Home›Blog›Project stakeholders: who counts, and on what criterion

FundamentalsBeginner

Published September 6, 2026 · By Vincent KENNEL

Project stakeholders: who counts, and on what criterion

Only one of four criteria appears in every definition, and it is not the one usually quoted. Six English-language references compared, the power/interest grid returned to its actual authors, and the 1963 report where the word first appears.

A wooden and plastic model seen from above. At the centre, a wooden disc carries the word PROJECT. Six pieces are tied to it by blue cords: an office block, a figure in a suit, a group of three figures, a factory, a tree and a columned building. Three signs stand over them, reading AFFECTS THE PROJECT, AFFECTED BY THE PROJECT and PERCEIVES ITSELF AS AFFECTED. To the right, beyond a dashed line drawn on the surface, an orange block is tied to the disc by an orange cord, under a sign asking INTEREST ONLY ?

In brief

A stakeholder is any person, group or organization that has interests in, or can affect, be affected by, or perceive itself to be affected by a project (ISO 21506:2024). The four criteria are not carried by the same texts: PRINCE2 7, PM² and GovS 002 do not mention interest, and the APM Body of Knowledge, 8th edition, mentions neither the ability to affect the project nor self-perception.

A stakeholder register has an unpleasant property: it becomes wrong the moment it stops being maintained, and nothing signals it. A regulator enters the picture halfway through, an internal department discovers that the project concerns it, a local community organises. The document sits where it always sat, with its creation date and its air of being complete.

Every reference standard says to reopen it. ISO 21502:2020 asks that project stakeholders be "identified, analysed, documented and engaged throughout the project" (clause 7.12.1). GovS 002, the UK government functional standard, requires the engagement plan to be "implemented, monitored and updated to reflect newly emerging stakeholders and changes in the position of existing stakeholders" (version 2.1, 1 September 2025, clause 7.12). And Johnson and Scholes noted long ago that stakeholder groups recompose themselves according to the decision at hand (Exploring Corporate Strategy, 5th edition, 1999, p. 215).

There is a question that comes before that one, and it usually gets answered without being asked: on what criterion does a third party enter the list at all? The definitions do not settle it. And the references do not draw the line in the same place.

What a stakeholder is

The ISO vocabulary gives the reference definition. A stakeholder is a "person, group or organization that has interests in, or can affect, be affected by or perceive itself to be affected by, any aspect of a project, programme or portfolio" (ISO 21506:2024, clause 3.86). ISO 21502:2020 carries that sentence word for word at clause 3.27, and both trace it to the same origin, ISO/TR 21506:2018 clause 3.79, a technical report that became a full standard in 2024.

Three things sit in that sentence that a quick reading passes over.

The subject is not necessarily a person. The definition holds the person, the group and the organization. A register is therefore not a list of names, and an entire division, a user community or a public body earns its line on the same footing as an individual.

The relation bears on an aspect of the project, not on the whole project. A supplier concerned by a single work package is a stakeholder as much as the customer who holds the contract. That is what rules out reasoning by apparent importance when the list is first drawn up.

Perception is enough. "Perceive itself to be affected" appears in five of the six definitions examined here, and it is the clause that most surprises practitioners. It is not a drafting indulgence: opposition founded on an inaccurate fear does exactly the same thing to a schedule as opposition founded on an accurate one.

One last point, and it is a habit worth acquiring: the definition above carries a year, and that is not bibliographic fussiness. The same sentence existed in 2018 as a technical report, a document class that carries no requirements, and in 2024 as a published standard. Writing "according to ISO" without a year names no text in particular, and the same trap waits further down this article, where two editions of one guide give the same list of grids without attributing any of them, and a third qualifies the lot.

Where the boundary falls

Four criteria circulate in the definitions: having an interest in the project, being able to affect it, being affected by it, and perceiving oneself as affected. Only one of the four appears in every definition examined here, and it is not the one usually quoted.

A table crosses six reference documents in rows against four criteria in columns. The rows are ISO 21506:2024, ISO 21502:2020, the APM Body of Knowledge 8th edition, PRINCE2 7, the PM2 Guide 3.1 and GovS 002 version 2.1. The columns are interest alone, can affect, is affected, and perceives itself affected. The column for being affected reads yes for all six and is visually muted. The interest column splits the field: yes among other routes for the two ISO texts and for the APM Body of Knowledge, not in the text for the three others. The can affect and perceives itself columns read yes for five documents and not in the text for the APM Body of Knowledge alone. — Only being affected by the project is common to all six definitions. The interest criterion splits the field in half, and the most recent reference of the six is alone in dropping two of the other three.

Two divergences run through the table, and they are not of the same kind.

