Choose a research question, inspect the steps and start with a compatible example. Each lab connects its inputs to a declared calculation and an interpretable result.
Define the problem, select a suitable published method or clearly identified application rule, record the data and assumptions, calculate, challenge the result and retain a reproducible record. Specialist review can then assess the exact analysis and its evidence.
01 · Trace the calculation
Method variants, directions, weights and settings affect the result. Keep them with the source data.
02 · Challenge the conclusion
Check uncertainty, holdout performance or sensitivity where the lab supports it. Missing evidence remains visible.
03 · Read the review status
Published sources and software checks have a defined scope. Independent validation and expert approval require their own evidence.
A scoped current/target assessment with evidence dates; asset risk scenarios; control costs, coverage judgements, dependencies and a budget.
How it works
Organize evidence under the six CSF functions, calculate declared weighted priorities, then compare feasible control bundles. Recalculate after weight and budget changes.
Get
Known gaps and unknowns, explained remediation priorities, a feasible portfolio or an explicit infeasibility finding, and a downloadable decision record.
Import the project JSON or use the lab’s example button. Replace illustrative assets, evidence, costs and coverage estimates with your scoped records. Up to 12 scenarios and 18 candidate controls.
FIRST EPSS daily data — join by CVE and retain score date/model version. This external likelihood signal is not an organization’s breach probability.
Review relevant records and record evidence references in the project. Raw feeds are not directly imported or automatically mapped to 1–5 ratings. Source conditions
NIST CSF 2.0 and SP 800-30 inform organization and risk context. The weighted priority and coverage rules are declared application rules; their outputs are relative decision aids, not NIST scores, compliance certificates or measured risk-reduction percentages.
02 / Predictive evidence
Prediction & Decision Lab
What can the data predict, and how can that evidence support a decision?
A tabular CSV with a named target, meaningful input features and independent observations. Use classification for classes or regression for a numerical outcome.
How it works
Python runs in this browser. Preprocessing fits inside training folds; cross-validation compares models. A separate final holdout evaluates the selected approach.
Get
CV comparisons, holdout measures, new-row predictions and reproducible Python exports. Declare directions and preferences before passing a decision matrix to the decision labs.
Inspect repeated measurements or shared batches before treating rows as independent. Both source pages specify CC BY 4.0; retain dataset authors, DOI and transformations.
Current limits: 2 MiB, 2,000 rows and 30 features, with at least 40 usable target rows. The available split assumes independent rows; grouped or time-ordered studies need a suitable validation design. Model importance is not automatically a decision weight.
03 / Criterion importance
MCDM Weighting Lab
How should criterion importance be represented for this decision?
Defined criteria and ranges; genuine expert judgements for a subjective method, or a comparable numerical matrix for a data-derived method.
How it works
Choose the named formulation and enter its required comparisons, ratings or data. Inspect normalization, aggregation and the method’s applicable diagnostics.
Get
A normalized weight vector, calculation trail and method-specific diagnostics. Export the compatible weighting report for Ranking or Sensitivity.
The matrix supports entropy, CRITIC, standard-deviation and the data component of a hybrid. The worksheet is a collection aid: transfer direct allocations or FWZIC ratings manually. AHP needs pairwise comparisons; ROC/SWARA need ordered inputs.
External data · Preparation needed
Practice data-derived weighting
World Bank development indicators can supply country-by-criterion observations. Select a common reference period and comparable definitions; reshape the selected observations into the matrix template and handle missingness explicitly.
Statistical variation or conflict does not establish stakeholder importance. Public indicator data cannot replace expert elicitation. Check metadata and reuse conditions
The generic CSV importer accepts 3–200 alternatives and 2–10 criteria. Directions are required; record units and the source in the study. A numerical consistency diagnostic alone does not validate the decision’s preferences.
04 / Alternative comparison
MCDM Ranking Lab
How do the alternatives compare under an explicit preference model?
A numerical decision matrix with comparable criteria, benefit/cost directions, a justified weight vector and a suitable ranking formulation.
How it works
Validate the matrix, declare weights, and apply the selected normalization and aggregation rule. Compare method results with their assumptions visible.
Get
Method-specific scores, ordered alternatives, intermediate calculations and an exportable report. Follow with sensitivity testing before interpreting a narrow lead.
Import the CSV, then enter weights or explicitly select equal weights. The file does not assign preferences. For the existing workbook case, inspect the numerical example with its separate audit.
External data · Preparation needed
A country comparison study
World Bank development indicators can support a scoped comparative question. Define countries as alternatives, choose the same period, justify criterion directions and record measurement definitions.
Reshape to the template; do not treat missing values as zero or rank an arbitrary collection of indicators as a universal country-quality score. Source conditions
CSV imports support 2–200 alternatives and 2–20 criteria. Scales, compensation assumptions and the chosen method affect ranks; cross-method agreement is useful evidence with a limited scope.
05 / Decision stability
MCDM Sensitivity Lab
Which plausible changes could alter the conclusion?
A baseline matrix, directions, weights and ranking method, plus defensible weight changes, data ranges or other supported scenarios.
How it works
Recalculate the declared ranking across the configured scenarios. Compare ordering, leading alternatives and the changes responsible for reversals.
Get
Rank changes, scenario comparisons and uncertainty summaries within the selected experiment. Retain the baseline and scenario settings with the findings.
