Stakeholder Matrix

Purpose

The matrix orders the stakeholder population by the two variables that decide the value of a contact, power and interest. An initiative goes off the rails when the engagement effort concentrates on a peripheral actor while the one who can stop it stays unnoticed. The grid makes that mistake visible while it is still repairable.

The decision it supports is an allocation. Who sits in the elicitation workshop and who receives the minutes, how deeply a requirement gets negotiated, how often and through which channel a stakeholder is kept informed, whose refusal shifts the schedule and whose refusal costs nothing. These trade-offs get made either way, with or without a matrix; the matrix makes them visible, arguable and revisable. It consumes the stakeholder list and does not rebuild it: if the register is incomplete, it is the register that has to be reworked.

The deliverable comes in two parts. The filled-in grid: two axes, four named quadrants, every stakeholder or stakeholder group placed in a single cell and for each quadrant the engagement strategy it imposes. And the register that goes with it: one line per stakeholder, the justification for the placement and the trigger that would move it. The grid alone reads at a glance and is hard to challenge; the register makes it transferable and re-runnable. Together they feed the engagement plan and the communication plan, which take from them the depth, the frequency and the channel per actor. The whole thing is dated, because its validity is dated too.

Usage

When to use it

• Stakeholder list established: turn it into an engagement order before planning a single session.
• Elicitation workshop to compose: decide who sits in the room and who receives the minutes.
• Politically sensitive change: identify who can block the initiative and who can accelerate it.
• More stakeholders than available time: allocate the effort according to each one's power and interest.
• Communication plan to write: calibrate depth, frequency and channel per quadrant.
• Milestone, scope change or reorganisation: re-run the placement; positions move.
• New analyst or new project manager arriving: hand over an explicit political reading of the ground.

When not to use it

• Question of how responsibilities are divided: the grid does not say who answers for what; the RACI matrix settles that activity by activity.
• Question of distance from the solution: the onion diagram places actors by proximity; it answers "who touches the solution".
• Need to understand users in order to design: the grid sets the level of engagement; personas carry knowledge of use.

Description

The two axes and their variant

Power is a stakeholder's capacity to weigh on the fate of the initiative. It is made of formal authority (a signature, a budget, a regulatory veto), of control over resources the initiative depends on (developers, a dataset, an access right) and of informal influence, which appears on no org chart and is read from the facts: who is consulted before decisions, who gets rulings outside the formal bodies, whose wording management adopts. The org chart hides one further point: a stakeholder with no individual power can acquire some, through personal relationships, by mobilising an external group (a professional association, a union) or by acting collectively with their peers. That is latent power, invisible for as long as the group has not decided to use it, and it appears in no document because it has not yet happened.

The interest carried on the axis is manifest interest: the concern the stakeholder expresses today for the outcome. How heavily the outcome will affect them is a different quantity, and the two come apart. An actor who will be heavily affected but who does not yet know the project exists is at low interest today and will stay there until the day they find out. That gap does not go on the grid; it is recorded beside it: for every actor at low interest, write down the trigger that will wake them up. The placement says where you stand; the trigger says where you are heading.

Both axes are read relative to this initiative. The medical adviser of a health insurer holds considerable power over coverage decisions and none over the choice of a hosting supplier. Defining the two axes at the start of the session, for this project and its scope, avoids half the placement disagreements.

One variant circulates and causes confusion. BABOK's figure labels its axes "influence of stakeholder" and "impact on stakeholder", that is influence against impact, while its prose describes an influence against interest grid. It is the same tool mounted on different axes: the power/impact grid replaces interest with the force with which the change hits the stakeholder, and it usually serves change management planning, when the question is "who takes the shock" rather than "who feels concerned". The reference grid is power against interest, Mendelow's. The labels on BABOK's figure designate the neighbouring variant, and a reader who reopens the standard will recognise the same instrument there. The choice of axis is declared once, written on the grid and held to the end of the exercise.

The name of the vertical axis varies with the source as well. BABOK writes influence, Mendelow writes power. Power is the right term, because the components listed fall into two families that the name of the axis has to carry together. Decision power is the ability to say no and make it stick, whether the no comes from a signature, a budget envelope, a regulatory veto or a hand on a resource the initiative cannot do without: the medical adviser runs nothing, and the de facto veto they hold over the assessment of a contested case binds as tightly as a clause in the regulations. Influence power is the capacity to move a decision one does not own.