The first splits the field down the middle. The two ISO texts admit interest as one route among others, through an "or". The four other references do not carry the word at all. PRINCE2 7 defines a stakeholder as "any individual, group, or organization that can affect or be affected by (or perceives itself to be affected by) the project" (2023, glossary). The European Commission methodology and the UK functional standard say the same thing in almost the same words: "any individual, group or organisation that can affect, be (positively or negatively) affected by, or perceive itself to be affected by the project" (PM² Guide 3.1, glossary), and "any individual, group or organisation that can affect or be affected by, or perceive itself to be affected by an initiative" (GovS 002, version 2.1, Annex B). Interest on its own does not open the door.

The second is the work of a single reference, and it is the most recent of all. The APM Body of Knowledge, 8th edition, 2025, defines a stakeholder as "an individual or group that has an interest or role in the project, programme or portfolio, or is impacted by it" (glossary). Interest is in, and so is holding a role, which no other definition mentions. But the ability to affect the project is absent, and so is self-perception. The broadest reference on one axis is the narrowest on the other two.

A word of caution on how to read the empty cells. What the table records is an absence in the text, established by reading it. None of these documents states that it is excluding a criterion, and nothing in them establishes an intention. An absence is defensible as an observation; it is not evidence of a decision.

The consequence is immediate and operational. A regulator that watches without holding any power over this particular project, a trade body that follows the file without taking part: they enter the register under the ISO definition, they do not enter it under the four others. Reading a definition does not decide the case, and the definitions do not agree.

What follows from this is not that one reference is right against the others. It is that the perimeter of the register is a decision, that it is taken once, and that it is worth writing at the top of the document. A register where nobody knows whether interest without leverage counts is a register that can be neither challenged nor completed.

The main categories

Once the criterion is settled, things have to be sorted. The project management standard lists eight categories: "Stakeholders should include, but are not limited to: a) the sponsoring organization and project team (see 4.5); b) customers; c) partners and suppliers; d) special interest or pressure groups; e) regulatory bodies; f) finance providers; g) shareholders; h) relevant external third parties" (ISO 21502:2020, clause 7.12.2, p. 39).

Eight categories handle badly. They regroup into five families, which have the advantage of matching genuinely distinct counterparts on a project under contract.

Family	What it holds
Contractual	Customers, partners and suppliers. Everything bound to the project by a written commitment
Internal	The sponsoring organization and the project team
Financial	Finance providers and shareholders
Regulatory	Regulatory bodies
External	Special interest or pressure groups, and relevant external third parties

Finance deserves a line of its own rather than a place among the external parties: a shareholder and a neighbouring resident share neither leverage, nor horizon, nor channel, and merging them under one label makes the category unusable.

A useful counterpoint, because it shows the same object can be cut differently. PRINCE2 7 does not classify categories at all. It holds that "all projects have the following primary stakeholders: the business, users, and suppliers" (2023, chapter 2), and builds its project board on those three interests rather than on a list of types. "Who are they" and "whose interests must be represented in the governance" are two different questions, and two references in force answer them separately.

The grids, and how to use them

The most reproduced grid crosses power with level of interest and distributes four postures: minimal effort when both are low, keep informed when interest is high but power low, keep satisfied in the reverse case, and key players when both are high (Johnson and Scholes, Exploring Corporate Strategy, 5th edition, 1999, exhibit 5.5, p. 216).

This matrix is not Mendelow's, contrary to what is written almost everywhere. The source note printed under the Johnson and Scholes plate points to a Mendelow paper given at the second International Conference on Information Systems, which it dates to 1991 when the paper is from 1981. And that paper does not contain the matrix: it crosses power with the dynamism of the environment, its quadrants are scanning modes rather than management postures, and the word interest does not appear once across its thirteen pages (Mendelow, "Environmental scanning: the impact of the stakeholder concept", ICIS Proceedings, 1981, p. 407-417). The Johnson and Scholes note says "adapted from" for good reason: the power/interest form and the four labels are theirs.

The useful part is not the grid itself, which everyone knows, but the conditions of use its authors take the trouble to write down and that the reproductions leave out.

Find the right grain. There is "clearly a balance to be struck between describing stakeholders too generically and subdividing them to a point where the map is confused" (p. 217). A map with five entries is useless; so is a map with forty.

Place a role, not a person. The authors recommend asking whether a different holder of the same post would move the point, and warn that "serious errors of judgement can be made if this is not attended to" (p. 220). That is the difference between a map that survives a replacement and a map to be redrawn at every move.

Do not treat it as a permanent artefact. The analysis "is most useful when it is related to a specific strategic development" (p. 215). In other words there is no single map of the project; there is one map per structuring decision.

