COSMIN compliant · 17 AI actions · EFA + CFA built in

Build Psychometric Scales
From Construct to Pilot

18 development phases. 17 AI actions. Deterministic EFA/CFA fit verifier. Built on COSMIN, DeVellis 8-step, and AERA/APA Standards.

Build your scaleSee how it works
No credit card to start7-day money back15,000 Mindful AI Tokens/month
scale.rsminds.com / workflow
01Idea & Construct Definition
04Domain & Subdomain Mapping
08Item Generation (overinclusive)
13Expert Panel · CVI / CVR
18Pilot-Ready Synopsis
+ 13 more sections — see full workflow below
0

AI actions — largest surface across RSMinds

0

workflow phases — construct to pilot-ready

0

reporting families integrated · COSMIN + AERA/APA

₹0/mo

starts here · 7-day money back

How ScaleMinds works

The 18-phase workflow

Six phases — Conceptualization, Theory & Blueprint, Item Development, Scoring & Validation, Pilot Package, Output. Each step feeds the next: your construct shapes domains; domains shape items; items shape the CFA model.

01

A · Conceptualization

Idea & Construct Definition

Inputs

Construct idea text

AI action

predictScaleType

Output

Best-fit scale type (Likert, VAS, semantic differential…) + construct boundary draft.

Saves 30–60 min of scale-format deliberation
02

A · Conceptualization

Literature & Gap Analysis

Inputs

Construct

AI action

analyzeConstructLiterature

Output

Existing instruments mapped, adaptation pathway flagged, justification for new scale.

Saves 4–6 hours of database screening
03

A · Conceptualization

Operational Definition

Inputs

ConstructLiterature

AI action

generateOperationalDef

Output

Measurable, behaviour-anchored definition suitable for item generation.

Saves 1–2 hours of refinement
04

A · Conceptualization

Domain & Subdomain Mapping

Inputs

Operational definition

AI action

generateDomainMap

Output

Dimensions, subdomains, and reflective vs formative model decision.

Saves 2–3 hours of conceptual blueprinting
05

A · Conceptualization

Research Question & FINER

Inputs

ConstructDomains

AI action

draftScaleQuestion + auditResearchQuestion

Output

Question drafts + FINER scoring (Feasibility, Interest, Novelty, Ethics, Relevance).

Saves 1–2 hours vs manual drafting
06

B · Theory & Blueprint

Theory & Conceptual Framework

Inputs

QuestionDomains

AI action

discoverTheories + generateFramework

Output

Theory candidates with citations + framework diagram tying construct to domains.

Saves 3–5 hours of literature triangulation
07

B · Theory & Blueprint

Blueprint & Specification Table

Inputs

DomainsFramework

AI action

generateBlueprint

Output

Items-per-domain target, coverage balance, positive/reverse split spec.

Saves 1–2 hours of planning
08

C · Item Development

Item Generation (Overinclusive Pool)

Inputs

Blueprint

AI action

generateItemPool (parallel per-domain SSE)

Output

Overinclusive item pool — deductive + inductive items, keying tagged, reading level checked.

Saves 4–8 hours of drafting
09

C · Item Development

Item Quality Guardian (15-pt Audit)

Inputs

Item pool

AI action

auditItemQuality

Output

Per-item flags: double-barrelled, negation, jargon, social desirability, construct drift.

Saves 2–3 hours of manual review
10

C · Item Development

Item Refinement

Inputs

PoolAudit

AI action

refineItems

Output

Cleaned item set with rewrites, traceable to original phrasing for audit.

Saves 1–2 hours of rewriting
11

C · Item Development

Response Format & Anchors

Inputs

ConstructRefined items

AI action

generateAnchors

Output

Recommended scale points (5/7/VAS) + anchor wording with cross-cultural notes.

