TextDataView TextData on GitHub

Eisenhardt Case Methodology

Toolkit · Qualitative

An AI-assisted toolkit for Eisenhardt case-method research preparation.

What has been built

A method map, reference collection and maintenance skill support preparation (Eisenhardt 1989, 2021). Software preserved research materials during reorganization. Case analysis remains proposed.

What comes next

Implement and evaluate case selection, within-case and cross-case analysis, and theory development.

What it is for

A method map, a reference collection and a working maintenance skill support preparation for case-based theory building. Explore the method, find candidate studies, and try a reproducible example of preserving research packages during reorganization.

Works today

  • Explore eight method phases with source-linked explanations.
  • Search, filter and download 22,856 candidate-reference records.
  • Run the maintenance example offline: inventory, plan, apply and verify.

Proposed, not built

  • Case selection
  • Within-case analysis
  • Cross-case analysis
  • Theory development

These analytical capabilities remain to be implemented and evaluated.

The method, mapped

Select a phase to see what it asks, what evidence it should leave, and a common failure. The reference map explains the method; the software currently supports preparation.

  • Dashed: a phase a proposed capability would serve.
  • Cobalt ground: what works today.

The sources behind the map

Explore source notes by topic. Each states its evidence type, confidence and material limits; some entries refer to the same work.

What is core, and what is folklore

Core to the method

  • Theory building is the organizing aim; constructs alone are insufficient without relationships and why-mechanisms.
  • Cases are theoretical replications and analytic units, not observations pooled under statistical sampling logic.
  • Careful case selection creates theoretically useful similarity and difference and helps address alternative explanations.
  • Constant comparison, within-case analysis, cross-case analysis, literature iteration, boundary conditions, and theoretical arguments work as one system.
  • The desired theory balances parsimony, empirical accuracy, logical coherence, testability, and relevance beyond the immediate setting.

Not requirements of the method

  • A fixed count of cases. Four to ten is a historically common range, not a rule.
  • Interviews or qualitative data only. Archival, quantitative, observational, ethnographic, and mixed evidence may be appropriate.
  • Performance as the outcome.
  • Variance theory rather than process theory.
  • A rigid positivist philosophy or a fully detached researcher.
  • A standardized visual data structure analogous to the Gioia display.
How it differs from neighbouring methods
ApproachWhat it aims to produceWhat warrants itRelation to Eisenhardt
EisenhardtExplanatory theory: constructs, relationships, mechanisms, and boundaries across casesTheoretical sampling plus within/cross-case constant comparison and replication logicFocal method.
GioiaInformant-grounded concepts and a transparent progression from first-order terms to second-order themes and a grounded modelInformant-centric coding, data structure, and traceable movement from participant terms to researcher conceptsShares induction and iteration, but foregrounds informant meaning and coding architecture; it should not be treated as the same workflow.
YinCase-study design, evidence convergence, and analytic replication for explanatory or deductive as well as exploratory studiesCase protocol, multiple sources, replication logic, chain of evidenceSupplies much of the case and replication logic that Eisenhardt integrates into inductive theory building.
Grounded theoryTheory generated through theoretical sampling, constant comparison, and saturationIterative category/theory development grounded in dataSupplies core inductive logic; Eisenhardt adds a particularly explicit multi-case comparison and theoretical-argument architecture.
Langley/process researchExplanations of how phenomena unfold over time using multiple process-theorizing strategiesTemporal patterning, sequencing, narrative/visual/bracketing and related strategiesComplementary rather than categorically opposed; Eisenhardt 2021 explicitly embraces time and process possibilities.

Combining techniques is defensible only when the research question, assumptions, unit of analysis, and output logic remain coherent. A paper that cites all approaches but merely borrows their visible reporting devices has not established method fit.

Two neighbouring methods have their own pages in the toolkit: Grounded Theory and Gioia Methodology.

The reference collection

Browse 22,856 candidate-reference records, observed through 26 August 2026. These are records, not distinct works; duplicates and versions may remain. Inspect the linked study before treating a screening label as evidence of method adherence.

22,856 candidate-reference records
Records, not distinct works: duplicates or versions of one work may remain.
20,732 with a DOI
A persistent identifier for the publication; each DOI appears once.

Records by publication year

  • Metadata-qualified
  • Provisional

12 records have no year. Select a bar to show that year’s records.

