Eight Examined Theses Compared
Seven of these works were written to one section skeleton and the eighth arrives with its whole apparatus. Read side by side they show what a design promises, what a sample size permits, and how far a conclusion travelled past its own evidence.
The eight works, and what each one actually is
Eight examined master's works in project management sit behind this page. Seven are written to one section skeleton — the same eight headings in the same order — treated on its own at The Thesis Pro-Forma. The eighth was supplied with its complete research apparatus and is handled at A Complete Research Project, End to End. All eight were examined and passed, and every defect below is recorded as a defect in the document, never in the person who made it.
Two columns matter more than the rest. Design as claimed is what the work calls itself. Design as used is what the chapters do. Where the two differ the difference changes what the work may conclude, and it is invisible from the abstract.
THE EIGHT WORKS — DESIGN CLAIMED AGAINST DESIGN USED
| Work | Topic domain | Design as claimed | Design as used | Data from people |
|---|---|---|---|---|
| F | Stakeholder engagement between public agencies and a community that holds its knowledge collectively | Qualitative; two methods, "literature review and focus group" | Narrative review plus one focus group of four, analysed against four themes fixed in advance | 4 |
| G | Construction-related pressure and workforce safety | "Concurrent mixed-method approach" via a "triangulation convergence model" | One web questionnaire of 100, plus the work's own literature review designated the qualitative strand | 100 |
| H | Change-management practice inside project management | "A rigorous literature based study and secondary data analysis" | A narrative review with no reported search and no secondary dataset | None |
| J | Risk management in event projects | "A mix of qualitative and quantitative data collection methods", named across six sub-headings | One online survey of 46 plus two interviews. The other four named methods report no data | 46 + 2 |
| K | Agile adoption in a defence organisation's ICT projects | Systematic quantitative literature review plus questionnaire, self-described as mixed methods | As claimed. 348 records screened, 33 retained, plus a questionnaire of 17 | 17 |
| L | Lessons-learned frameworks across two repeat maintenance projects | Named four ways in four places, from "qualitative case study analysis" to "mixed method approach" | A two-project case study: 10 interviews plus five project performance metrics | 10 |
| M | Scope creep in a defence industry | "A hybrid Delphi influenced interview and survey" | Five interviews, then a 20-item survey of 16 whose panel overlaps round one by an unstated amount | 5 + 16 |
| The dataset thesis | Supplied with instrument, consent sheet, raw file, workbook, schedule and completion report | Not compared on this dimension here | A survey with 50 usable responses, every stage of it preserved | 50 |
Topic domains are the works' own, scrubbed of jurisdiction, employer, programme and community. Every count is that study's own figure, never a benchmark.
Scale: four people to one hundred
Those four tiles are this corpus's own figures. None is a requirement, and the supplied teaching source sets none. Read them as the spread of what passed.
One work states a target and reports against it: fifteen to twenty sought, sixteen achieved. That is the only stated-and-reported target in the library, and it carries no justification — no power calculation, no saturation argument, no citation. Sizing rules belong on Sample Size in Quantitative Studies.
The absences are more uniform than the numbers. Purposive, convenience, snowball, stratified, random and saturation were searched exhaustively across all seven works and return a single hit, which is a verb in an unrelated sentence.
What each design could establish, and what each one demonstrated
Each work was audited by building a claim ladder: every claim the conclusions make, set against what the study supplies for it. In every case the best-evidenced claim is more modest — and more interesting — than the one the conclusions lead with.
