Strategy making
Strategy as text, cascade, scenarios, and the gap between talk and action.
Strategy making has the same shape as planning: a text that speaks for an organisation, and a body of action that may or may not follow it. The literature here is large, and it already makes most of Holagram's rules, one at a time. There are working methods for reading a strategy as a text: discourse analysis of a Finnish city's strategic plan (strategy-01), a performative reading of a city strategy (strategy-02) and a study of ambiguity in strategy texts (strategy-03). The case that no level sums up another is made from several sides: talk, decision and action are decoupled by design (strategy-06, strategy-07), realised strategy is the pattern in action and not the text (strategy-05), unit-level readers translate a strategy in their own routines (strategy-08), and a figure at one level distorts the level below it (strategy-15, strategy-16, strategy-17). Scenarios are the field's own labelled fiction. The best evidence on them is mixed, and the clearest failure was published because failures usually are not (strategy-09, strategy-10, strategy-11, strategy-12). A meta-analysis finds that formal planning helps on average (strategy-19), so a sceptical reading has to be earned text by text. What the literature does not offer is a protocol that keeps the reader and the tester apart and states the level of every claim. The main risk for Holagram is its own tallies: any count of held or broken entries can be gamed, as the metrics literature predicts.
Sources
- Eero Vaara, Virpi Sorsa and Pekka Pälli (2010). On the force potential of strategy texts: a critical discourse analysis of a strategic plan and its power effects in a city organization. Organization 17(6), pp. 685-702checkedmust read
A worked example of reading a strategy document as a text. This gives the Reading register at organisation level a checklist drawn from a real plan, set in a Finnish city, and a text of the kind the Cities method also reads. The study reads what the text does to its readers and does not test it against what the city later did; that Test is what Holagram adds. The authors themselves say the features can be generalised only 'with due caution'.
Read first: The abstract lists the five features. The 2009 companion (Pälli, Vaara and Sorsa, Discourse & Communication 3(3): 303-318) analyses the same city's plan as a genre and is the shorter way in.
- Martin Kornberger and Stewart Clegg (2011). Strategy as performative practice: the case of Sustainable Sydney 2030. Strategic Organization 9(2), pp. 136-162checked
Reads a city strategy, Sustainable Sydney 2030, as a practice with effects. The authors argue that the strategy rendered the city knowable, mobilised the public and legitimised outcomes while silencing other voices, and that strategising works as an aesthetic performance joining facts and values. It is the nearest bridge between the Strategy and Cities methods: one kind of document, read once as a plan about a place and once as an organisation's strategy. It adds a performative reading (what the text does) beside an argumentative one (what it claims), and it names the voices a text leaves out, which are ready Test targets for readers at street level. An Australian case from 2011; transfer to Estonian or Gulf settings is untested.
Read first: The abstract states three theoretical contributions and five practical implications; start with the empirical analysis of the strategy-making process.
- Chahrazad Abdallah and Ann Langley (2014). The double edge of ambiguity in strategic planning. Journal of Management Studies 51(2), pp. 235-264checked
Shows that ambiguity in a strategy text is not simply a defect. In a case study of strategic planning in a cultural organisation the authors find three forms of ambiguity in the text and show that members consumed them in different ways. Ambiguity at first enabled action; over time it produced internal contradiction and overextension. This is the literature's nearest match to Holagram's rule that ambiguity stays. It lets the Reading register mark ambiguous phrases and predict who will take them which way, and it gives the Test a target: consumption by another group. It also warns against a Reading that treats ambiguity as a fault, since the ambiguity may be doing work.
Read first: The abstract names the three forms of ambiguity and the move from enabling to overextension; the findings on how members consumed the text come next.
- Richard P. Rumelt (2011). Good Strategy Bad Strategy: The Difference and Why It Matters. Crown Business (UK edition: Profile Books)checked
The practitioners' reference on what a strategy is, and so the likeliest objection from the field. Secondary summaries describe a 'kernel' of diagnosis, guiding policy and coherent action, and name symptoms of bad strategy: fluff, failure to face the challenge, and goals mistaken for strategy. As a grammar, the kernel lets an Entry classify a passage of a strategy document before reading it for meaning, as Mandelbaum's three readings do for plans (held documents-01). As an objection: a document is not the strategy, and reading fluff for hidden depth is over-reading. The book has not been opened in this run, so check the summaries against it before relying on them.
