gate Store and Market Lane Foods, from two different DCs.

Pass the exercise to finish this task.

Fix five interview questions

A colleague drafted these questions for reps. Rewrite each one so it is open, about a real past event, and does not suggest the answer. (1) "Don't you think shops find the app too slow?" (2) "Would shops use a reorder button?" (3) "How often do you get case and unit wrong?" (4) "Is your bonus the reason you prefer phone orders?" (5) "What features should the app have?"

  • Every rewrite asks about a specific past event ("the last time…"), not an opinion or a future guess
  • None of the rewrites names the answer you suspect (slow, case and unit, bonus)
  • Question 4 is rebuilt so a rep can raise incentives without feeling accused, or moved to document analysis
  • Question 5 becomes a question about a time the app could not do what was needed
Show a model answer

(1) "Tell me about the last order a shop phoned in. How did the call start?" (2) "Walk me through the last order you took from a shop that orders the same things each week." (3) "Tell me about the last order that was corrected after delivery. What happened?" (4) "How is your month judged? Walk me through how last month's numbers were worked out." Check the bonus scheme document as well. (5) "Tell me about the last time a shop said the app couldn't do something. What did you do?"

Check yourself

The head of partner sales gives you a list of eight shops to interview, all friendly and all near the North DC. What is the best response?

Ten minutes into your workshop, a senior rep says "the real fix is a WhatsApp ordering bot" and the room starts designing it. What should you, the facilitator, do?

An interview says shops phone because of patchy data, the survey says about 21% lose data most days, and on your one visit the app worked fine. Which statement is best?

Flashcards

Open my review

Key takeaways

  • Map stakeholders with three questions and a power and interest grid, then write an engagement plan and a RACI with one Accountable person per decision.
  • Pick the technique from the question: interviews for why, observation for what and how long, documents for scale, workshops for agreement, surveys for how many.
  • Run workshops with a written goal, ground rules, a facilitator and a scribe; diverge in silence before you converge.
  • Play back notes and triangulate. At Mangrove, five reasons held up, and two of them (credit and rep bonuses) are not screen problems at all.