Import the matrix and declare weights/method, or import a compatible saved Ranking project. Set scenario ranges after reviewing the criterion meanings.
External data · Preparation needed
Challenge a real comparison
World Bank development indicators provide source observations and metadata for a prepared country comparison. Keep the common-period baseline; use documented revisions, measurement bounds or justified scenarios.
An unrelated year is a new context, not automatically a measurement-error interval. State the reason for every range or distribution. Source conditions
Use 2–50 alternatives and 2–20 criteria. A stable result is conditional on the scenarios tested; it does not establish correctness outside their scope.
06 / Explanation & scoped checks
CECE Research Lab
What source conditions and recorded checks support a benchmark comparison?
Select the supplied supplier benchmark to inspect its archived explanations, or bring your own decision matrix and weights for the separate exploratory calculator.
How it works
The explorer displays saved pairwise evidence and scoped checks from a pinned research release. The own-data calculator applies its declared exploratory formulation.
Get
Benchmark explanations with source/provenance details, or an own-data ranking report. Keep each result’s method and scope visible.
The JSON is for inspection and reproducibility, not an arbitrary project import. The calculator accepts its two-sheet Excel template or manual entry: 2–50 alternatives and up to 20 criteria.
The CECE adaptation changes the downstream pipeline; inspect its provenance rather than treating original ranks as ground truth. CECE materials v2.0.0 archive supplies the versioned research record and reuse details.
Own-data rankings do not automatically receive CECE certificates. The explorer’s archived checks apply to the named finite benchmark and supplied conditions; they are not a platform-wide certification.
07 / AI system evidence
Trustworthy AI Lab
What evidence supports confidence in this AI system for its intended use?
A named system/version, intended use, requirements and evidence references. Optionally add held-out binary labels, predictions/probabilities and relevant group/cohort labels.
How it works
Record component judgements and evidence gaps. Where prediction outputs support it, calculate classification, calibration and subgroup measures against declared criteria.
Get
An evidence profile, technical measurements, critical unmet/unknown requirements and an auditable assessment record with its actual review status.
Required: y_true (0/1) and complete y_pred (0/1), y_prob (0–1), or both. Optional: group,cohort,id. Define the positive class. This six-row example is too small for substantive assurance claims.
External data · Model outputs needed
Build evaluation evidence
UCI Banknote Authentication offers a small binary research task under CC BY 4.0. Train a model with a documented split, then export predictions on untouched test cases into the output template.
Raw banknote feature rows are not directly compatible with this assessment importer. Do not invent demographic groups or treat this task as evidence about an unrelated deployment.
Prediction CSV limits: 20,000 rows and 2 MiB. Privacy, human oversight and organizational practices need separate evidence; a prediction dataset alone cannot establish them. Small group denominators limit interpretation.
08 / Whole-decision assurance
Trustworthy MCDM Lab
Do the context, data, preferences, constraints and evidence support this decision?
A decision matrix and weights, purpose, stakeholders, constraints, evidence references and a declared ranking/sensitivity plan.
How it works
Check feasibility and numerical assumptions, calculate the declared ranking, inspect sensitivity and review evidence coverage across the decision process.
Get
A decision record linking feasible alternatives, ranks, stability and unresolved evidence requirements. Imported provenance remains distinguishable from independently verified claims.
Import the CSV, then set weights, intended use, constraints and evidence in the lab. It supplies none of these judgements or claims automatically. Compatible saved projects can preserve more context.
External data · Context required
Document a real comparison
World Bank development indicators can support a prepared country-by-criterion matrix. Keep source metadata and transformation records; add stakeholder preferences, feasible-choice constraints and evidence for the actual decision.
The data provider does not endorse the resulting ranking or supply the decision’s normative assumptions. Source conditions
CSV inputs support 2–50 alternatives and 2–20 criteria. Weight provenance, fairness of criteria, uncertainty bounds and domain validation require evidence beyond a final numerical vector.
09 / Software engineering decisions
Software Engineering Decision Lab
How do the releases compare, which tests matter first, and which improvements fit the available effort?
Choose one workflow or combine all three; only enabled workflows need inputs. Quality needs comparable release metrics, units, directions, anchors and evidence. Testing needs component risks, test coverage, durations and a sequential test-time budget in minutes. Backlog planning needs improvement priorities, effort, dependencies and capacity in person-days.
How it works
Normalize quality metrics against fixed declared anchors, while enforcing hard gates separately. Select an exact feasible test subset by unique priority-weighted component coverage; order selected tests with a dependency-aware additional-coverage-per-minute heuristic. Select improvements under person-day capacity and examine changed assumptions.
Get
Explained quality comparisons and evidence gaps, a resource-aware test plan, improvement priorities and a feasible portfolio or explicit infeasibility finding. Download the project and calculation record.
Import the example project JSON or select the lab’s example button. Enable quality comparison, testing, backlog planning, or any combination; inactive sections do not need dummy data. Replace illustrative measurements, test coverage, durations, priorities and effort estimates with records from the same stated system and assessment context. The file is a decision project, not source code to execute. Limits: 8 quality criteria, 50 alternatives, 12 components, 18 candidate tests and 18 improvement items.
Missing quality measurements and failed hard gates leave an alternative unranked. Metrics, component risks and improvement value are separately declared inputs. Test-component coverage is a planning input, not proof of fault detection. Exact subset selection does not imply a globally optimal test order; the order uses a stated heuristic. Scores are preference indices, not defect probabilities or proof of release readiness.