Saves 1 hour of anchor crafting
12

D · Scoring & Validation

Scoring Model

Inputs

ItemsAnchors

AI action

generateScoringModel

Output

Total + subscale scoring rules, reverse-key handling, missing-data policy.

Saves 1–2 hours and pre-empts pilot rework
13

D · Scoring & Validation

Expert Panel (CVI / CVR)

Inputs

ItemsConstruct

AI action

planExpertPanel

Output

Expert profile spec + S-CVI / I-CVI / CVR rating form ready to send.

Saves 2–4 hours of CVI form design
14

D · Scoring & Validation

Cognitive Interview Protocol

Inputs

Items

AI action

generateCognitiveProtocol

Output

Think-aloud + verbal-probe protocol surfacing comprehension failures before fielding.

Saves 2–3 hours of protocol drafting
15

E · Pilot Package

Pilot Study Design

Inputs

DomainsItems

AI action

generatePilotPlan

Output

Sample size for EFA/CFA, recruitment plan, data-collection protocol.

Saves 2–3 hours of pilot planning
16

E · Pilot Package

Instrument Assembly

Inputs

All prior sections

AI action

assembleInstrument

Output

Formatted draft instrument — instructions, items, response options, scoring key.

Saves 1–2 hours of formatting
17

E · Pilot Package

Planned Analysis Protocol

Inputs

Pilot plan

AI action

generateAnalysisPlan

Output

Pre-specified item statistics, internal consistency, EFA/CFA decision tree with thresholds.

Saves 2–4 hours of analysis prep
18

F · Output

Creation Synopsis

Inputs

All previous sections

AI action

generateScaleSynopsis (6-group parallel SSE)

Output

COSMIN-aligned development synopsis exportable as DOCX, PDF, or Markdown.

Saves a full day of synopsis drafting

Scale-type coverage

14 scale types — all covered

From classic Likert and visual analogue scales to modern IRT and computer-adaptive testing. Each scale type gets its own item guidance, anchor logic, and analysis-plan defaults.

Rating

4 types

  • Likert (5 / 7 / 9 pt)

    Classic ordered agreement scale.

  • Semantic Differential

    Bipolar adjective anchors.

  • Numerical Rating

    0–10 magnitude judgements.

  • Stapel Scale

    Unipolar single-adjective rating.

Continuous

3 types

  • Visual Analogue (VAS)

    100mm line, fine-grained.

  • Graphic Rating

    Illustrated anchors for clarity.

  • Slider Scale

    Digital VAS with snap option.

Comparative

3 types

  • Thurstone Equal-Appearing

    Pre-scaled judge-rated items.

  • Guttman Cumulative

    Hierarchical, unidimensional.

  • Paired Comparison

    Forced choice between options.

Modern Test Theory

4 types

  • Rasch (1-PL)

    Item difficulty, person ability.

  • IRT (2-PL / 3-PL)

    Discrimination + guessing parameters.

  • Multi-dimensional IRT

    Several latent traits jointly.

  • Computer-Adaptive (CAT)

    Item-bank-driven testing.

Compliance

Built on the standards
your reviewers expect

Integrated

COSMIN Risk of Bias

Methodology quality for measurement studies

Integrated

COSMIN OMP

Outcome Measurement Property checklist

Integrated

COSMIN Methodology

Study-design taxonomy for PROMs

Integrated

DeVellis 8-step

Scale Development: Theory & Applications

Integrated

AERA / APA / NCME

Standards for Educational and Psychological Testing

Integrated

Streiner & Norman

Health Measurement Scales — practical guide

Integrated

ITC Test Translation

Translating and Adapting Tests guidelines

Integrated

ISPOR PRO Good Practice

Patient-Reported Outcome research practices

Integrated

FDA PRO Guidance 2009

Patient-reported outcome support for labelling

Integrated

IRT / Rasch Reporting

Item Response Theory reporting standards

Integrated

Boateng et al. 2018

Best practices for scale development in health

Integrated

Lynn 1986 (CVI)

Content validity index expert-panel norms

Integrated

Messick 1995

Unified validity framework

Integrated

GRADE-PRO

Evidence quality for PRO measures

Integrated

CREDES

Conducting and Reporting of Delphi Studies

Your instrument is scored against every applicable standard in real time.