BEST-EVIDENCED CLAIM AGAINST THE CLAIM THAT OUTRAN IT
| Work | What the study establishes | What the conclusions add |
|---|---|---|
| F | Four people from a standing committee, asked whether four themes taken from the literature were relevant, did not disagree — and volunteered six concrete local additions | The conclusions report that "four themes emerged". The method says the themes were carried in and put to the group for endorsement |
| G | Pressure to maintain production is present in the sample and unevenly spread across roles and organisation sizes, on 100 self-reported responses | The headline prevalence figure is stated at three different values in three consecutive chapters — see When a Thesis Disagrees With Itself |
| H | The four sources most often given for a widely quoted failure rate do not evidence it, and a published fifth author found the same across five more works | The conclusions say the findings "disprove" the rate, and add a licence to proceed "with a view to success" that nothing in the study bears on |
| J | Among the 46 who answered, risk management is universally endorsed, unevenly executed, rarely reviewed and almost entirely self-taught | The headline recommendation is the work's own stated aim restated as a finding, and no survey item asks any respondent about it |
| K | The literature carries no common account of the organisation's ICT environment. The review flow reconciles: 348 screened, 315 excluded, 33 retained | The two rungs the abstract leads with are attributed to other publications rather than derived from this study — correctly, at every appearance |
| L | Rework fell 33% between the two projects — 138 items against 92 — verified against the chart's own stored values | The conclusion claims superior performance on all five deliverables, including safety, which the work's own data contradict |
| M | Five causes of scope creep supported by both rounds, and three round-one findings overturned by round two and reported as overturned | The headline novelty — that scope creep can be positive — is the fifteenth-ranked item of twenty in its own survey, and the rank is not reported |
| The dataset thesis | Five of six figures traced end to end from instrument through raw file, tabulation and chart to the thesis sentence survive intact | One fails: an engagement figure reported at 36% where the count is 15 of 50, with the chart beside it labelled 30%. See Tracing a Figure Back to Its Source |
The strongest analytical work in the corpus
One work does something no other work in this library's student corpus does, and it is set out in full here because the method transfers to any number you are asked to take on trust.
Thesis H is a literature-only study. It takes a change-failure rate that "is touted more often than not" — two in three, or 70%, quoted routinely in practice — and goes to the four sources most often given for it. It reads each in the original, quotes what each actually says, and names the transformation between what was said and what is now repeated.
- A 1995 article. Its author "never actually claimed a 70% fail rate nor implied any actual research leading to any such claim". What it states is an observation of more than 100 companies, most falling "somewhere in between, with a distinct tilt toward the lower end". An observation becomes a statistic.
- A 1993 book. What is actually stated is "our unscientific estimate is that as many as 50 percent to 70 percent of the organisations …". An explicitly unscientific estimate of a range becomes a point estimate, and the upper bound becomes the headline.
- A 2000 article. "The brutal fact is that about 70% of all change initiatives fail" — and "the authors provide no statistical findings or data to support their claim". An assertion is quoted, never evidenced.
- A 2009 consultancy article reporting a worldwide survey of 3,199 executives and one in three change efforts succeeding, again with no statistical findings or data supplied. A survey is named and its results are not shown.
- A published fifth source, used as corroboration rather than as the finding — an author who reviewed five works professing the same rate and found none "embedded with valid as well as dependable empirical substantiation".
It then names what would have been present had research been done: no data, tables, graphs, case studies or reference lists in any source, and six unreported study parameters — where the process broke down, the organisation, the industry, the country, the size, and the span of time.
Why these findings exist at all
Almost everything above is visible for one reason: raw material survived into the delivered file — a property of what these works kept, not of their quality.
- One work prints its complete raw dataset — twenty questions by sixteen respondents — so every mean and mode could be recomputed and eighteen of its own claims audited.
- One work left its chart data inside live chart objects, so a headline safety claim could be tested against the numbers behind the picture and a cost section shown to disagree with itself.
- One work printed both its instruments, which is the only reason an orphaned survey item and an appendix carrying an earlier thesis title are findings at all.
- The eighth supplies everything — instrument as issued, participant information sheet, raw response file, analysis workbook, schedule and completion report.
- Two works display nothing beyond their references, and nothing below the surface of either can be checked.
The corollary matters more than any single finding. These defects are not more common in these works than in others; they are more visible. A work that publishes less is not cleaner, only less auditable. Nothing here generalises to works that publish less — and of the three preservation routes above, only the printed dataset was deliberate.