External research data · Prepare first
Use traceable software evidence
Defects4J: reproducible Java bugs and test infrastructure — pin a project, bug and revision. Produce comparable metrics, measured test durations and a coverage map in an appropriate external environment before entering the decision project. Raw Java repositories are not directly imported by this lab.
Historical bug labels, repeated files and commit histories need contextual review. They do not supply your stakeholder weights, acceptance gates, future defect probabilities or improvement costs. Keep transformations and exclusions with the exported project. Source conditions
Scientific context: ISO/IEC 25010:2023 provides a product-quality reference model; ISTQB CTFL v4.0.1, section 5.1.5 discusses risk, coverage and dependencies in test prioritization; NIST SSDF v1.1 supplies secure-development context. Scoring, optimization objectives and thresholds are declared application choices. The lab does not implement an official ISO score or issue ISO, NIST or ISTQB certification.
10 / Organizational systems and evidence
Information Systems Decision Lab
Which system fits the organization, how do its processes perform, and what supports user adoption?
Enable selection, process diagnostics, adoption assessment, or any combination. Selection needs requirements, fixed metric anchors, mandatory thresholds and lifecycle cost estimates. Process diagnostics needs case IDs, activities and time-zoned timestamps with declared start/end activities. Adoption needs a declared item/construct map, actual response data and separately documented usage counts.
How it works
Compare eligible systems with transparent weights and fixed scales; check nominal lifecycle cost against the budget and report discounted costs separately. For observed process cases, examine boundary completeness, ordered variants, repeated activities and elapsed recorded gaps. Reverse declared survey items and summarize complete responses per construct; qualify internal-consistency statistics and keep observed use separate from intention.
Get
A system shortlist and explained ranking, cost and weight scenarios; process timing profiles and data-quality findings; construct-level response profiles, descriptive group results and usage evidence. Export the native project and calculation record. Inactive workflows do not need dummy data.
The full project defines criteria, system costs, process boundaries, survey items and scale settings. Event CSV columns are case_id,activity,timestamp; timestamps require an explicit time zone. Response CSV columns are respondent_id,group followed by the current item IDs. Blank answers remain unknown. Import a CSV into its matching workflow after defining its context.
All example cases, respondents, evidence and cost estimates are synthetic. The illustrative questions are not a validated DeLone–McLean or TAM instrument. For a real study, use genuine responses, document the exact questionnaire and permissions, and assess measurement suitability for the intended context. Replace the vendor evidence and costs with dated, comparable records.
Current bounds: 8 selection criteria, 50 systems, a 1–10-year horizon, 10,000 events and 2,000 process cases, 28 items across up to 7 constructs and 2,000 respondents. Imports are limited to 2 MiB. A nominal lifecycle budget is distinct from the discounted cost reported alongside it.
Public process research · Prepare first
Prepare scoped event evidence
BPI Challenge 2020: Domestic Declarations records university travel-expense claims. Its source record identifies CC BY-NC 4.0; retain attribution and follow its noncommercial-use conditions. Check provider availability and current metadata before reuse.
Convert a scoped XES extract into the native CSV; map the case identifier, activity label and timestamp, preserve time zones and choose lifecycle semantics explicitly. Select a declared observation window and whole cases within the row/case limits; never sample random event rows that truncate a case. Keep unfinished and ambiguous cases visible in the data-quality findings. The full source exceeds this browser lab’s limits and is not directly imported. Preserve source version, filtering, exclusions and the meaning of the chosen start/end activities.
For system selection, collect actual organizational requirements, integration constraints and vendor cost/support evidence. For adoption, collect authorized survey responses and operational usage evidence; a process benchmark or public feature table cannot supply those judgements. Source conditions
Scientific context: the DeLone–McLean IS Success model and Davis’s user-acceptance research motivate constructs; they do not validate this implementation’s questionnaire or create a universal success score. Incomplete or tied process sequences remain visible and are excluded from completed-case comparisons. Elapsed event gaps are not separate processing or waiting times. Repeated activities are candidates for investigation, not proof of error. Cronbach’s alpha, when available, is a qualified internal-consistency statistic, not evidence of causal impact, validity or ROI.
11 / A documented starting dataset
Data Quality & Preparation Lab
Which values need attention, and what exactly changed during preparation?
A comma-separated tabular CSV with unique, nonempty headers. Record the source, observation context and intended use, then declare each column’s type, unit, required status, permissible range or labels and missing-value tokens. Defaults treat columns as text; the lab does not infer the correct measurement meaning.
How it works
Profile the original records against declared rules. Apply only selected trimming, missing-token normalization, exact label replacements and explicit affine numeric unit conversion. Optionally remove rows missing named fields and deduplicate exact rows or complete keys. Conflicting keys keep all records for review by default; selecting the first or last source record is an explicit rule. Inspect searchable findings with original and prepared values, and interpret unit conversions separately from quality changes.
Get
Before/after missingness and declared-rule validity counts, duplicate summaries and numerical profiles. Download prepared CSV, original data, a complete findings CSV, project JSON, an audit record or an HTML report. Original row lineage, conflicting-key groups, transformations and exclusions remain available. There is no imputation or automatic outlier deletion; factual accuracy and suitability need contextual evidence.