Sources and further reading

Episode transcriptFive reasons hiding behind the phone · about 11 minutes · tap any line to play from there
  1. NadiaImran, when the head of partner sales started this project, what did the brief say?
  2. ImranMake shops use the app. About fifteen hundred partner orders a week still come in by phone to forty sales reps. She wants that number down.
  3. NadiaAnd if a new analyst took that brief straight to Lagoon Digital and asked for nicer screens?
  4. ImranThey'd build nicer screens. And I suspect the phones would keep ringing.
  5. NadiaThey would. Two of the real reasons shops phone have nothing to do with screens. Today is about how you find reasons like that. Who to talk to, how to ask, what to watch, and how to know you heard it right.
  6. NadiaStart with a car mechanic. Before she touches a spanner, she asks the driver about the last time the noise happened. Then she drives the car herself. Then she plugs in the reader and looks at the stored error codes.
  7. ImranSo the driver says brakes, the drive says front left wheel, and the codes say wheel sensor.
  8. NadiaAnd she trusts the point where all three agree. She also knows the driver isn't the only one with a stake. The owner pays, and the partner who drives it every weekend hears noises the driver never mentions.
  9. ImranAsk, watch, read the records, and talk to everyone who uses it or pays for it.
  10. NadiaThat's elicitation. And it starts with the people. A stakeholder is anyone who affects the change or is affected by it. Three questions find them. Who does the work today? Who decides, approves or pays? And who loses something if this succeeds?
  11. ImranFor us that's shop owners like Ravi, the forty reps, the head of partner sales as sponsor, me, finance credit control, DC operations, the drivers, support, and Lagoon.
  12. NadiaThen you place them on the power and interest grid. If you can see the screen, the first diagram has four boxes. If not, picture a square cut into four. High power and high interest is manage closely. High power, low interest, keep satisfied. Low power, high interest, keep informed. Low on both, monitor.
  13. ImranI'm in manage closely, with the sponsor, the reps and Lagoon. Drivers and marketing are in monitor.
  14. NadiaAnd there's an arrow in the diagram. Credit control starts in keep satisfied. They barely care about the project. But the moment you propose letting shops ask for a credit increase in the app, they own that decision.
  15. ImranSo they jump to manage closely. What about Ravi? Shop owners are in keep informed. Low power, one by one.
  16. NadiaThat's the trap. The grid tells you how much to communicate. It doesn't tell you whom to learn from. Ravi's behaviour decides whether this project works. You interview him and you watch him order.
  17. ImranAnd decisions? I've been burned by meetings where three people thought they were deciding.
  18. NadiaThat's what a RACI is for. Responsible, Accountable, Consulted, Informed, for each decision you can see coming. The rule that matters is one Accountable person per row. Credit control's manager is accountable for the credit increase. Not the sponsor, not you.
  19. ImranOkay. Techniques. I know interviews. Ask about the last time, not opinions.
  20. NadiaRight, and as an analyst you add a must-learn list. Pack sizes, credit, timing. You tick them off as they come up, without asking leading questions. Was the app too slow puts the answer in their mouth.
  21. ImranSo if I want to know how long reps spend re-keying an order, I just ask them?
  22. NadiaThat's the guess I hoped you'd make. People estimate their own task times badly. They'll say a few minutes. You shadow two reps for a day and time it. It's about ten minutes an order.
  23. ImranSo observation for what people do and how long it takes. Interviews for why.
  24. NadiaAnd documents for scale. Read the call logs, the correction credit notes, the bonus scheme, before you meet anyone. It costs nobody's time and gives you numbers to test every claim against. Surveys come last, to count how common something is once you know what to ask.
  25. ImranAnd who I interview matters too. If the sponsor hands me eight friendly shops near the North DC?
  26. NadiaThank her, keep a couple, and pick the rest from data. The heaviest phone users, some mixed shops, and a few app-only shops, across all four DCs. App-only shops tell you what already works.
  27. ImranThat's the practice task. A query that finds shops with at least three phone orders in the last four weeks and their phone share, highest first.
  28. NadiaAnd watch the date window. One shop has four phone orders in total but only two inside the four weeks. Filter by date before you count.
  29. ImranNow the workshop. Reps, shop owners and someone from credit control in one room. What goes wrong?
  30. NadiaThe room follows its loudest or most senior voice. So you prepare a goal written as an output, a ranked list of reasons with evidence. A timed agenda. Ground rules read out at the start. And two roles, a facilitator who stays neutral, and a scribe who writes what people say in their own words.
  31. ImranAnd the run itself?
  32. NadiaTwo phases. Diverge first. Everyone writes reasons on cards, one per card, in silence. That stops the first speaker setting the agenda, and the quiet rep gets the same space as the confident one. Then converge. Cluster the cards, name the clusters, and vote with dots.
  33. ImranShould the sponsor be there?
  34. NadiaFor the last fifteen minutes, when the group presents. No rep will say my bonus counts orders I key in with the head of sales watching.
  35. ImranAnd if someone says the real fix is a WhatsApp bot, and the whole room starts designing it?
  36. NadiaPark it on the board, ask what problem it would fix for shops, and bring the group back to reasons. That's solution-jumping, one of the three big biases, with the loudest voice and leading questions.
  37. NadiaThen comes the step people skip. The second diagram shows the cycle from the BABOK guide. Prepare, elicit, confirm, communicate, top to bottom, with a dashed loop from confirm back up to elicit, marked gaps or conflicts.
  38. ImranSo confirming isn't just reading my notes again.
  39. NadiaTwo things. Play back, meaning you send each person a short summary within a day or two and ask what you got wrong. And triangulate. Check each finding against what people say, what they do, and what the records show. Seen in all three, it goes into requirements. Seen in one, it's a hypothesis.
  40. ImranSo what did we actually find? Twelve shop owners, six reps, two shadowing days, two hundred credit notes, a workshop and a survey.
  41. NadiaFive reasons held up. One, speed. Ravi reorders mostly the same thirty things every week, and a three-minute call is easier than about forty taps while serving customers.
  42. ImranTwo, credit. When an order goes over the limit, the app says payment declined, and the owner phones the rep, who can ask finance for a one-off extension. About one phone order in six waits for a finance approval.
  43. NadiaThree, missing items and promotions. Reps sell DC-only promotions and substitutes the app doesn't show. Four, rep incentives. Three of six reps said privately that phone orders count for their bonus, and the bonus scheme confirms it.
  44. ImranAnd five, patchy mobile data. The survey says eighty-eight of four hundred and twelve replies, about twenty-one percent, lose data most days.
  45. NadiaWhich is probably low. The shops with the worst data are the least likely to answer a survey sent through the app.
  46. ImranAnd the credit notes?
  47. NadiaHere's a conflict we had to resolve. Reps said most corrections are DC picking mistakes. But a hundred and four of the two hundred notes, fifty-two percent, were case versus unit mix-ups.
  48. ImranSo who wins?
  49. NadiaNobody wins by vote. You take a few real notes to two reps and ask them to walk you through each one. Either the notes are miscoded, or the reps remember the dramatic cases. Then you know.
  50. ImranAnd credit and bonuses aren't app problems at all. That's what you meant at the start.
  51. NadiaYes. And the bonus finding is sensitive, so you report it as a pattern, never by name, and take it to the head of partner sales one to one before the summary goes out.
  52. ImranSo the Do it, in order. List and map the stakeholders, with an engagement plan and a RACI. Read the records first. Write a thirty-minute interview guide with the must-learn list at the bottom.
  53. NadiaThen shadow two reps from different DCs before the workshop. Run a ninety-minute workshop with silent writing, clustering and dot votes, and the sponsor joining at the end. Then play back, build a said, seen and records table, and send a one-page summary to everyone you talked to.
  54. ImranQuestions for my team before we book the first interview. Who loses something if this works, and have we talked to them? Who's the one accountable person for each decision?
  55. NadiaAnd which findings come from only one kind of source. That one saves projects.
  56. NadiaOne to think about. An owner tells you data is patchy, the survey says about twenty-one percent lose data most days, but on your one visit the app worked fine. Is the finding dead? Take a second.
  57. ImranNo. Two sources support it, and one quiet visit to one shop is thin. I'd plan visits to shops in the areas where people report it.
  58. NadiaExactly. Triangulation weighs sources. It doesn't let any one of them win automatically.
  59. ImranMy summary. One, map stakeholders and engage by power and interest, but learn from the users whatever their corner. Two, pick the technique from the question. Watch for what people do, ask for why, read records for scale, survey to count.
  60. ImranThree, diverge in silence before you converge, and keep the sponsor out until the end.
  61. NadiaGood. I'd add the step you didn't say. Play back and triangulate before anything is called a finding.
  62. NadiaNext time, we take those findings and draw the phone order as it really runs today, every hand-off and wait, and find the waste.
Finished reading?Mark it complete to fill in its stop on your course line.

Use it at work

Data-quality checks

Duplicates, missing values, orphans and impossible values, each as one row with a count.

Open in the SQL Lab
Next stopModelling processes
Was this lesson helpful?
Questions and discussion

Loading the discussion…