Records by year as a table
YearMetadata-qualifiedProvisionalAll
1988101
1989000
1990415
1991426
1992213
1993538
1994549
199516925
199610616
199712517
1998161127
1999291443
2000452974
2001402161
20027338111
200310837145
20049740137
200514049189
2006164100264
2007219115334
2008288120408
2009385131516
2010444172616
2011593208801
2012665271936
20137632781,041
20147832971,080
20158513301,181
20168163381,154
20179063961,302
20181,0574181,475
20191,0344401,474
20201,0925091,601
20211,1324541,586
20227365211,257
20237375771,314
20247245701,294
20256326901,322
20265604511,011

22,856 records

  1. Gopalakrishnan Narayanamurthy, Anand Gurumurthy, Roger Moser · 2018 · Journal of Organizational Change Management

    Metadata-qualified · Confirmed case study

  2. Ganesh Prasad Shukla, Gajendra K. Adil · 2021 · Sustainable Production and Consumption

    Metadata-qualified · Confirmed case study

  3. Marina Yue Zhang, Mark Dodgson · 2007 · Journal of World Business

    Metadata-qualified · Confirmed case study

  4. Jacqui Basquill · 2018 · Lancaster EPrints (Lancaster University)

    Metadata-qualified · Confirmed case study

  5. Nicoletta Policek · 2019 · Advances in higher education and professional development book series

    Metadata-qualified · Confirmed case study

  6. Thomas Sorreda, Xavier Philippe · 2020 · Society and Business Review

    Provisional · Confirmed case study

  7. Ziad Elsahn, Frank Siedlok · 2021 · Industrial Marketing Management

    Metadata-qualified · Probable case study

  8. Bice Della Piana, Alessandra Vecchi, Vittoria Marino and one more · 2017 · International Conference on Advances in Business Management and Law (ICABML)

    Provisional · Probable case study

  9. David I. Levine, Frank Neuhauser, Jeffrey S. Petersen · 2002 · Journal of Policy Analysis and Management

    Metadata-qualified · Probable case study

  10. Anna Marrucci, Riccardo Rialti, Raffaele Donvito and one more · 2022 · International Journal of Emerging Markets

    Metadata-qualified · Confirmed case study

Download the collection

The appendix contains the same files. The manifest records their counts and checksums; source-specific terms and provenance qualifications are stated in NOTICE.md.

Try the maintenance prototype

The skill inventories a workspace, seals a move plan with a checksum, applies the approved plan and verifies the result. Completed packages move whole. Run the synthetic example locally with Python 3.9 or newer; no network or model is needed.

The software also preserved research materials during a documented reorganization. That single use and the synthetic example demonstrate preparation, not case analysis.

Two packages, one approved plan

Run “python3 example/run_example.py” from the appendix folder. The example copies two synthetic packages into a temporary workspace, checks the plan, moves them whole and verifies the result.

Selected outcomes from the runnable synthetic example
AttemptWhat the software did
Edit the plan after sealing and keep the old sealRefused: the plan no longer matches its seal.
Add a file to a package after sealingRefused: the package no longer matches the plan (167 bytes expected, 206 found).
Apply it with the exact checksumApplied and then verified: two packages moved whole and two relative links added. The plan allows no copy, edit or deletion.
Apply the same plan againRefused: the packages are no longer where the plan found them.

The output is “all_as_expected”: two renames, two relative links, zero copies and zero deletions. See example/output/SUMMARY.json and 04-refusals.json for the retained results.

Read the AI prompt

Open the Eisenhardt appendix

Use the appendix

Open the TextData repository and appendix/07-eisenhardt-case-methodology for the method map, collection, maintenance skill, runnable example and offline checks. The README links the matching AI prompt.

What it does not do

  • Case selection, within-case analysis, cross-case analysis and theory development remain proposed.
  • The collection is a bounded set of candidate records, not a census or a guarantee of method adherence. The source ledger is private, so the exporter cannot regenerate it from this appendix.
  • The map is a bounded interpretation of the literature, not a systematic review. Its software comparisons are based on documentation, not performance tests.
  • Moves are sequential, not transactional. A partial failure needs inspection before recovery. Aggregate snapshots and selected hashes can miss same-size changes; do not modify a workspace during application.
  • The example tests file maintenance, not research quality or general control of an AI agent.

Sources and credits

  • Method: Eisenhardt 1989, 2021. The toolkit is an independent project, not affiliated with the method’s author.
  • Collection sources: Crossref, OpenAlex, DataCite, OpenAIRE, OpenCitations and Semantic Scholar. See NOTICE.md for attribution, source-specific terms and the remaining API-terms qualification.
  • OpenAlex asks to be cited: Priem J, Piwowar H, Orr R (2022) OpenAlex: A fully-open index of scholarly works, authors, venues, institutions, and concepts. arXiv:2205.01833.
  • Licences, as in the repository: the code under the Apache License 2.0; the project’s documents and records, the method map, and the collection’s screening labels and compilation under CC BY 4.0, to the extent the project holds rights in them. Titles, author names and other publication metadata remain under their sources’ terms; OpenAlex data is CC0. The works cited in the map are cited, not included.