The project carries the original CSV and declared schema/cleaning choices. The CSV alone requires you to define those rules. Examples are fictional teaching records; they do not establish scientific findings. Save project and audit files explicitly to retain work; dataset contents are processed on this device.
Current bounds: 2 MiB CSV, 5,000 data rows, 40 columns and 100,000 cells, with up to 4,000 characters in a cell. A valid CSV may still need the target lab’s special headers and shape. Spreadsheet-safe exports may prefix formula-like text; use the explicitly labelled exact-data export when preserving machine-readable values is required.
Original public data · Prepare first
Keep meaning and provenance with the values
UCI Iris provides a small tabular teaching source. Add unique column headers when needed, retain measurement units, source version and dataset citation, and retain attribution under the source’s CC BY 4.0 licence.
World Bank indicators require a scoped selection, reshaping and explicit country/year/indicator definitions. Preserve units and metadata; a blank value is not automatically zero.
Choose source-appropriate checks. Unit conversion requires a known common unit and a declared factor/offset; a file cannot establish that its unit labels are factually correct. For prediction, keep any learned preprocessing within the training/validation protocol; do not use holdout outcomes to choose cleaning rules. Source conditions
Scientific context: Wang and Strong’s Beyond Accuracy treats data quality as context dependent. NIST boxplot guidance supplies IQR fence context. This release uses type-7 interpolated quartiles and 1.5-IQR flags when supported by the numeric sample; these are diagnostic application choices. The original and prepared profiles use the same declared target rules: a conversion can change a range flag because target units describe the prepared values. A changed count does not by itself establish better data quality. No universal quality score, accuracy certificate or independent expert approval is inferred.
12 / Study design and statistical evidence
Research Statistics & Study Design Lab
What does the study estimate, and what uncertainty remains?
Declare a research question, descriptive/independent/paired/observational design, and exploratory or prespecified analysis status. Supply a CSV with unique headers and selected numeric columns. Welch comparison needs a numeric outcome and two declared group labels; paired analysis needs two measurements in the same row for the same unit. Pearson correlation needs two comparable numeric variables. Precision planning needs a mean or proportion, desired margin and planning SD or proportion; it requires no CSV.
How it works
Choose one supported formulation and run the fixed Python/SciPy engine on this device. Descriptive means use a Student-t confidence interval; independent groups use Welch’s t-test and a mean-difference interval; paired rows use the difference between the selected columns; Pearson reports linear correlation with its supported confidence interval. Exclude missing observations only for the variables needed by that analysis and retain source-row reasons. Invalid present numeric values stop the run. Normal-approximation precision planning rounds the calculated sample size upward.
Get
Descriptive summaries, an effect estimate with a confidence interval where supported, a two-sided p-value when applicable, or a precision sample-size estimate. Retain used/excluded row counts, method and runtime versions, assumptions and the exact input project. Download the audit JSON, summary CSV, HTML report and fixed Python engine for reproduction. Precision planning is not power analysis and does not establish adequate design for a particular study.
Import the project to retain its selected variables, groups and confidence level. Importing only the CSV requires declaring those settings again. All observations are synthetic teaching data. The paired workflow assumes row-matched measurements; an ID column alone does not establish a correct pairing.
Current data bounds: 2 MiB, 5,000 rows, 40 columns and 100,000 cells. The same-origin Python/NumPy/SciPy runtime has a substantial first download, runs in a dedicated browser worker and executes only the fixed analysis workflow. Dataset values are not sent to an analysis server. Save a project explicitly to retain work.
Public research data · Prepare first
Prepare a small, meaningful comparison
UCI Iris is identified by its provider as CC BY 4.0. Add descriptive headers and retain the authors, version, provider corrections and centimetre units. For an independent comparison, select two species as the declared group labels and one measurement as the outcome; other group labels are excluded with a recorded reason. These observations do not establish that species membership causes a measurement difference.
For a paired study, collect authorized paired observations of the same measurement scale. For any inferential use, document the sampling process, independence between units, outcome definition and intended population. Repeated measurements, clustered records and cross-validation folds are not automatically independent observations. Do not apply an independent-group Welch test to cross-validation fold metrics by default. Clustered, time-dependent, weighted or complex-survey designs need a suitable analysis outside this bounded release. Prepare data with the Data Quality Lab while keeping exclusions and units visible.
Provide one flat fact CSV with unique headers, a documented row grain and comparable measures and units. Define up to six KPIs using sum, mean, count or ratio-of-sums, with an optional scale multiplier and higher/lower/neutral target direction. Optionally declare one calendar-date column, one category dimension, inclusive date bounds, selected category values and current or alternative target thresholds.
How it works
Filter the source records explicitly, then calculate each KPI from its own contributing records. Ratios use paired-complete numerator and denominator cells and divide their sums; missing cells are not zero. Totals and time/segment views are recomputed from the underlying rows rather than by averaging group ratios or adding group means. A target gap is observed value minus target; target status follows the declared direction. An alternative threshold is evaluated against the same observed value. Target comparisons apply only to the filtered total; segment and time views are descriptive. Targets are not automatically prorated or rescaled when the filter scope changes.