- Henry Mintzberg and James A. Waters (1985). Of strategies, deliberate and emergent. Strategic Management Journal 6(3), pp. 257-272checkedmust read
The base for any Test that sets a strategy text against action. The abstract treats deliberate and emergent strategy as two ends of a continuum and lists the types found in research: planned, entrepreneurial, ideological, umbrella, process, unconnected, consensus and imposed. Read with Mintzberg's definition of strategy as a pattern in what an organisation does (recalled; not in the abstract), it tells Holagram what the ground is for a strategy Entry: the stream of moves at lower levels, not a second document. It also offers more verdicts than held and broke: intended and carried out, intended and dropped, found in action without being intended (the paper's scheme, from memory).
Read first: The abstract lists the eight types (planned, entrepreneurial, ideological, umbrella, process, unconnected, consensus, imposed); they are the vocabulary to borrow.
- Nils Brunsson (translated by Nancy Adler) (1989). The Organization of Hypocrisy: Talk, Decisions and Actions in Organizations. Wiley, Chichester and New York (1989); 2nd edition Copenhagen Business School Press (Open Library gives March 2003)checkedmust read
The main reason a strategy Test cannot treat a gap between text and action as a refutation. As described in the reviews seen in this run, the book separates talk (aimed at legitimacy), decisions (formal policies and programmes) and action (day-to-day operations), and argues that organisations facing conflicting demands let the three drift apart on purpose. A Reading that says 'this strategy is talk' is therefore not shown wrong by finding talk and action apart; it is shown wrong only if they turn out to be coupled. The book names three things the organisation level produces and sets the polarity of the Test. Commentary describes the hypocrisy as organised, not as a moral charge. The book has not been opened.
- Patricia Bromley and Walter W. Powell (2012). From smoke and mirrors to walking the talk: decoupling in the contemporary world. Academy of Management Annals 6(1), pp. 483-530checked
Updates the decoupling idea of Meyer and Rowan (1977) and Weick (1976). The authors argue that the usual picture, a gap between policy and practice, hides a more common gap between means and ends: practices are in place but their link to the stated goal is opaque. They list consequences: more complex internal structures, perpetual reform and resources diverted from core goals. For Holagram this gives two separate Tests for one strategy claim (is the practice there; does anything connect it to the stated end) and a reason why a cascade with a measure at every level may still sum up nothing.
Read first: The abstract's distinction between policy-practice and means-ends decoupling, then the section the abstract describes on when to expect each.
- Linda Rouleau (2005). Micro-practices of strategic sensemaking and sensegiving: how middle managers interpret and sell change every day. Journal of Management Studies 42(7), pp. 1413-1441checked
A move-level and unit-level study of how middle managers in a clothing company interpret and sell a strategic change in everyday routines and conversations. It names four micro-practices: translating the orientation, overcoding the strategy, disciplining the client and justifying the change. It supplies the lowest levels of the strategy method and shows what 'another reader' can mean in practice: the people who must act on the strategy read it, and an outsider writes it up. It studies a private firm by ethnography; a first study on public material cannot copy it and can only approximate it through minutes, job adverts and tender notices.
Read first: The abstract names the four micro-practices; the case material on routines and conversations follows.
- Gerard P. Hodgkinson and George Wright (2002). Confronting strategic inertia in a top management team: learning from failure. Organization Studies 23(6), pp. 949-977checkedmust read
A first-hand account of a largely unsuccessful attempt to use scenario planning and cognitive mapping with a senior management team. The authors say the popular management literature suggests such methods invariably succeed, that the absence of documented failures may mislead would-be users, and that their own intervention failed because participants used defensive avoidance, amplified by psychodynamic processes started by the chief executive (the abstract breaks off at that point). It is the closest published precedent for Holagram's failure log, and an argument for keeping one. It also names a failure mode for the whole method inside organisations: the host avoids the reading.