The umbrella association of the service providers decides nothing in a health insurer's project: it signs no budget, rules on no scope, holds no veto. It can nevertheless make the insurer give up the single billing channel it meant to impose, by mobilising its members and taking the matter public. No decision power, considerable influence power. On today's grid that power stays latent: the providers sit at low power because none of them weighs alone on the design, and they rise on the day the group uses what it holds.

Naming the axis "influence" keeps only the second half of it. Executive management, whose signature releases the budget and who has never needed to convince anyone, is judged on its powers of persuasion and drops a rung. Its refusal stops the initiative all the same. The engagement effort follows the grid, so it turns away from executive management. Latent power, in turn, becomes impossible to show on the drawing: an axis named influence reads influence as it is exercised. A capacity that has not yet been used gives nothing to read. It scores zero there, however decisive it will be on the day it is spent. The axis loses what makes the most dangerous quadrant of the grid legible, the one where low individual power adds up. The word power, by contrast, also names a capacity that has not yet been used.

Interest is the concern that what is at stake gives rise to in the stakeholder: what the change, the solution or both of them reach into, their work, their data, their room for manoeuvre. The subject-matter expert whose daily work the solution takes over is therefore the most interested stakeholder on the grid, whether they want to be or not, and an actor opposed to the project is an interested actor. The engagement budget is therefore owed in proportion to what is at stake for the stakeholder. The axis carries that concern as it stands today, and a stake that has not yet given rise to concern is an exposure, which the register records as a trigger.

By convention, power occupies the vertical axis and interest the horizontal axis, increasing upwards and to the right. That is the orientation of BABOK's figure and of the classic Johnson and Scholes rendering. Some academic sources invert it, which is harmless as long as the axes are labelled on the drawing.

The four quadrants and the strategy each one imposes

The placement only counts through what it commands. Every quadrant carries an engagement strategy, and the name of the quadrant is a verb addressed to the analyst. Some renderings name the cell after its occupants instead, "key influencer" for the top-right cell: same people, same conduct.

Manage closely (high power, high interest). They can make the initiative land and derail it, and they want to know. This quadrant consumes most of the engagement budget. It is the only one where bad news is given in person and in advance: a key actor who learns of a slip from a report circulated to thirty recipients is not reacting to the slip; they are reacting to the channel.

Keep satisfied (high power, low interest). They can block and they do not follow. The strategy is asymmetric, and two symmetrical mistakes lie in wait: drowning them in reports they will not read, which makes them deaf to the message that will matter, and forgetting them until the day they discover a decision and put themselves across your path. The guard is the trigger: identify what would wake their interest and warn them before it happens.

Keep informed (low power, high interest). They are concerned first and foremost and hold few individual levers. Handled well, they are the best relays the initiative has: they know its operational detail and they carry the message to their peers. Neglected, they become the most effective source of opposition the project will meet, because weak individual power adds up: this is where collective action materialises, and a quadrant that coalesces does not slide to the right of the grid; it rises.

Monitor (low power, low interest). General communication, no targeted effort. The quadrant is held through its trigger: a stakeholder who has not yet met the project will meet it at go-live, and their interest will jump without anything having changed in the organisation.

BABOK describes the same four cells in the influence/impact vocabulary of its figure, with near-identical actions: engage the key stakeholders regularly, meet the needs of those who hold power without being concerned while working to raise their interest, solicit the supporters for their input since they can become the initiative's ambassadors, keep the rest informed through general communication that additional engagement can turn into active support. The two vocabularies overlap. A single document picks one and holds to it.

Running the session

1. Start from the stakeholder list. The matrix analyses an existing register. A session that opens with "who have we forgotten" is a list session, and it is held earlier, separately.
2. Choose the granularity. Named individuals, roles or groups: the grid supports all three, though not simultaneously and not without consequence. Naming individuals gives a usable placement and a politically flammable artefact; staying at role level gives a circulable artefact and a coarser placement. The organisation's culture settles it, and the choice is made before the session.
3. Define the two axes for this initiative, out loud, in front of the group: which power counts here, what the interest is an interest in. The placement disagreements that follow are then substantive disagreements.
4. Place as a group, one label per stakeholder. Bryson describes the method: each participant places in turn, labels move, the discussion continues until the group accepts the position. Group placement carries the whole exercise. An analyst working alone judges power on what he can see, which is to say on job titles.
5. Keep the register: one justification per placement, one trigger per low interest. The register is written in a table beside the grid, one line per stakeholder. The justification is what makes the placement arguable, transferable and re-runnable: without it, the next grid starts from zero and nobody knows what changed. The trigger is what makes the placement forward-looking: for every actor placed at low interest, write the event that will wake them, and a cell left empty says that you can see none.
6. Write the engagement strategy quadrant by quadrant, with the frequency, the channel and the depth expected. A matrix that does not produce an engagement plan is a drawing.
7. Refine if the population demands it. A finer variant carries three levels on each axis and groups the stakeholders into six categories. It serves when a quadrant is too crowded to command a single strategy. The method does not change, only the granularity of the strategies increases and the move is justified by the number of stakeholders.
8. Date the grid and fix the moments when it will be re-run: milestones, scope change, reorganisation, arrival of a new actor.