Get
A scoped KPI dashboard, segment drilldowns, time summaries and accessible charts/tables, with current and alternative-target gaps. Retain included/excluded source rows, metric-specific contributing and missing counts, ratio numerators and denominators, filters and assumptions. Download a summary CSV, project/audit JSON and standalone HTML report. Zero-denominator ratios are unavailable. A target scenario is not a forecast, causal analysis or simulated change in business operations.
The synthetic example supplies a fact table, measure definitions and target context. Import the project to retain the configuration; importing only the CSV requires declaring the KPI rules and scope again. All business records are fictional, not customer performance findings.
Current bounds: 2 MiB, 5,000 rows, 40 columns, 100,000 cells, six KPIs, 50 segment groups and 120 time bins. Dates use actual YYYY-MM-DD calendar values. When no date filter excludes them, rows without dates may contribute to overall totals but remain unassigned to time bins, with a reported count. This release does not join tables, connect to external BI accounts or implement a general database model.
Public indicators · Prepare first
Prepare a comparable indicator extract
World Bank indicators provide original definitions, units, periods and source metadata. Choose a small comparable extract, remove provider metadata rows during documented preparation and reshape to one clear row grain. Check each indicator’s use conditions, missing-value conventions, revisions and coverage. A count of indicator records is not a count of people or transactions.
Do not sum percentages, stocks or rates across countries or periods merely because they are numeric. Declare a justified aggregation and denominator; a ratio of available sums only describes its paired-complete contributing rows. Do not mix currencies, price bases or unequal observation frequencies without an explicit preparation decision. Use the Data Quality Lab for documented preparation and record all source conditions.
Read the case, source and import requirements alongside each file. Use the lab-specific guides above to check compatibility before importing. Archived examples and synthetic exercises retain their labels.
XLSX
Decision matrix workbook
The supplied CECE matrix template, also linked from the weighting, ranking and sensitivity labs.
Illustrative judgements and synthetic data are examples of use, not collected expert responses or outcomes from a platform customer.
Prepare once, interpret carefully
A dataset link is a starting point.
File
Where it fits
What to complete
Research statistics CSV / project JSON
Research Statistics & Study Design Lab
Declare the analysis, design, numeric columns, groups or row pairing and missing rules. Precision sample-size planning uses declared assumptions without a CSV.
Business fact CSV / project JSON
Business Intelligence & Performance Lab
Declare row grain, units, KPIs, date/dimension filters and targets. Measures are computed from contributing records; arbitrary multi-table models are not imported.
Tabular preparation CSV / project JSON
Data Quality & Preparation Lab
Unique headers and bounded rows/cells; explicit types, units, missing tokens, validation and transformation rules. Prepared CSV must still match the destination lab’s format.
Decision matrix CSV
Weighting’s data methods, Ranking, Sensitivity, Trustworthy MCDM
Alternative header; Direction row; numeric cells. Enter weights where required; record source, units and assumptions.
Expert collection worksheet
Manual preparation for direct allocation / FWZIC
Collect actual judgements and rationale; transfer to the matching method editor. This worksheet is not directly imported.
Binary outputs CSV
Trustworthy AI technical assessment
Held-out truth and complete predictions/probabilities; class definitions, evaluation plan and system evidence.
Training / new-row CSV
Prediction & Decision
Target in training; matching input features in new rows. Inspect dependence, leakage and validation suitability.
Two-sheet Excel template
CECE own-data calculator
Decision Matrix and Criteria sheets; directions and weights. Arbitrary CSV is not the calculator’s import format.
Cybersecurity project JSON
Cybersecurity Decision Lab
Scoped evidence, ratings, scenario IDs, control coverage, dependencies and budget. Raw CVE feeds need contextual preparation.
Software engineering project JSON
Software Engineering Decision Lab
Inputs for enabled workflows only: release metrics/gates, test risks/coverage/durations, or backlog effort/dependencies/capacity. Raw repository files and research datasets need preparation.
Information systems project JSON
Information Systems Decision Lab
Inputs for enabled selection, process and/or adoption workflows; keep criteria, lifecycle cost assumptions, boundaries, scale and item definitions with the data.
Event-log CSV
Information Systems process diagnostics
case_id,activity,timestamp with explicit time zones. Declare start/end activities, observation scope and lifecycle interpretation; raw XES is not directly imported.
Survey response CSV
Information Systems adoption assessment
respondent_id,group and exact item-ID columns. Define constructs, reverse items and scale first; genuine missing answers stay blank. Usage evidence is entered separately.
Source, version and use conditions.
Source links and published access conditions were reviewed on 7 November 2026. External datasets are linked rather than copied. New guide examples are clearly marked synthetic; the template notes explain permitted reuse and compatibility.
For UCI datasets, retain the authors and dataset DOI under the licence shown on the specific source page. For World Bank data, consult the summary terms and each indicator’s metadata: the default is CC BY 4.0 with additional terms and possible third-party conditions. The CISA mirror licence identifies KEV data as CC0; linked third-party materials retain their terms. For NVD and FIRST, follow their published access/use guidance and preserve retrieval dates. A source’s availability does not imply its endorsement of this platform.
Defects4J publishes its framework under the MIT licence; inspect the selected subject project and dependency terms separately. The SmartSHARK PROMISE challenge repository states Apache-2.0 and requests citation of Trautsch and Herbold’s versioned data record (DOI 10.5281/zenodo.8370181). Preserve notices and source-specific conditions. ISO and ISTQB materials are linked for reference; their contents retain their original rights.