Read first: The abstract carries the argument for publishing failures; read the account of the failed intervention and the analysis using Janis and Mann's conflict theory of decision making that the abstract describes.
- George Wright, Kees van der Heijden, George Burt, Ron Bradfield and George Cairns (2008). Scenario planning interventions in organizations: an analysis of the causes of success and failure. Futures 40(3), pp. 218-236checkedopen access
Compares a successful scenario intervention with the failed one in strategy-09. According to the repository record the authors argue that a pre-intervention interview can show how receptive an organisation is; link 'strategic inertia' (bolstering a failing strategy, procrastination, buck-passing) to a reduced sense of threat; and warn that a scenario exercise that does not surface a clear alternative strategy can reactivate those coping patterns and reinforce the inertia. For Holagram this means a receptiveness check before a study in a host organisation, and a reason to pair every Fiction with a move someone can take. It is evidence that fiction can do harm as well as good.
Read first: The comparison between the successful case and the failed case of strategy-09, and the implications for practitioners at the end (both as the repository record describes them).
- Shardul Phadnis, Chris Caplice, Yossi Sheffi and Mahender Singh (2015). Effect of scenario planning on field experts' judgment of long-range investment decisions. Strategic Management Journal 36(9), pp. 1401-1411checked
Three field experiments with practising experts judging long-range investment decisions. Contrary to earlier findings, using several scenarios did not raise or lower the experts' confidence overall; their judgement moved according to how an investment fared in the scenario used, and afterwards they preferred more flexible options. This supports the use Holagram wants for Fiction: a world in which a move is tried, not a claim about the future. It also fits the labelling rule, since the effect shown is on the judgement of the move. The setting is investment decisions, not public strategy.
Read first: The abstract; the results on confidence and on the preference for flexible options are the two findings to take.
- Mark P. Healey and Gerard P. Hodgkinson (2024). Overcoming strategic persistence: effects of multiple scenario analysis on strategic reorientation. Strategic Management Journal 45(8), pp. 1423-1445checkedopen access
A laboratory study using a multi-round strategy simulation in which players had to change strategy to succeed. Multiple scenario analysis reduced strategic persistence by encouraging reorientation, indirectly, by fostering belief in a new strategy, and the effect depended on prior performance. It is recent, open access and the clearest controlled evidence in this thread that imagining several futures changes what people do. It also carries the risk for Holagram: the route is belief, so a Fiction can persuade whether or not it is sound. That is the case for the Adversary's Test coming after the Fiction and being written by a different hand.
Read first: The research summary and the managerial summary in the abstract.
- Sohail Inayatullah (1998). Causal layered analysis: poststructuralism as method. Futures 30(8), pp. 815-829 (reprint on the author's site)checkedopen access
Causal layered analysis reads an issue through four layers (litany, social causes, worldview, myth and metaphor) and treats movement between them as the work. The abstract on the author's reprint says its aim is to open space for alternative futures, not to predict. It is a ready format for a strong, even paranoid, Reading that stays inside the Reading register. Its layers measure depth of discourse, not scale, so they must not be mistaken for Holagram's levels (move, unit, organisation, field, world). 'Deeper is more real' is a risk built into the method and needs an Adversary.
Read first: The abstract and the section on the four layers; the page shows the full text with case studies.
- Stuart Candy and Kelly Kornet (2019). Turning foresight inside out: an introduction to ethnographic experiential futures. Journal of Futures Studies 23(3), pp. 3-22checkedopen access
A cycle for using made futures as ethnographic instruments, derived from five projects, with five steps: Map1, Multiply, Mediate, Mount, Map2. The first and last steps map before and after the artefact appears, so the second map records what readers who were not the authors did with it. For Holagram it gives a published way to put Alter's Fiction in front of other readers and to treat their responses as Traces of reading, not of the world. It treats fiction as something shown to people and then mapped, not only written. Open access.
Read first: Section headings seen: Images of the Future; Five Projects; Surfacing a Structure; Discussion, which sets out the five steps.