Positions move

BABOK states the warning: a stakeholder's position on the matrix changes over the course of the project, under the effect of a change of organisation, of environment or of requirements scope. The analysis is therefore iterative and is reviewed often. A grid filled in once and then filed becomes false in silence, and its falseness only shows on the day the actor filed under monitor asserts a power nobody credited him with. Movement is also the most useful information the technique produces: an arrow between two quadrants says something no static placement says, namely that the project has just changed nature for that actor.

Movement is provoked as much as it is suffered: that is the active half of the technique. BABOK is explicit on two cells. Faced with a powerful and unconcerned actor, the strategy is to meet their needs while working to raise their interest. Faced with an actor who is neither powerful nor concerned, additional engagement can move them into the ambassador quadrant, where they bring the project support nothing obliged them to give.

On the "keep satisfied" cell the two conducts coexist and the choice between them is an analyst's decision. The defensive conduct is to identify the trigger and warn the actor before the event arrives: you leave them at low interest and settle for not surprising them. That costs little and it returns only the absence of a bad surprise. The offensive conduct is to raise their interest deliberately, because you need their attention, their arbitration or their political cover: you move them into "manage closely" and you pay the price for it, the time of an actor of that rank, in exchange for a decision nobody else can take. The same reasoning holds on "monitor" and on "keep informed", where the additional engagement seeks a relay rather than an arbitration: a well-handled user who explains the project to their peers does work the analyst cannot do in their place.

An arrow on the grid can therefore be suffered or provoked, and the analyst has to know which one he is looking at. The suffered arrow is an event to absorb: scope changed, an actor discovered the project, a reorganisation redistributed authority. The provoked arrow is the result of a strategy. It is judged on what it cost and what it returned. Both are drawn the same way and read differently. The register therefore carries, beside each movement, the reason that produced it.

A position can also be declared without being held. The ABPMP's BPM CBOK, which follows stakeholders through a process transformation, places the moment at which the gap becomes visible: the initial fact-finding study, that is the analysis of the current-state model, identifies those who say they support the project and hold it back. The signals to look for are behavioural: missed meetings, slow decisions, repeated changes of decision. Resistance may stay subtle: it then shows in the repetition of those behaviours. The project manager finds it provided he looks for it. For those who truly obstruct, the guide asks for mitigation steps designed with the project sponsor and taken, where necessary, to executive management.

What makes the exercise fail

Confusing the title with the power

The org chart gives formal authority only. The director who signs without reading weighs less on the initiative than the head of operations whose opinion decides go-live. A placement based on business cards files both on the wrong side.

Placing alone

Informal power is what an outside analyst cannot see. The placement is triangulated with people who know the house, and the position of a key actor is validated with the sponsor before being treated as settled.

Reading "low power" as "harmless"

A group with no individual power acquires some by coalescing, by taking the matter to its professional association or by refusing to use what was built for it. The "keep informed" quadrant is the most dangerous on the drawing, because it is the easiest to neglect.

Confusing manifest interest with exposure

An actor can be the most affected of all and the least interested today. Carrying them at high interest because they "ought" to be distorts the grid and the allocation that follows from it. The placement follows what is; the trigger carries what is coming.

Freezing the grid

With no date, no written justification, no trigger and no review milestone, the matrix is a snapshot that expires without warning.

Circulating names where the culture will not bear it

A senior manager who discovers in writing that he is classified as low power and low interest does not argue with the method: he argues with the author. The choice of granularity (step 2) is the guard, and it is made beforehand.

Stretching the grid beyond its population

The PMBOK Guide judges two-axis grids useful on small projects or on projects whose relationships stay simple. The three-levels-per-axis refinement (step 7) treats a quadrant that is too crowded; it does not treat a population whose relationships are themselves the problem: networks of alliances, representation through umbrella associations, cross-dependencies. The standard then names the stakeholder cube, which adds a third dimension to the grid. The salience model abandons the two axes and classifies actors by power, the urgency of their claim and the legitimacy of their involvement; it is this model that the PMBOK Guide intends for large communities or those crossed by dense networks.