The BPI Challenge 2020 Domestic Declarations record identifies CC BY-NC 4.0. Its noncommercial restriction and attribution requirements apply to that external data; no BPI files are redistributed here. Other BPI releases may have different terms. Research-model citations do not grant permission to reproduce every published questionnaire. Use appropriately authorized response data and retain instrument/version, consent scope, item coding and reverse-scoring decisions.
Retain a data snapshot, citation, licence, units, missing-data decisions, transformations and exclusions with the final analysis. Use only data appropriate for the stated purpose and authorized for that use. More on reproducible attribution
Ask DeciMethod
YOUR RESEARCH GUIDE
Ask DeciMethod
Clear answers. Your next step.
What is DeciMethod?
DeciMethod brings research methods, scientific laboratories, assessed learning, expert services and scholarly publishing into one research environment. Start with your question, select a suitable workflow and retain the evidence behind your decisions.
Researchers, postgraduate students, lecturers, professional teams and institutions can use the platform. Each lab and service explains the knowledge, data and scope needed to use it appropriately.
The labs use declared methods and structured research workflows. Read the scientific sources, assumptions and calculation steps for the tool you choose. Human expert review is a separate service; an automated result does not itself mean scientific approval.
Choose the research task you want to solve, then read the lab’s input requirements and worked example. If the question is still unclear, use a research consultation to define the next step.
Seven browser labs are free releases. The six connected research labs offer a seven-day trial for one project, without a card or automatic trial charge. Academy preparation and the Foundation route have separate access rules.
Yes. You can combine compatible lab workflows, study the methods in Academy, seek a defined expert contribution and prepare evidence for a journal manuscript. Each service retains its own access, scope and review requirements.
Start with the task: prepare data, estimate or predict, evaluate systems, weight criteria, rank alternatives or test a decision. The lab guide describes the inputs, steps and outputs for all thirteen labs.
Inspect missing values, duplicates, declared types, ranges and labels. Preview cleaning and unit conversions while preserving the original dataset and a record of changes. Bring a small tabular CSV with clear column meanings.
Use supported workflows for descriptive estimates, independent or paired mean differences, Pearson correlation and precision sample-size planning. Keep the study design, exclusions and interpretation beside the numerical result.
Define KPIs from a documented fact table, inspect segments and time summaries, and compare results with declared targets. The lab keeps coverage and assumptions alongside the performance report.
The Prediction & Decision Lab runs Python in your browser for supported classification and regression tasks. It separates cross-validation from final holdout evaluation and can export reproducible Python work.
Organise security evidence, prioritise remediation and compare control portfolios within a declared budget. Unknown evidence remains visible. These are scoped decision aids, not automatic compliance certificates.
What is the Software Engineering Decision Lab for?
Compare releases against quality gates, prioritise tests within available time and plan fixes or improvements within capacity. Bring comparable metrics, test coverage, effort and dependencies.
Can I compare information systems or analyse processes?
The Information Systems Decision Lab supports system comparisons, process diagnostics and adoption profiles through separate workflows. Bring requirements and cost evidence, event logs or genuine questionnaire responses as appropriate.
The Weighting Lab supports declared formulations using expert judgements, performance data or explicit hybrid choices. Select a method that matches your research question and inspect its inputs and diagnostics.
Use a numerical decision matrix, justified weights and benefit or cost directions in the Ranking Lab. Inspect the chosen formulation and intermediate calculations, then test sensitivity before interpreting a close result.
Use the Sensitivity Lab to vary supported weights, data and modelling assumptions under declared scenarios. Stability is conditional on those scenarios; it does not establish correctness beyond them.
Inspect archived benchmark explanations and scoped checks, or use the separate own-data calculator. Archived evidence applies to the named benchmark and conditions; your own calculation does not automatically receive a CECE certificate.
Record an AI system’s intended use, evidence and gaps. Supported prediction outputs can provide technical measures, while privacy, oversight and organisational evidence require their own records.
Connect the decision purpose, constraints, data, preferences, ranking and sensitivity in one evidence record. Imported claims remain distinguishable from independently verified evidence.
Formats and limits differ by lab. Check its import instructions and template before uploading. A CSV for one lab is not automatically compatible with another, and a larger file is not necessarily suitable for browser analysis.
Yes. The lab guide links compatible synthetic examples, templates and selected external sources. Synthetic examples are for practice; preserve provenance and reuse conditions when working with external data.
Compatible connected workflows can pass declared matrices, weights and saved results between labs. Check units, directions, alternatives and assumptions before accepting a handoff. Model importance is not automatically a decision weight.
Use the export options provided by the selected lab. Available reports, datasets and reproducibility files vary by workflow. Save the method, input data and settings together with the output.
The lab catalogue separates the thirteen available labs from planned additions. Check that distinction before selecting a tool or requesting work. A roadmap entry is not an active service.
Read the lab’s methods and references and the central methods page. They describe published sources, named variants, application rules, assumptions and the scope of implementation checks.
Does every result have independent scientific approval?
No. A method’s publication, a software check and an independent review of your exact analysis are different forms of evidence. Expert approval requires a separate review and its recorded scope.
Retain the original data and provenance, preprocessing steps, method variant, weights, criterion directions, parameter settings and exported outputs. Record exclusions and changes so another researcher can follow the same workflow.
