Eighteen months into a management doctorate you have a topic, a literature review of two hundred papers and a supervisor who keeps asking the same question: what, exactly, are you testing? There is no research problem an examiner would accept, no questions that data could answer, and no hypotheses with named variables. The upgrade panel meets in ten weeks.
The cost of leaving it there is not abstract. The confirmation of registration is deferred, the funded period keeps running while the design is rebuilt, and a self-funded DBA candidate pays another year of fees for a chapter that could have been written in a month. This guide sets out the chain — problem, questions, hypotheses, variables — with three worked models from the theories management doctorates actually use, and the workflow that gets the chapter drafted while the thinking is still fresh. Open a free Tesify workspace and build the research model alongside as you read.
Why management doctorates stall here
Management is a borrowing discipline. Its theories come from economics, sociology, psychology and information systems, and each arrives with its own idea of what a hypothesis is. A strategy thesis built on the resource-based view speaks of resources, capabilities and competitive advantage; an information-systems thesis built on UTAUT speaks of constructs, intentions and moderators; an organisational thesis built on institutional theory may not test hypotheses at all and works through propositions and mechanisms. Candidates read across all three, absorb the vocabulary of each, and end up with a research problem written in one language, questions in another and a conceptual model that belongs to a third. The examiner’s first task is to check that the chain holds; the fix is to build it in one language, deliberately, from the theory you have chosen.
The chain has four links, and each must be derivable from the one before it.
| Link | What it must do | Test the examiner applies |
|---|---|---|
| Research problem | State a gap or contradiction in what is known, in one paragraph, with the practical stake named | Could a reasonable scholar in the field disagree that this is a problem? |
| Research questions | Turn the problem into two to four answerable questions, each pointing at evidence | Does each question have a conceivable answer that would change the problem? |
| Hypotheses or propositions | State the expected relationship between named constructs, with direction | Is each one falsifiable, and does it follow from the theory cited? |
| Variables | Name the independent, dependent, mediating, moderating and control variables, and how each is measured | Can every variable be traced to a hypothesis, an instrument and a data source? |
Worked model 1: a strategy thesis on dynamic capabilities
The theory. Teece, Pisano and Shuen defined dynamic capabilities in 1997 as the firm’s ability to adapt, integrate and reconfigure internal and external skills, resources and competences to match a changing environment; Teece’s 2007 refinement split them into sensing, seizing and transforming. It sits on the resource-based view, from Wernerfelt in 1984 and Barney in 1991, which holds that firms differ because their resources differ and that advantage comes from resources that are valuable, rare, imperfectly imitable and non-substitutable.
Research problem. The literature agrees that sensing capability matters for SME internationalisation but disagrees about when: studies in stable sectors find weak effects, studies in turbulent sectors find strong ones, and no UK study has tested whether environmental dynamism explains the difference.
Research questions. RQ1: Is digital sensing capability associated with export performance in UK manufacturing SMEs? RQ2: Does environmental dynamism moderate that association? RQ3: Does seizing capability mediate it?
Hypotheses. H1: Sensing capability is positively associated with export performance. H2: The association is stronger under high environmental dynamism than under low. H3: Seizing capability mediates the association between sensing capability and export performance.
Variables. Independent: sensing capability, measured by a validated multi-item scale drawn from the dynamic-capabilities literature and cited by source. Dependent: export performance, measured as export intensity and a subjective performance scale. Moderator: environmental dynamism. Mediator: seizing capability. Controls: firm size, firm age, sector, prior export experience. Analysis: hierarchical regression for H1 and H2, a mediation model for H3, with the moderation and mediation tests named in the methods chapter.
Worked model 2: an information-systems thesis on UTAUT
The theory. The Unified Theory of Acceptance and Use of Technology, published by Venkatesh, Morris, Davis and Davis in MIS Quarterly in 2003, holds that performance expectancy, effort expectancy and social influence predict the intention to use a technology, that facilitating conditions predict use directly, and that gender, age, experience and voluntariness moderate those effects; the original longitudinal test explained about 70% of the variance in intention. UTAUT2, from Venkatesh, Thong and Xu in 2012, added hedonic motivation, price value and habit for consumer settings.