- Jerry Z. Muller (2018). The Tyranny of Metrics. Princeton University Presschecked
The general case against metric fixation, across education, medicine, business and finance, government, policing and the military, and philanthropy and foreign aid. The publisher's description says tying rewards to published numbers moves attention from the work to the measuring and produces gaming, and that metrics help when they complement judgement; the page reportedly mentions a checklist for when and how to use them. It supports the rule that no level sums up another by showing, case by case, a lower-level number standing in for a higher-level aim. Practitioners know it; use it as the shared reference and strategy-16 and strategy-17 for evidence.
- Christopher Hood and Barbara Piotrowska (2020). Goodhart's law and the gaming of UK public spending numbers. Public Performance & Management Review 44(2), pp. 250-271 (online 7 April 2020; issue dated March 2021)checked
Goodhart's law tested on UK public spending numbers. The authors argue that gaming occurs at several levels: aggregate input numbers at the top, performance indicators for output control at the lower level and, as they show, a meso level through creative categorisation of spending; and that creating new categories to stop gaming opens new ways to game. It is a direct demonstration that numbers at different levels are gamed in different ways, so one level's figure cannot stand for another's. Companions: Hood (2006) found that the central managers of the UK targets regime (introduced in 1998, per its abstract) did not put substantial resources into checking performance data and had no coherent antigaming strategy; Hood and Piotrowska (2024) name four arrangements for keeping gaming in check, authority, contestation or rivalry, 'bafflement' and 'togetherness', and conclude that no single one suffices.
Read first: The abstract: the three levels, the creative-categorisation mechanism, then the three UK cases it lists (protected spending, public-private partnerships, accounting changes).
- Wendy Nelson Espeland and Michael Sauder (2007). Rankings and reactivity: how public measures recreate social worlds. American Journal of Sociology 113(1), pp. 1-40checkedopen access
Shows how a public measure changes what it measures: people react to being ranked, through two mechanisms the authors name, self-fulfilling prophecy and commensuration. Based on media rankings of law schools. For the strategy method it describes the link from world level to field and unit level: a ranking or benchmark reshapes the behaviour it compares, so a strategy aimed at a rank is partly reading its own effect. Most useful where national or city indicators are set against global benchmarks. A sociological source with a US higher-education case.
Read first: The abstract: the two mechanisms and the patterns of effects they produce.
- Geoff Mulgan and Giulio Quaggiotto (2023). Creating useful policy steering rooms. Apolitical (solution article)checkedmust readopen accessdated 8 November 2023 (date printed on the page)
It proposes policy steering rooms whose four walls hold the patterns (key facts, trends and performance indicators), the proven (evidence on what works), the promising (innovations elsewhere) and the possible (how the whole system might look in 10 to 20 years). The four walls resemble the four registers closely enough that practitioners will ask for a mapping, and differ enough that the mapping has to be argued (idea 7). It is the nearest existing practitioner frame to what Holagram offers governance-innovation readers.
Read first: The whole article; it is short, and the four walls are its core.
- Bert George, Richard M. Walker and Joost Monster (2019). Does strategic planning improve organizational performance? A meta-analysis. Public Administration Review 79(6), pp. 810-819checkedopen access
The strongest evidence against a purely sceptical reading of strategy documents. A random-effects meta-analysis of 87 correlations from 31 empirical studies finds that strategic planning has a positive, moderate and significant effect on organisational performance, strongest when performance is measured as effectiveness and when planning is formal, and holding across public and private sectors and across countries. The abstract itself records the standing criticism that planning is overly rational and inhibits strategic thinking. For Holagram it is the counterweight: a Reading that calls strategy documents ceremonial has to explain why planning goes with better performance on average, and the first study should not assume the ceremony. It measures averages across studies, not what any one text does, so it cannot settle a single Entry.
Read first: The abstract: the sample (87 correlations from 31 studies) and the two moderators, how performance is measured and how planning is measured.
- Marianne Chuang, Gabriel Chuang, Cheryl Chuang and John Chuang (2025). Judging It, Washing It: Scoring and Greenwashing Corporate Climate Disclosures using Large Language Models. arXiv:2502.15094 (v1 20 February 2025; v2 23 May 2025)correctedopen accessdated 20 February 2025 (arXiv v1)
A recent test of LLMs both as judges of corporate climate disclosures (emissions targets and progress) and as writers of greenwashed versions of them. The authors report that numerical rating and pairwise comparison both separate stronger from weaker companies and that pairwise comparison holds up better against LLM-greenwashed responses. For Holagram it is a small model of the Adversary's job on corporate talk: what a synthetic reader can catch in invisible literature and what a synthetic writer can do to defeat it. It suggests pairwise comparison for any machine Test and tests of the Adversary on deliberately washed passages. Preprint; climate disclosure, not strategy documents.