Cite the underlying scientific sources and identify the tool, method variant and version used where available. Follow your journal’s software and data citation requirements; using the platform does not replace citing the original method.
Does agreement between ranking methods prove a result?
Agreement can be useful evidence but is not proof. Examine the methods’ assumptions, data quality, preferences and sensitivity, and explain why the selected approach fits the research question.
Can the platform invent missing data or expert responses?
Research evidence must come from legitimate sources and actual participants where required. Do not present synthetic examples or simulated judgements as collected research data. Record missingness and limitations explicitly.
Public guides and free examples can be explored without a saved account workflow. Private projects, requests, course records and assessments require the sign-in shown on their pages.
Open My research workspace using the same account that created the project. Classroom submissions, Academy attempts and service requests have their own workspaces.
New analysis and editing in connected labs require active trial or paid access. After access ends, existing saved work remains available to view and download. Check your plan and project status.
The trial covers one project. Active research plans support up to five active projects, with separate limits on retained projects, versions and storage. The pricing page lists the current limits.
Archiving frees an active-project slot while retained records still count toward storage limits. Check the action shown in your workspace before deleting or changing a record.
First check that you are signed in with the account used for that activity. Course access requires enrolment; exams require their published availability. If a record is still missing, send support its reference.
Use the account page for the available profile and account settings. For access or correction issues that cannot be resolved there, use the private support form.
Academy combines written preparation, worked examples, practice and assessment pathways. Start with the study guide for your topic and check the exam workspace for current enrolment availability.
The six subjects are evidence-based decisions, criteria weighting, ranking and prioritisation, sensitivity and robustness, AI system evaluation, and a reproducible MCDM project.
The Foundation route is free, including preparation, the required assessment workspace, review and an earned certificate with QR verification. It does not require a personal research-lab subscription.
Normal enrolment requires recorded approval of the exact syllabus, question bank and rubric. Check the assessment workspace. An owner pilot is for testing and cannot earn a certificate.
The published Foundation format allows 45 minutes for seven questions and three short practical responses. Read the exact rubric before starting and save as you go; the timer continues if you leave.
Foundation requires 70/100 overall. Each of the three essential practical criteria must receive at least 5/10 and be marked as met. An assessor reviews practical work and authorises eligible awards.
Can I use AI or get someone to write exam answers?
You may use the published guide, your notes and a basic calculator under the Foundation rules. Another person or generative AI must not produce or rewrite assessment answers. Read the instructions for your assessment.
The Foundation policy permits up to three starts per topic in a rolling 30-day period, with one active or submitted attempt at a time. Check the workspace for your own attempt status.
After passing and individual award authorisation, open My certificates in the assessment workspace. The printable version can be saved as a PDF; standard verification and replacement download are included.
Are Academy certificates accredited qualifications?
DeciMethod certificates recognise defined non-accredited learning achievements. They are not AQF qualifications, nationally recognised training awards, degrees or professional licences.
It opens the permitted achievement record and its current status. A QR record is not a general professional licence or a guarantee about unrelated work. Private achievement information remains protected.
Issued credentials are private by default. You can choose an eligible public directory listing and remove it later. Assessment answers and feedback remain private.
Contact support before starting if you need an accessibility arrangement. Support also handles assessment review and appeal requests; include the relevant attempt reference without sensitive unnecessary data.
Check Academy for the current status. The published page distinguishes the free Foundation route from proposed Applied and Advanced Project levels and their draft pricing. Do not treat proposed prices as an available purchase.
Choose consultation for a focused discussion, expert judgements for an agreed elicitation task, research support for mentoring or specialist work, and lab review for human examination of one fixed analysis.
Discuss the research question, suitable methods or interpretation of results with a defined scope. The consultation page explains session arrangements and the action plan you can expect.
They are actual human contributions to an agreed questionnaire, weighting or decision evaluation protocol. The quote specifies the experts, items, rounds, rationale and reporting scope.
No. One expert supplies one person’s judgements. A panel requires real additional participants, with the number and contribution agreed in advance. Identification or anonymisation follows consent.
A reviewer examines one fixed analysis and its supporting evidence within an agreed scope. The fee covers review work; approval or certification depends on the evidence and the review decision.
Send a brief describing the question and outcome you need. Review the expertise, deliverables, timing and total fee before accepting. Sending a request does not itself take payment or reserve an appointment.
Return to the relevant service page using the same account. Its private request area shows available status, quote and delivery updates. Include the reference when asking for help.
Submit a private application with your expertise, evidence and proposed services. Approval, partnership terms and public profile publication are separate steps; applying does not automatically authorise paid work or reviewing.
Yes. Describe the question, methods, scientific sources, inputs, outputs, examples and limitations in an expert application. Development, ownership, costs, maintenance and any revenue arrangement require a separate agreement.
Applications and supporting private information are reviewed privately. An approved public profile requires the separate publication process and consent. CVs, private contacts, notes and lab proposals are not automatically published.
Support covers one agreed topic and university requirements: problem and gap, objectives, questions, methods, feasibility, ethics, timeline and reference guidance, with collaborative drafting support and one consolidated revision. Confirm scope and timing before acceptance.
The package supports a researcher-led proposal for one topic, using the agreed university template or length. It includes research-design and writing guidance, methodological feedback and one consolidated revision within the agreed scope.
Provide the university, programme, current guidelines, template, word limit and supervisor instructions. Support is tailored to the supplied requirements after review; requirements outside Australia can also be submitted for consideration.