Research problem. NHS procurement teams are being asked to adopt AI-assisted contract analysis, adoption is uneven between trusts, and the acceptance literature has been tested mainly on consumers and private-sector employees rather than public-sector professionals whose use is mandated.
Research questions. RQ1: Which UTAUT constructs predict procurement managers’ intention to use AI-assisted contract analysis? RQ2: Does mandated use change the relative weight of social influence? RQ3: Does professional experience moderate the effect of effort expectancy?
Hypotheses. H1 to H3: Performance expectancy, effort expectancy and social influence are each positively associated with behavioural intention. H4: Facilitating conditions are positively associated with use behaviour. H5: The effect of social influence on intention is stronger where use is mandated. H6: The effect of effort expectancy on intention weakens with professional experience.
Variables. Independent: the four UTAUT constructs, using the published items adapted to the setting and reported with reliability. Dependent: behavioural intention, then use behaviour measured by system logs where access is granted. Moderators: voluntariness, experience. Controls: trust size, role, age. Analysis: structural equation modelling, with the measurement model reported before the structural model.
Worked model 3: an organisational thesis on institutional theory, without hypotheses
The theory. Meyer and Rowan argued in 1977 that organisations conform to the rules and belief systems of their environment to gain legitimacy; DiMaggio and Powell in 1983 named the three mechanisms by which they come to resemble one another — coercive, mimetic and normative isomorphism; Scott’s three pillars, regulative, normative and cultural-cognitive, are the later synthesis. Much doctoral work in this tradition is qualitative, and a qualitative thesis does not fake hypotheses. It states propositions and mechanisms.
Research problem. UK mid-cap firms adopted sustainability reporting at very different speeds in response to the same regulatory pressure, and the literature cannot say whether the variation reflects coercive pressure from regulators, mimicry of sector leaders or the professional norms of finance and sustainability staff.
Research questions. RQ1: Through which isomorphic mechanisms did sustainability reporting spread among UK mid-cap firms between the regulatory announcement and the compliance date? RQ2: How did finance and sustainability professionals inside the firms interpret and translate those pressures? RQ3: What explains the early and late adopters?
Propositions. P1: Early adopters were driven mainly by mimetic pressure from sector leaders rather than by the regulator. P2: Normative pressure operated through professional bodies and was carried into firms by newly hired sustainability staff. P3: Late adopters experienced coercive pressure as a compliance cost and decoupled reporting from practice.
What replaces the variables. Concepts with definitions and indicators: for each isomorphic mechanism, what would count as evidence of it in an interview or a document, and what would count against it. This is the qualitative equivalent of the variables table, and examiners look for it just as hard. Our companion piece on writing up qualitative management research shows what the findings chapter built on these propositions looks like.
The workflow that gets the chapter written
The chain above is a document, not an insight, and the fastest way to produce it is to write it as one structured chapter from the start. Here is the sequence supervisors use, with the Tesify step at each stage.
- Write the problem paragraph first, in one language. Pick the theory, name its canonical source, and state the gap in its vocabulary. In Tesify, open a chapter for the research design, paste the problem paragraph and the theory’s definition, and keep every later section under it so the vocabulary cannot drift.
- Derive the questions from the problem. Two to four, each answerable, each pointing at a data source. Draft them as a numbered list under the problem, so that a question with no source or no theory behind it is visible immediately.
- Turn questions into hypotheses or propositions. One per relationship, with direction. Where the theory offers a moderator or a mediator, decide now whether you are testing it, because that decision sets the sample size and the analysis.
- Build the variables table. Independent, dependent, mediator, moderator, controls; for each, the construct definition, the instrument and its source, and the data source. Generate the table in your workspace and keep it live: it becomes the spine of the methods chapter and the results chapter.
