Review

Step 2 of the review: what each strand holds, by question. Existing reviews come first. Terms are in frame; grades and reading depth per source are in sources. A number from an abstract or a summary is marked (secondary).

Shape: review

Strand 0: existing reviews

Nawaz 2022 reviews 23 studies of rework, 2009 to 2020, in global software development only. It sorts causes into 6 categories; communication and requirements management are reported most (secondary).
Dybå 2008 reviews agile studies; its abstract was not retrieved.
Hembree 1992 is a meta-analysis of problem-solving research; two secondary summaries put it at 487 reports.
Chalmers 2009, Macleod 2014 and Glasziou 2014 are a series on research waste.
No review of design science process models turned up, and none of what a project trail must hold.

Strand A: the process from problem to result

Peffers 2006 synthesises seven process models into six activities: problem identification and motivation, objectives of a solution, design and development, demonstration, evaluation, communication.
The order is nominal. A project may start at any activity and move outward; Peffers 2006 lists four entry points.
Objectives are inferred from the problem. They can be quantitative, in terms in which a solution would be better than current ones.
That step needs "knowledge of the state of problems and current solutions". The comparison target sits there, with no review step named.
Evaluation compares the objectives with observed results, then decides whether to iterate to design or to go on.
Communication is the last activity, and the paper may follow its structure.
Grade: convention. The model is a synthesis and a demonstration on two cases, not a test.
Mettler 2014 reviews experiments in design science and proposes an evaluation framework: questions to ask and pitfalls to avoid (secondary).
The abstract names no review step and no comparison target, so the framework sharpens evaluation without adding the missing step.
Hevner 2004, Gregor 2013 and Wieringa 2014 were located, not opened.

Strand B: reviewing before building

Lund 2016 makes the call: "no new studies should be done without a systematic review of existing evidence".
Glasziou 2014 reports, in its abstract, that investigators of new trials rarely set their findings in the context of a systematic review.
It reports that 40 to 89 percent of published trials were not replicable, because their interventions were poorly described (secondary).
The evidence measures the neglect. No source found measures that doing the review changes the outcome of a project.
Grade: convention, untested for effect; the neglect is measured, descriptive.
Webster 2002 and Booth 2016 were located as guides to reviewing; not opened.

Strand C: heuristics

Polya 1945 and Schoenfeld 1985 were located, not opened.
Hembree 1992, in secondary summaries: checking work, drawing diagrams and using heuristics go with higher problem-solving scores.
The same summaries say heuristics had a clear effect only where teachers were trained in them, and that students told to follow fixed steps scored lower than students using flexible approaches.
Hohn 2002 reports two controlled experiments with 223 elementary students (secondary).
A named five-phase strategy, taught over several lessons, did better than traditional instruction, short term and delayed.
The abstract does not say whether students followed the phases as fixed steps, so the older warning about fixed steps is not tested here.
Grade: measured, for students of mathematics, now in two controlled experiments and the summaries; transfer to a person's own project is not tested.

Strand D: frameworks and personal process

Source	What it prescribes before building	Measured
Schwaber 2020	a Product Goal, a Sprint Goal and a Definition of Done, as commitments; a sprint of one month or less	nothing in the guide
Singer 2019	shaping an idea before a bet; a pitch; a six-week cycle and a two-week cool-down	nothing in the contents page read; below the threshold on the second rating, an example of practice only
Chapman 2000	phase 1: business objectives, situation, success criteria, a project plan; the process loops back	nothing in the sections read
Ubl 2020	a design doc before coding, to find design issues while changes are cheap; an example only, below the threshold	the owner's claim
Serrador 2015	agile use against project success	correlational, 1,002 projects (secondary)
Prechelt 2001b	a taught personal process: plan, log, review	controlled experiment

Prechelt 2001b found positive effects, but smaller than proponents assume. The trained group estimated productivity better and made fewer trivial mistakes.
CRISP-DM names success criteria in phase 1. It is the closest practice to a named baseline.
Grade for the prescriptions: convention, untested.

Strand E: rework and its causes

Boehm 2001, section TWO: current projects spend about 40 to 50 percent of effort on avoidable rework. Section ONE: a late fix is "often" 100 times dearer, and 5:1 for small noncritical systems.
Section THREE: about 80 percent of avoidable rework comes from 20 percent of the defects. Two major sources are hastily specified requirements and nominal-case design.
Fairley 2005 adds that rework is not all bad: iteration needs some, and too little may signal too little review (abstract).
Hofmann 2001 quotes Jones: requirements engineering is deficient in more than 75 percent of enterprises (secondary).
Grade: the shares are expert reports. The cause, requirements specified in haste, is supported with caveat by Boehm and by Nawaz 2022 for global development.

Strand F: the trail

Noble 2009: one folder per project with data, results, doc and source; dated experiment folders; a dated notebook. The reader is "someone else", often the author later.
Wilson 2017: an overview file (3a), a folder per project (4a), a change log serving as a lab notebook (5f). Its workshops reached over 11,000 people since 2010.
Sandve 2013: ten rules, first "for every result, keep track of how it was produced".
Altmann 2002 models memory for goals and the cost of interruption, fitted to a puzzle task. Parnin 2010 on programmers was located, not opened.
Grade: convention, untested. These guides agree with each other. No study tests that a set of files lets a second person resume.

Answers to the four questions

Question	Answer	Grade
Q1 round trips	the process models name objectives and evaluation; the cause of avoidable rework is hasty specification; the benefit of a written scope, baseline or decision list is not measured	convention, untested
Q2 topics	an outline raised judged writing quality in an experiment (secondary, Kellogg 1988); a plan written as a topic is not tested	not tested; outline measured for writing
Q3 trail	practice guides converge on a README, a folder per project and a dated log	convention, untested
Q4 boundary	the DSR process ends in communication; a document's shape is outside the process	convention