Read first: The abstract: the three steps (judge, greenwash, test whether the judge survives).
Audit: The arXiv comments line reads 'To appear, ClimateNLP 2025', so it is a workshop paper as well as a preprint; the scout's 'no peer review shown' should be softened.
What it implies for the proposals
- Levels. Define the five strategy levels by the scale of the actor, not the depth of the text. Move: one decision or routine, such as a hire, a budget line or a meeting (strategy-05, strategy-08). Unit: a team, department or project that re-reads and reworks the strategy (strategy-08, strategy-07). Organisation: the authorised text, its drafts and its tools (strategy-01, strategy-02, strategy-03). Field: peers, regulators, funders and the rankings or benchmarks that compare organisations (strategy-17, strategy-07). World: the trends and scenarios the strategy assumes (strategy-11, strategy-12). Every Entry names one level for its Trace and one for its Reading, and a Test has to come from a different level or a different hand than the Reading it tests. The layers of causal layered analysis (strategy-13: litany, social causes, worldview, myth) measure depth, not scale; they belong inside the Reading register and must not be mapped onto the five levels.
- Documents to read (the invisible literature). Read the strategy text with its siblings and not alone: earlier drafts and comments, the mission statement, the annual report narrative, board or council papers, KPI or OKR sheets, budget and procurement notices, organisation charts, job adverts, and the slide decks that carry the strategy in daily use. The literature gives three reasons. A strategic plan has directive features that act on its readers (strategy-01). Ambiguity in the text is consumed differently by different groups (strategy-03). Indicator sheets are where numbers get gamed (strategy-16). For a first study on public material the unit-level documents are the thin part; budget lines, tender notices, job adverts and published minutes are the likeliest substitutes, and each needs its own source and date as a Trace.
- Ground to check against. For a strategy Entry the ground is the pattern in what the organisation does, not a second document (strategy-05). Four public streams give it: what is funded (budget lines, grants), who is hired or appointed (job adverts, organisation chart changes), what is bought (tender notices) and what is stopped (closures, withdrawn programmes). Each is a Trace at move or unit level; the strategy's claim is a Trace at organisation level; the link between them is a Reading and cannot be a Trace. Reported figures need their own check: the UK targets regime studied in the 2006 companion to strategy-16 had central managers who did not put substantial resources into checking performance data, so a unit-level check of a reported number is itself a Trace worth taking.
- Three kinds of Test, each by another hand or at another level. (a) Coupling test: for a claim in the text, check whether a practice exists (policy-practice) and whether anything connects that practice to the stated end (means-ends) (strategy-07, strategy-06). (b) Other-reader test: someone at unit level, or their documents, translates the same passage, and the result is compared with the organisation-level Reading (strategy-08, strategy-03). (c) Fiction test: Alter writes two or three worlds, a move is tried in each by another reader, and the Test records whether the judgement of the move shifts with how it fares in each world, not whether anyone becomes more confident (strategy-11, strategy-12). Because decoupling is often designed in (strategy-06), a Reading has to say which coupling it predicts: 'this is talk' is held if coupling is absent and broke if it is present; 'this guides action' is held only if unit-level traces show it was followed.
- Failure modes to log from the start. (a) A text-action gap read as a fault, or as proof of the Reading (strategy-06, strategy-07). (b) Defensive avoidance by the host: published scenario interventions have failed because the senior team avoided the exercise (strategy-09, strategy-10); run a short receptiveness interview before any study that involves people, and give the outside reader a stop rule. (c) Fiction persuading: the controlled evidence on scenarios runs through belief in a new strategy (strategy-12), so a vivid Fiction can move people whether or not it is sound. (d) The method's own numbers being gamed: never report held or broke counts as a score (strategy-15, strategy-16). (e) Front-stage selection: public text is where talk aimed at legitimacy is most likely to sit (strategy-06), so a gap there is expected and proves little; require at least one lower-level Trace per Entry. (f) Depth taken for scale (strategy-13).