You retain intellectual direction, decisions and final authorship. Support must follow your institution’s permitted-assistance rules. The service does not complete assessed work that you are required to undertake personally.
You can enquire about independent research mentoring and supervision support. The scope, availability and relationship with any university-appointed supervisor must be agreed; an enquiry does not establish a formal university appointment.
Can I get help with a figure, presentation or review protocol?
Specialist research services include focused feedback and support for scientific communication and research development. Describe the material, intended use and required contribution so the deliverables can be scoped.
Does buying support guarantee admission, funding or publication?
No. Proposal support develops your own research preparation. Universities and funders decide applications, and journals make independent editorial decisions. Publication is not included automatically in a proposal package.
Classroom connects course lab activities, documented student attempts, fixed submissions, rubrics and lecturer feedback. Course work remains separate from personal research projects.
Sign in first, then ask your lecturer to enrol the email linked to that account. Only courses associated with the signed-in account appear in Classroom.
Submit the unit description, learning outcomes, assessment instructions and rubric. An authorised university representative reviews the exact proposed design; development and technical acceptance are agreed separately.
The published pilot is 30 days for one approved course, up to 50 students and five assignments, with no card or automatic paid renewal. Institutional authorisation and course readiness are required.
The site provides preparation routes for institutional LMS integration. Check the current connection status and arrange institutional approval and testing; a preparation guide does not itself enable live launch or grade return.
They document what an account submits, not who independently produced it. Use your institutional integrity policy and, where useful, a practical or oral explanation alongside the record.
Explore official degree and funding routes, compare preferences and build a preparation shortlist. Applications are made on the provider’s official website.
Does a high match score mean I will win a scholarship?
No. The score expresses declared preferences, not an admission or award probability. Eligibility checks and missing evidence are separate; verify requirements and round dates with the provider.
The study tool is session-only unless you choose device saving. Device saving is separate from your account, and others using the same browser may see it. Export your plan if you want a copy.
The catalogue is a curated starting point, with official search portals for wider discovery. Recheck current deadlines, eligibility and funding details on the linked provider website.
How is Digital Fusion Press connected to DeciMethod?
DeciMethod provides research tools and support that can help authors prepare methods and evidence. Digital Fusion Press provides a separate scholarly publishing pathway with its own journal scopes and editorial policies.
Compare the six journals’ aims, scope and author instructions on the publisher website. Match the actual study and article type to the journal rather than choosing on title alone.
Does using a lab submit my paper or secure acceptance?
No. Lab work and service requests do not submit a manuscript. Authors use the selected journal’s submission process, and acceptance depends on independent editorial assessment and peer review.
Where can I check my manuscript before submission?
Use the publisher’s Manuscript Pre-check and your selected journal’s instructions. A preparation check identifies issues within its scope; it does not replace editorial review or guarantee acceptance.
Where are journal templates and publishing policies?
Open your chosen journal’s author and resource pages for the applicable template. The publisher policies explain shared requirements; check journal-specific instructions as well.
Are publication fees part of a DeciMethod subscription?
Lab subscriptions and research service fees are separate from journal publishing arrangements. Check the publisher’s current fee policy and the chosen journal’s instructions before submission.
The pricing page lists research plans, student concessions and service guide prices in Australian dollars. Check the current checkout availability and the exact amount shown before any purchase.
No. The seven-day project trial does not require a card and does not automatically charge when it ends. A paid purchase is a separate action with its own terms.
A student concession is listed on the pricing page and requires eligibility review before purchase. A university email alone does not automatically grant the concession.
No. A request begins scope and availability discussions. The accepted quote and available payment stage determine the later checkout; a form submission or accepted quote alone does not charge you.
Use the checkout offered for your confirmed request and agreed payment stage. Check the amount, service and reference before payment. If checkout is unavailable or unclear, ask support rather than paying through an unverified link.
How do I cancel a subscription or request a refund?
Use the billing account for available renewal controls. For service cancellation, rescheduling or refunds, send a private support request and follow the applicable terms and agreed scope.
No. Research-lab access, Academy pathways, expert services and journal publishing arrangements have distinct scopes. Read the selected offer rather than assuming one payment includes the others.
Where do I ask a question that is not listed here?
Use the private support form with a short description and the relevant project, course, assessment or service reference. This guide provides published information and cannot inspect your private account records.
Email support using the address on the Help page. Include DeciMethod, the page and any error message. Leave out passwords, card details and research files. You do not need to sign in to email support.
Can I put passwords or participant data in a support request?
Do not include passwords, card details or identifiable participant information. Describe the problem using the minimum information needed, and agree an appropriate route before sharing confidential research material.
The service policies describe data handling, permitted research support, booking and cancellation arrangements. Use support for a privacy, access, correction or deletion request.
Check the selected lab’s file format, size limit, required columns and numeric assumptions. Try its supplied example to separate a file issue from a workflow issue, then contact support with the error and lab name.
Does Ask DeciMethod use an AI model or send my search elsewhere?
This help panel searches a curated set of platform questions and answers in your browser. It does not call an AI model, submit the search as a support request or access private research records.
Public guidance and this help panel support small screens and keyboard controls. For data-heavy lab work, check the lab’s requirements and use a suitable device. Escape closes this panel, and reduced-motion preferences are respected.