Answering the wrong question

The matrix measures power and interest. Asked outside its domain, it produces a wrong answer.

AI considerations

The first pass of the placement is a task AI accelerates. Fed with the stakeholder list, the org chart, the minutes of the last governance meetings and the history of a comparable project, a language model proposes an initial grid in minutes, with a justification per actor. That grid is a draft which the session corrects: it moves the workshop's starting point from a blank page to an object the group attacks. The gain is in time and in the quality of the discussion, because a disagreement is easier to formulate against a proposal than against a void.

The second use is the re-run. Positions move, and revision is the step where the technique is most often abandoned: the marginal cost of that revision has to come down. A model given the old grid, the register line by line and the events that have occurred since (decisions, rulings, scope changes, people moving) produces a statement of the gaps: which triggers have fired, who moved, in which direction, on the strength of which fact. The analyst validates or rejects; he does not rebuild. The third use is the derivation into communication: once the quadrant is fixed, writing the one-page note for a "keep satisfied" and the detailed announcement for a "keep informed" is a bounded drafting job, and the model is good at it.

What AI cannot do follows from the very matter of the exercise. Placing a real, named person on a power axis is a judgement with consequences, and the model has only what its inputs give it: titles, an org chart, documents. It therefore reproduces the bias the technique exists to correct, the confusion between rank and influence, and it cannot spot the latent power of a group that could coalesce, since that coalition appears in no document until it has happened. A position validated by a human who knows the politics of the house is the only usable placement.

The legal point remains. The revised Federal Act on Data Protection (revFADP), in force since 1 September 2023, governs profiling, that is the automated processing of personal data intended to evaluate characteristics of a person. A facilitation grid drawn in a workshop, discussed and then thrown away, stays outside that qualification. A tooled chain that automatically generates, stores and reuses assessments of the power and influence of named people comes close to it, because it keeps a file on those people. The rule of conduct: an AI-assisted placement is a workshop draft that a human confirms. It never becomes an automated, durable record about a named individual.

Examples

A health insurer is digitising its claims handling: a self-service portal for the insured persons and automated intake of the invoices sent in by service providers. The stakeholder list was established beforehand. The placement session was held at role and group granularity.

Beside the grid, the register carries the other half of the deliverable: why each actor sits where they sit and what would move them.

Three trigger cells are empty: executive management, the IT department and the claims handlers are already at high interest, there is nothing to wake in them. The grid is then read through what it imposes. Executive management, the programme's sponsor, and the IT department that builds it occupy the "manage closely" quadrant: they decide and they follow, engagement there is bilateral and continuous.

Three actors sit in "keep satisfied" for different reasons. The FOPH supervises social health insurance under LAMal/KVG and has expressed no interest in the project as such, but any change to the statutory handling of claims will bring it in at once. The FDPIC supervises the application of data protection law and its power bears on the project as soon as health data is involved, which is sensitive data; on the portal itself it has not spoken. The medical adviser holds a de facto veto over any automation that touches the medical assessment of a contested claim; on the rest of the scope he has voiced no view. Three high powers, three distinct triggers, three different messages: the "keep satisfied" strategy is settled actor by actor.

The claims handlers and the service providers, among them physiotherapists and medical practices, are in "keep informed": their daily work changes entirely and their individual power over the design is low. The insured persons and the cantons stay in "monitor". The former will be heavily affected by the portal and today do not know the project exists, which puts them at low interest and makes go-live their trigger. The latter pay health insurance premium subsidies, which the digitisation of claims handling does not touch. The placement of the insured persons sets how much engagement effort goes to them during the project. Knowledge of how they will use the portal comes from personas.

Six months later the grid is re-run, and three positions have moved. Scope has been extended to the automated triage of reimbursement claims, which sorts health data: the FDPIC's trigger has fired, its interest rises, it moves into "manage closely". The portal goes live: the insured persons discover the project and move from "monitor" to "keep informed". Automated intake becomes the only accepted channel for invoices: the service providers, through their umbrella association, acquire a capacity to block the cutover and move into "manage closely". The first two movements are rises in interest, and the register announced them, trigger in hand. The third is a rise in power, an event of a different nature: the register could record the trigger, it could not announce that the providers would coalesce to convert it into blocking capacity. On the drawing, the difference is read from the direction of the arrow: interest is won horizontally, power is won vertically.