- First study. A retrospective, public-only study of one strategy whose term has ended, so outcomes are on record: three weeks, one strategy text with its indicator sheet, its later progress or annual reports, and 8 to 12 Entries across the five levels, with at least two Fictions by Alter and two Tests by Adversary. Two kinds of candidate suit it. A city strategy of the kind already analysed at Lahti (strategy-01) or Sydney (strategy-02), where a published discourse analysis gives a second reader to set against Holagram's. Or a national agenda with indicators set against global benchmarks, which suits the field and world levels (strategy-17). One verified example of the second kind: the UAE National Agenda, launched in 2014 as a seven-year agenda with six national priorities and 52 national key performance indicators that, the government portal says, 'generally compare the UAE against global benchmarks'; the Cabinet added Government Accelerators in October 2016 (u.ae, National Agenda page, undated, fetched 14 October 2026). Whether its progress reports are public, and in what form, was not checked, and no emirate-level or authority-level plan was verified.
- Care rules specific to strategy. The outside reader should hold practice in the host field (a strategy-unit or planning-unit practitioner who can call a stop), because critical discourse analysis reads strategy texts for their power effects (strategy-01, strategy-02), a strong reading mode of the kind the care rules say must be switched on and off deliberately. Keep the fieldwork block short and fixed, for example one reading day per level, and record in the failure log whenever the reading mode was on. Where the host is a live organisation, treat the study as an intervention in the sense of strategy-09 and strategy-10: agree in advance what the host receives and what is published.
- Offer the Entry inside forms that practitioners already run. A policy steering room with four walls (patterns, proven, promising, possible; strategy-18) and portfolio or sensemaking reviews hold material similar to the four registers, but with no level on each item and no separate hand for the Test. Show each wall's contents with a register stamp and a level (idea 7). The point to make to such readers is what the Entry adds, not that it replaces what they do. The vocabulary they are likely to use in reply is the OECD's anticipatory innovation governance (Tõnurist and Hanson 2020, OECD Working Papers on Public Governance No. 44, DOI 10.1787/cce14d80-en; abstract read via RePEc; not listed here).
- Synthetic hands on strategy text. Alter writes Fiction and Adversary argues Tests; neither supplies Traces. For corporate or public talk there is a first test of LLM judging: ratings and pairwise comparison both separated stronger from weaker climate reporters, and pairwise comparison held up better against text an LLM had been asked to wash (strategy-20, a preprint). Use pairwise comparison for any machine Test, try the Adversary on deliberately washed passages, and keep the washing model and the judging model apart. Word counts or phrase frequencies produced by a script are logged instruments with a human responsible, not synthetic Traces.
Ideas to take
Give every strategy Entry a coupling pair: one line for policy-practice (is there a practice at unit level?) and one for means-ends (is anything traceable from that practice to the stated end?), and require the Reading to predict which coupling it expects.
Changes: Gives the Test register a template for strategy claims, separates 'gap found' from 'Reading broken', and answers the objection that a text-action gap is expected.
Risk: The pair becomes a scorecard (strategy-15, strategy-16): a tally of coupled and uncoupled claims will be read as a verdict on the organisation. Report pairs singly, never as a count.
Build the first study around a reconstructed stream of moves instead of a second document: choose three or four public streams (funding, appointments, purchases, closures) and date each move against the strategy's term. The realised strategy is the pattern the stream shows.
Changes: Lets the diagram carry a stream node beside the document node and moves the first study from reading one PDF to reading one organisation over time.
Risk: Public records are thin at unit level, and the reader chooses the streams, so the pattern can be an artefact of the choice. Name the streams before reading the strategy and log the choice as a Reading.
Mark ambiguous phrases in the strategy text and have the Reading predict how a named group at unit level will take each one; the Test reads that group's own documents. Treat strategic ambiguity as possibly working as designed.
Changes: Adds a text-level Entry type with a built-in Test at another level, and gives the rule 'ambiguity stays' a use: an ambiguous phrase is a place where readings can be tested, not a defect to be fixed.