- Draw the model and read it back. Every arrow is a hypothesis; every box is a variable in the table. If an arrow has no hypothesis or a box has no measure, the chain is broken. Fix it here, not at the viva.
- Write the chapter around the table. The introduction chapter carries the problem and the questions; the design chapter carries the hypotheses and the variables table. Tesify drafts the connecting prose from your problem, questions and table, in your own register, and the Tesify AI Editor tightens the hypothesis wording so that direction and construct names are consistent from H1 to the results.
The judgement in every step is yours: which theory, which construct, which relationship is worth three years. What the workspace removes is the drift that comes from holding a problem, twelve hypotheses and thirty variables in separate documents over eighteen months. A candidate who runs this sequence over a fortnight has a design chapter for the upgrade panel; one who leaves it in notes has the same conversation with the supervisor for another term.
DBA candidates: the practitioner problem is not yet the research problem
A DBA starts from practice — Henley’s programme, for one, admits senior executives with a master’s degree and at least five years’ experience, runs over four to six years, and puts candidates through research design modules before the thesis. The recurring failure at the design stage is that the practitioner problem (our firm is losing bids) is written up as the research problem. It is not. The research problem is the gap in what the literature can explain about why firms like yours lose bids, and the questions, hypotheses and variables have to be derivable from that gap, not from the boardroom. The three worked models above all began as practitioner problems; the move that made them doctoral was the sentence that located the gap in a named theory. Our comparison of the professional doctorate and the PhD covers how the assessment differs; the design chain does not.
Build your research model in Tesify today: problem, questions, hypotheses and the variables table in one workspace, drafted into a design chapter your supervisor can read this week. The free plan is enough to build the model; the thinking stays yours.
Frequently asked questions
Does a qualitative management thesis need hypotheses?
No. A qualitative or interpretive design states propositions and defines its concepts and indicators instead. Examiners look for the same derivability — propositions that follow from the theory and questions that the data can answer — not for hypotheses forced onto a design that cannot test them.
How many hypotheses should a management PhD have?
As many as the research model has arrows, and no more. Six to ten is typical for a quantitative thesis with one or two moderators or mediators. Every hypothesis has to be traceable to a research question and to a variable in the table.
What is the difference between a moderator and a mediator?
A moderator changes the strength or direction of a relationship, such as environmental dynamism strengthening the effect of sensing capability on performance. A mediator carries the relationship, such as seizing capability transmitting the effect of sensing on performance. The two need different tests and different sample sizes.
Which theories do UK management doctorates most often build hypotheses on?
The resource-based view and dynamic capabilities in strategy, UTAUT and its predecessors in information systems, institutional theory in organisation studies, and stakeholder and agency theory in governance and accounting. Each has a canonical source that the hypotheses should be derived from and cite.
How much does Tesify cost for a doctoral researcher?
There is a free plan that is enough to build the problem, questions, hypotheses and variables table and draft the design chapter. Paid plans add capacity for a full thesis; the pricing is on the site and there is no long contract.
Will my examiners object to using an AI writing tool for the design chapter?
Examiners object to work that is not the candidate’s own. The theory choice, the problem, the hypotheses and the variables are your decisions, made and recorded by you; the tool drafts connecting prose in your register and keeps the documents consistent. Check your institution’s policy on AI assistance and declare use where it asks you to.
Is my unpublished research model safe in Tesify?
Your workspace is private to you, and unpublished thesis content is not used to train models. A research model is not participant data; if your thesis later holds identifiable interview or company data, keep that in the approved environment and use the workspace for the writing.
What if my supervisor rejects the hypotheses?
That is the process working. Because the chain is written as one structured document, a rejected hypothesis is edited in place with its variables and its arrow in the model, rather than rebuilt across separate notes. Most rejections are about derivability, and the table shows the fix.