Visualisations

The matrix is a spatial artefact: the information is carried by the position. A table listing the same actors with a "power" column and an "interest" column would lose what the grid gives at a glance, the distribution of the population across the plane and the fact that two actors with the same power fall under two opposite strategies because they are not in the same column. The drawing is therefore the grid itself, axes labelled, quadrants named, stakeholders placed inside their cell. The register that goes with it is text in rows and columns, so it is a table.

The axes are declared on the drawing

Power on the vertical axis, interest on the horizontal axis, increasing upwards and to the right, labels written out: the impact variant circulates, and a grid where nobody knows what the horizontal axis carries can no longer be read.

The quadrants carry their strategy

"Manage closely" is an instruction, "high power, high interest" is an observation and the reader who was not in the session needs the first.

The population stays legible

Beyond fifteen labels or so the grid saturates: group by role or move to the six-category variant.

The arrows carry the movement

Movement is drawn on the same grid: an arrow from the old position to the new one, dated, carrying in one line the reason for the move. The horizontal arrow says interest changed, the vertical arrow says power changed, the two call for different responses. The ceiling is low: beyond three or four arrows the grid becomes illegible, and two dated grids side by side, the old one and the new one, serve better.

Cost

Execution is brief; preparation and upkeep weigh more.

Tooling

The whiteboard and repositionable notes remain the best support for the placement session, and Bryson describes that setup: two axes drawn with a marker, one label per stakeholder, labels physically moved during the discussion. The cost of moving a label must be zero, otherwise the group stops moving positions and settles for the first proposal, which removes the step that carried the value. A spreadsheet works for preserving the result and re-running it: a power column, an interest column, a justification column, a trigger column and a scatter plot as output, which makes the gap between two versions immediate to produce.

Collaborative whiteboards (Miro, Mural, Microsoft Whiteboard) carry the same metaphor for a distributed team, with persistence and versioning on top, which serves the re-run at milestones. The technique presupposes none of these tools: it holds just as well on a flip chart, and an organisation whose workstations are locked down to the office suite runs the session without installing anything.

Some portfolio and requirements management tools embed a stakeholder register where power and interest are attributes on each actor's record, and the grid is a generated view of it. The value is in keeping the matrix tied to its source: when the list changes, the grid follows. That only holds for an organisation that already maintains such a register; buying one for this technique alone would be disproportionate. As soon as the grid names individuals, its support has to carry access control and restricted circulation.

HERMES, the Swiss Confederation's project management method, makes the positioning an outcome in its own right, the Stakeholder interests (outcome 4.4.1.48, Stakeholderinteressen), whose prescribed content is the position of the stakeholders, their description and the conflicts of interest and objectives identified. The outcome is distinct from the stakeholder list, and HERMES calls this analysis a subjective assessment, internal to the project, which is not a public outcome.

Sources

• IIBA, A Guide to the Business Analysis Body of Knowledge (BABOK Guide) v3, §10.43 Stakeholder List, Map, or Personas.
• Mendelow, A. L., "Environmental Scanning: The Impact of the Stakeholder Concept", ICIS 1981 Proceedings, paper 20 (AIS eLibrary): the origin of the power/interest grid. The date is 1981 and not 1991, contrary to what many secondary sources repeat.
• PMI, A Guide to the Project Management Body of Knowledge (PMBOK Guide), 8th edition, §5 Tools and Techniques, "Stakeholder mapping/representation": two-axis grids judged useful on small projects or on populations with simple relationships, then the stakeholder cube and the salience model, the latter for large or densely connected communities.
• HERMES 2022, Project Management Reference Manual, Swiss Confederation, Federal Chancellery, outcome 4.4.1.48 Stakeholder interests: the prescribed content, position of the stakeholders, description and conflicts of interest and objectives identified, an outcome distinct from the stakeholder list and its qualification as a subjective assessment, internal to the project, which is not a public outcome.
• ABPMP, Guide to the Business Process Management Common Body of Knowledge (BPM CBOK), §7.3.6 Stakeholder Management: declared support covering real resistance, which may stay subtle but is spotted through repeated behaviours by whoever looks for them, from the analysis of the current-state model onward, then the mitigation steps designed with the sponsor and taken where necessary to executive management.
• The Open University, OpenLearn, "Mapping stakeholders' power and interest" (§3.3): the Johnson and Scholes rendering of the grid and its four categories, where the top-right cell is named "key influencer".
• Bryson, J. M., "What to do when Stakeholders matter", Public Management Review, vol. 6, no. 1, 2004, pp. 21-53: the method for building the grid in session, label by label.