Risk: Over-reading: every vague phrase becomes a hidden intent. Predictions have to be written before the unit-level documents are read.
Add a receptiveness interview and a stop rule to the care rules for any study inside a live organisation, and file defensive avoidance by the host as a named failure type.
Changes: Gives the care rules two concrete steps taken from published scenario failures, and a failure category the log can describe by kind and not by score.
Risk: The interview can screen out the organisations that most need a reading, and can itself become the intervention. Keep it to a few questions and record that it was asked.
Write wind-tunnel Entries: Alter writes two or three worlds, another reader tries a named move in each and records how their judgement of the move changes, and a second map after the exercise records what readers made of the Fiction.
Changes: Gives Fiction a testing role that never makes it evidence: the move is tested in the story, the story is not tested. The published experiments suggest judgement shifts with the move's fate in each world.
Risk: The controlled evidence runs through belief in a new strategy (strategy-12), so the exercise may persuade without informing. The Adversary's Test must follow, by a different hand.
Treat the Adversary as one of several arrangements against gaming and rotate them: authority (the outside reader), contestation (the Adversary) and the two less familiar arrangements Hood and Piotrowska call 'bafflement' and 'togetherness'. Record the mix in the failure log and change it between blocks.
Changes: Gives the care rules a basis in the gaming literature and explains why one safeguard is not enough: the authors say no single arrangement works and point to changing mixes.
Risk: More machinery than a three-week study carries.
Offer the Entry to practitioners as stamps on a steering room's four walls: Patterns carry Traces, Proven carries Test results, Promising carries Readings and Possible carries Fictions, and every item gets a level and a hand.
Changes: A route into forms that governance-innovation practitioners already run, and a way to show what the Entry adds: levels, separate hands, and a Test that someone else wrote.
Risk: The mapping is mine and may be forced: the walls are kinds of evidence, the registers are the status of one claim. Practitioners may read it as relabelling. Try it on one real room before using it.
Specify a machine Test for corporate or public talk: the Adversary compares pairs of passages (the strategy's claim against a lower-level trace, or two versions of one passage), is tried first on passages an LLM has been asked to wash, and is never asked for a score.
Changes: Turns 'Adversary argues Tests' into a format that can be checked, and makes washing a known threat and not a surprise.
Risk: A preprint on climate disclosures, not strategy documents. The LLM-judge literature already held in the review (critic threads) reports self-preference and sycophancy, which the protocol has to control.
Counter-arguments
- Strategy-as-practice and discourse analysis already read strategy texts and follow them into practice (strategy-01, strategy-02, strategy-08). A reviewer from that field will ask what the Entry adds beyond a critical discourse analysis plus fieldwork. The honest answer is the rules, not the object: stated levels, separate hands for reading and testing, labelled fiction and a failure log. The first study should be built so that it can be compared with a plain discourse analysis of the same text. A 2026 perspective on thirty years of the approach (Whittington, 'Strategy as Practice: Thirty Years Young'; SSRN record seen through Crossref, abstract only, which proposes four directions for future research) is worth reading before making the claim.
- Decoupling is often designed, not a failure (strategy-06, strategy-07). A Test that finds talk and action apart has not found a fault, and a Reading that calls a strategy ceremonial cannot be broken by a gap. If Readings do not say which coupling they predict, the method can confirm anything.
- The strategy is not the document. The practitioner literature treats a strategy as a diagnosis, a guiding policy and coherent action (strategy-04, as summarised), and the research literature treats it as a pattern in action (strategy-05). Reading the text first studies the layer that matters least, and reading fluff for depth is the over-reading Holagram's own safeguards warn against.
- Planning works, on average. A meta-analysis of 87 correlations from 31 studies found that strategic planning has a positive, moderate effect on organisational performance, strongest for formal planning, in public and private organisations (strategy-19). A method that starts from suspicion of strategy documents has to account for that, and has to say why a single text can be ceremonial while planning in general helps.