Reading the comparison onto your own project
The corpus is most useful as a set of decisions already taken by people working under your constraints: a fixed submission window, an approval process that can overrun, and a supervisor rather than a research team.
What this corpus tells you to do next
What to carry forward
- Read the design as used, not the design as claimed. In most of these eight the two differ, and the difference is invisible from the abstract.
- One section template accommodated four participants and one hundred without changing a heading. Structure tells the reader nothing about the evidence; only the method chapter can.
- The most credible move available to you costs nothing: report the result that went against you, and put the qualification in the same chapter as the claim it qualifies.
- Audit a quoted number the way one of these works did — originals, verbatim quotation, named transformation, named missing evidence — and stop at what the sources fail to show.
- Every figure here belongs to one study, and the defects are visible only because raw material survived.
Frequently asked questions
Are these works being criticised?
No. All eight were examined and passed. Every defect on this page is recorded as a defect in the document, not in the researcher, and several are the direct consequence of constraints the works themselves state — a thirteen-week window, an ethics process that overran, a pandemic. The register throughout is what the document does and what its own evidence shows.
Why are the works identified by letter?
Because every name, institution, employer, supervisor, place, community and programme has been removed from the extracts these pages are written from. The letters F, G, H, J, K, L and M carry no meaning beyond keeping seven works apart; the eighth is referred to by what makes it distinctive, which is that its entire research apparatus was supplied with it.
How many participants does a master's project need?
This corpus supplies a range and no rule. The seven works on the template ran between four and one hundred participants, one stated a target in advance, and none reported a pilot. The supplied teaching source sets no figure. Treat 4–100 as the observed spread of what passed examination, not as guidance.
Which of the eight should I model my thesis on?
None whole; several in parts. Specific transferable moves are: the four-source audit of a quoted statistic; the epistemic rather than logistical justification for choosing a focus group; the fully reported search strategy in the systematic review; the 249-word single-paragraph abstract that names its design in the second sentence; and the 157-word gap section that explains why the literature is thin rather than only that it is.
Two works count their literature review as a data-collection method. Is that ever legitimate?
It is legitimate when the review is conducted as data collection and reported as such — a stated search, named databases, criteria, counts at each stage. Across the seven works, six rest substantially on literature searching and one reports the search. That one is entitled to call its review a strand; the others are describing background work in the vocabulary of method.
Can I quote the failure-rate finding in my own work?
Quote it as what it is: four named sources do not evidence a widely repeated rate, and a published fifth author found the same across five more works. Do not quote it as disproof of the rate, which is the one word the work's own conclusions overreach on. Note also that three of the audited sources are absent from that work's reference list, so you would need to reach the originals yourself.
References and source attribution
- Eight examined master's works in project management, supplied as student work and used here as observed practice rather than as model answers. Researchers, supervisors, institutions, jurisdictions, employers, communities and programmes scrubbed. Seven are written to one section template; the eighth was supplied with its complete research apparatus.
- The consolidated extract of four of those works (a focus-group study, a construction safety survey, a change-management literature study and an event risk survey), including every heading with its section length, every reference list audited entry by entry, and every quantitative claim checked against the printed evidence.
- The consolidated extract of three further works (a systematic quantitative literature review, a two-project case study and a hybrid Delphi study), including a complete raw dataset of twenty questions by sixteen respondents, recomputed mean by mean and mode by mode.
- The supplied teaching source: weekly study notes, slide decks and assessment activities for a master's-level research methods subject in project management, which sets a 300-word abstract guidance, names the pilot study four times, and teaches no focus-group, paradigm or systematic-review content. Author, institution and year not stated in the supplied files.
Suggested questions for Ask KEVOS
- How do I tell whether my design as written matches the design I actually ran?
- What should a claim ladder for my own conclusions look like?
- How do I audit a widely quoted statistic before I cite it?
- What sample size did examined master's theses in project management actually achieve?
- How do I justify choosing a focus group over interviews?
- What should I display in my appendices so a reader can check my analysis?