And above all, draw two maps. This is the move almost nobody reproduces, and the one that makes the tool actionable: establish where stakeholders currently stand on a given decision, then where they would need to stand, and read the gaps between the two. "By comparing these two maps and looking for mismatches, political priorities can be established" (p. 220). One map describes; two maps say what to do and in what order.

Two identical charts side by side, power on the vertical axis and interest on the horizontal one, scaled from low to high and divided into four quadrants carrying no labels at all. The left chart, titled current position, holds four anonymous stakeholders marked A, B, C and D. The right chart, titled desired position, carries the same four, marking each current position with a dashed circle and each displacement with an arrow. A moves towards more interest and more power, B towards more interest only, C towards more of both, D towards more power only. The four displacements are of different lengths. An arrow between the two charts is labelled read the gaps.

The other grids exist and are worth as much. The PMBOK Guide lists four, power/interest, power/influence, influence/impact and the salience model, in a list identical from its 4th edition (2008, p. 249 of the English edition) to its 5th (2013, p. 396), and attributes none of them to anyone. Its 6th edition (2017) is the first to qualify them: these classification models "are useful for small projects or for projects with simple relationships" (section 13.1.2.4). The salience model, for the record, crosses three attributes, power, legitimacy and urgency, not two (Mitchell, Agle and Wood, "Toward a theory of stakeholder identification and salience", Academy of Management Review, 22(4), 1997).

What a register contains

The typical contents of a register are stable from one edition of the PMBOK Guide to the next, and they hold in three blocks (4th edition, 2008, p. 250; 5th edition, 2013, p. 398).

Identification information: name, organizational position, location, role in the project, contact information.

Assessment information: major requirements, main expectations, potential influence, and the phase in the life cycle where this stakeholder's interest is highest. This is the block that gets filled in least and serves most: it turns a list into a sequencing tool.

Stakeholder classification: internal or external, supporter, neutral or resistor. The 6th edition (2017) opens that third block to "any other classification model chosen by the project manager" (section 13.1.3.1), which explicitly authorises carrying the five families above rather than the internal/external pair.

Two additions are worth borrowing from the NASA systems engineering handbook, freely available, that are found nowhere else (NASA Systems Engineering Handbook, NASA/SP-2016-6105 Rev2, section 4.1). The first is to separate primary stakeholders from the rest at the point of identification, which avoids treating forty lines at one level of attention. The second is to place expectations under baseline, that is, to freeze them and trace their evolution as one would a requirement. On a project under contract, this is the move that makes it possible to demonstrate that an expectation changed along the way, which is an entirely different conversation from observing that it has not been met.

One last point, which saves time: there is no universal template. The European Commission methodology describes its stakeholder matrix without publishing its columns, which live in a separate distributed model, and states that it "should be adapted to the needs of the project" (PM² Guide 3.1, section 6.3). The APM Body of Knowledge prescribes no register with fixed contents. Looking for the right template is therefore a search without an object: one has to settle on a template and stay with it.

Where the word comes from

The word appears in 1963, at the Stanford Research Institute. What is interesting is not the date but the way it was handed down, because that transmission is the reason so much of what circulates about the origin is second hand.

The reference text on the origin writes that the word was "coined in an internal memorandum at the Stanford Research Institute in 1963" (Freeman and Reed, "Stockholders and stakeholders: a new perspective on corporate governance", California Management Review, XXV/3, 1983, p. 89). But note 7 of that same article, p. 104, establishes that its authors never saw the document. Their sources are an anonymous reviewer and a correspondence with William Royce, a consultant at SRI, twenty years after the fact.

Fifteen years later, a researcher at the Centre for Business Research in Cambridge identifies the item, cites it with page numbers, attributes the word to a member of the Long Range Planning Service, and corrects the nature of the object: it is not an internal memorandum but a subscription report circulated to outside clients (Slinger, Spanning the gap, CBR Working Paper 111, 1998). His principal source is the same William Royce.

Between the two, the most cited article in the literature on the subject opens its chronology with the entry "Stanford memo, 1963", followed by "cited in Freeman and Reed, 1983" (Mitchell, Agle and Wood, 1997, table 1, p. 858). The admission of second-hand sourcing is explicit, and their bibliography carries no Stanford Research Institute entry at all. It is that shorthand, "Stanford memo", picked up afterwards, that fixed the object in the discipline's memory as something it was not.

Three links, then, and a single witness at both ends. The two accounts of the origin do not corroborate each other: they are the same source, better documented the second time. This matters beyond the anecdote, because a citation chain that looks like three independent confirmations and turns out to rest on one testimony is the ordinary failure mode of a well-established fact.