- Foresight practitioners already label scenarios as not predictions and treat plausibility, not probability, as the standard (Ramírez and Selin 2014, Foresight 16(1): 54-74, abstract read; not listed). 'Fiction, never evidence' is therefore not new to them. What they lack is a Test written by someone else. They will also ask what the Entry adds to premortems (Klein 2007, Harvard Business Review, page seen), red teams (held safeguards-15) and portfolio or sensemaking reviews of what an organisation does against its strategic intent; the UNDP pages that describe such reviews were seen in search results only and are undated.
- A Test that asks for accuracy may misdescribe how strategy is made, and the method has to leave room for accounts that coordinate action while staying inaccurate.
- Metric critics will say the method creates its own metric. Any count of held and broken entries, or of failures logged, becomes a target once someone funds or ranks on it (strategy-15, strategy-16), and the four registers and five levels are categories that can be used creatively: the UK study found that new categories made to stop gaming opened new ways to game (strategy-16).
- Public-only material and three weeks are the wrong conditions for strategy. Public text is the front stage, where talk aimed at legitimacy is most likely to sit (strategy-06), so the study may find decoupling by construction; strategy effects take years; and a retrospective design brings hindsight risks (see held safeguards-12 on explanation and confidence).
- Innovation labs and offices may not be able to host the method. The literature describes policy labs as typically small, short-lived, easy to shut down or ignore, and dependent on political patronage (Lewis 2020, Policy Design and Practice 4(2): 242-251, abstract read; not listed), and a global survey found labs created partly for emulation and legitimation (Tõnurist, Kattel and Lember 2017, Public Management Review 19(10): 1455-1479, abstract read; not listed). A reading that shows a lab's own strategy as decoupled can read as an accusation, and an outside reader with the power to stop a study may not exist inside such a unit.
Gaps
- No protocol found that separates the reader of a strategy text from the tester and records the level of each claim. The nearest are critical discourse analysis (strategy-01), ambiguity consumption (strategy-03) and middle-manager studies (strategy-08), each with one hand. This is the space the Entry would fill.
- No independent ex-post evaluation of scenarios used in strategy making, for example of Shell's, turned up. What exists is insider history (Cornelius, Van de Putte and Romani 2005, California Management Review 48(1): 92-109; Kupers and Wilkinson 2014, The Essence of Scenarios, Amsterdam University Press; both Crossref-verified, not opened) and the experiments (strategy-11, strategy-12). A search for track-record evidence returned anecdote about two or three well-known cases.
- Peer-reviewed work on OKRs and goal cascades is thin and mostly practitioner. A search returned a systematic mapping of OKR literature in software contexts (seen in search results only) and one exploratory interview study of 11 academics on an OKR-based appraisal system at an Omani higher-education institution (Al-Saadi and colleagues, SAGE Open 13(2), 2023; Crossref abstract read, CC BY). Neither addresses gaming or alignment directly.
- Reading board papers or minutes as texts: nothing found in this run. Annual-report narratives are covered by the impression-management literature (Merkl-Davies, Brennan and McLeay 2011, Accounting, Auditing & Accountability Journal 24(3): 315-344; Brennan and Merkl-Davies 2022, Accounting and Management Review 26(1); Crossref-verified, not opened). Reading a mission statement as a genre is covered (Swales and Rogers 1995, Discourse & Society 6(2): 223-242, abstract read, not listed).
- LLMs detecting decoupling in strategy documents: no study found. The closest are strategy-20 (climate disclosures) and a handful of arXiv papers on 10-K sections and ESG extraction seen in search results only.
- Evaluation of government foresight: found OECD practice reports (2022 to 2025, titles seen in search results only, not opened) and a comparative review of the UK, Singapore and the Netherlands (Habegger 2010, Futures 42(1): 49-58, Crossref-verified, abstract not seen), but little on the effect of foresight on decisions.
- Finnish, Estonian and Italian items. Nothing in Estonian or Italian was searched. The English route is strategy-01 and its 2009 companion.
- Dubai and layered jurisdictions: only the federal UAE National Agenda page was verified (u.ae, undated). No emirate-level, authority-level or free-zone strategy document was checked, and Goodhart's own 1975 paper was not checked; the law is reached here through Hood and Piotrowska, Muller and a Crossref record for Campbell (1979).