A diagram in two bands. The upper band, titled what circulated, starts from William Royce, sole witness and consultant at SRI. Two arrows leave him, one to Freeman and Reed 1983, the other to Slinger 1998. A third arrow runs from Freeman and Reed to Mitchell, Agle and Wood 1997 and is marked second hand. Three annotations state that note 7 of Freeman and Reed records that its authors never saw the document, that Mitchell cites them at second hand, and that Slinger corrects the nature of the object to a subscription report. A dashed line separates the two bands and carries the statement no link, none of the publications above ever saw the document. The lower band, titled what was read, connects report R168 of SRI, 1963, presented as the original document, to Puyt 2025, by an arrow annotated read on the physical copies.

The report itself has since been found and read, by a team that worked on the physical copies of the series. According to Puyt, it carries on its very first page a list of categories, customers, employees, lenders, owners, suppliers, and on page 6 a figure titled "A method for analyzing stakeholder expectations" (Puyt, The origins and institutionalization of SWOT analysis, doctoral thesis, University of Twente, 2025). Identifying and categorising is therefore not a late development of the subject, added once the discipline had organised itself. It is the founding gesture, present on the first page of the first document.

In short

Two moves are enough to make a register usable, and neither of them requires a tool.

Write the criterion before writing the list. Decide whether interest on its own is sufficient, in this context and for this project, and carry that decision at the top of the register. The references do not agree on the point, so nobody will settle it on the project's behalf, and a register whose entry criterion is implicit can be neither challenged nor completed.

Reopen the list at every structuring decision, rather than once at launch. Not out of documentary discipline, but because the map of positions recomposes itself at each arbitration: that is what the three texts quoted at the outset all say, and it is also what makes the two-map method workable at all.

Stop guessing. See the real impact.

Frequently asked questions

Q.Is a competitor a stakeholder?

Yes, under all six references compared here, but not through the door one expects. A competitor matches none of the eight categories in ISO 21502:2020, clause 7.12.2. It enters through the criterion: it can affect the project. Mendelow counted competitors among his own eight categories in 1981.

Q.Is a stakeholder the same as an interested party?

A vocabulary question, not a boundary one. ISO 31000:2018, clause 3.3, admits the second in a note as an alternative term for the first; ISO 31073:2022, clause 3.3.2, reverses their rank and drops the note; ISO 21506:2024 defines neither, across its hundred entries.

Q.Is a stakeholder register mandatory?

None of the texts cited here imposes one. ISO 21502:2020 and the PMBOK Guide, 6th edition (2017), are guidance, not requirements. GovS 002, version 2.1, is sharper still: it treats stakeholder engagement in should where its neighbouring clauses 7.11 and 7.13 end in a shall.

References

A. L. Mendelow - Environmental Scanning - The Impact of the Stakeholder Concept - 1981
APM - APM Body of Knowledge - 8th edition, 2025
AXELOS - PRINCE2 7 - Managing Successful Projects - 2023
Gerry Johnson, Kevan Scholes - Exploring Corporate Strategy - 5th edition, 1999
Giles Slinger - Spanning the Gap - The Theoretical Principles That Connect Stakeholder Policies to Business Performance - October 1998
ISO - ISO/TR 21506:2018 - Project, programme and portfolio management, Vocabulary - 2018
ISO - ISO 21502:2020 - Project, programme and portfolio management, Guidance on project management - 1st edition, 2020
ISO - ISO 21506:2024 - Project, programme and portfolio management, Vocabulary - 1st edition, 2024
ISO - ISO 31000:2018 - Management du risque, lignes directrices - 2nd edition, 2018
ISO - ISO 31073:2022 - Management du risque, vocabulaire - 1st edition, 2022
NASA - NASA SP-2016-6105 - Systems Engineering Handbook - Revision 2, 2016
Office des publications de l'Union européenne - PM² Project Management Methodology Guide 3.1 - EN, 2023
PMI - A Guide to the Project Management Body of Knowledge (PMBOK Guide) - 4th Edition - 2008
PMI - A Guide to the Project Management Body of Knowledge (PMBOK Guide) - 5th Edition - 2013
PMI - A Guide to the Project Management Body of Knowledge (PMBOK Guide) - 6th Edition - 2017
R. Edward Freeman, David L. Reed - Stockholders and Stakeholders - A New Perspective on Corporate Governance - 1983
R. F. Stewart, J. K. Allen, J. M. Cavender - The Strategic Plan - LRPS Research Report R168, Stanford Research Institute - 1963
Richard Willem Puyt - The Origins and Institutionalization of SWOT Analysis - From an Underappreciated Strategy Tool in Theory to a Widely Adopted Viral Idea in Strategic Management Practice - 2025
Ronald K. Mitchell, Bradley R. Agle, Donna J. Wood - Toward a Theory of Stakeholder Identification and Salience - Defining the Principle of Who and What Really Counts - 1997
UK Government - Government Functional Standard GovS 002 - Project delivery - Version 2.1, 2025

In this area: Stakeholders & coordination