Why this matters

Factor structure,
but with a second opinion

Other tools

Suggest a factor model. That's it.

If the AI hallucinates — wrong number of factors, mis-loaded items, or a CFA model that won't actually converge — you don't know until your psychometrician opens the lavaan output and the χ² is on fire.

ScaleMinds

AI proposes. Verifier checks the fit.

A deterministic verifier re-computes CFI, TLI, RMSEA, and SRMR from the AI's proposed factor model and checks each against the accepted thresholds. If any index flunks, we flag it before you commit pilot data.

  • Deterministic, not stochastic — same model, same indices.
  • Hu & Bentler 1999 thresholds applied per index.
  • Disagreement flagged with respecification suggestions.
section 17 / planned analysis · 3-factor reflective model
Fit indexAI estimateVerifierThresholdVerdict
CFI0.9510.948≥ 0.95Pass
TLI0.9420.939≥ 0.95Flag
RMSEA0.0580.061≤ 0.08Pass
SRMR0.0460.044≤ 0.08Pass

Respecification suggestion

TLI = 0.942 is below the 0.95 threshold. Consider correlating residuals on items 7–9 (same domain, parallel wording) or revisiting cross-loadings before pilot launch.

Plans

Simple pricing

7-day money-back guarantee

Access 1m

₹299/ 30 days

15,000 Mindful AI Tokens

  • All 14 scale types
  • 18-phase workflow
  • COSMIN-aligned exports
Choose Access 1m
Most popular

Access 2m

₹499/ 60 days
Save ₹99

15,000 Mindful AI Tokens / month

  • Everything in 1m
  • Priority AI throughput
  • Save ₹99 vs monthly
Choose Access 2m

Access 3m

₹699/ 90 days
Save ₹198

15,000 Mindful AI Tokens / month

  • Everything in 2m
  • Quarterly project cadence
  • Save ₹198 vs monthly
Choose Access 3m
Compare all plans

FAQ

Frequently asked questions

Common questions from scale developers, psychometricians, and PROM researchers.

What sample size do I need for EFA / CFA?

ScaleMinds applies established rules of thumb (10 participants per item, minimum 200 for EFA, 300+ for CFA) and adjusts based on factor structure complexity, expected loadings, and communalities. The deterministic verifier cross-checks the recommendation against MacCallum, Widaman, Zhang & Hong (1999) guidance for your specific design.

How many items should I write per domain?

The blueprint phase (step 7) generates 2–3x the final target — typically 8–15 items per domain for a 4-item final subscale. Overinclusion is intentional so the expert panel (CVI) and pilot EFA can trim weak items without leaving the domain under-represented.

Can I import existing items from another scale?

Yes. Paste the source instrument and ScaleMinds extracts items, response anchors, and scoring rules. The item quality audit (step 9) then re-evaluates them against COSMIN criteria so you can decide what to keep, adapt, or replace.

How do I handle translation and cross-cultural adaptation?

ITC Guidelines for Translating and Adapting Tests are integrated. The cognitive interview protocol (step 14) generates language-specific probes, and the analysis plan includes measurement invariance testing (configural / metric / scalar) for multi-language pilots.

What about test-retest reliability planning?

The pilot study design (step 15) includes test-retest sample size recommendations and the analysis protocol (step 17) pre-specifies ICC(2,1) for continuous scores and weighted kappa for ordinal scoring, with intervals tied to construct stability assumptions.

Will my expert panel data work with this?

Yes. Step 13 generates the rating form your experts complete (Lynn 1986 4-point CVI rubric by default). Paste back their ratings and the platform computes I-CVI per item, S-CVI/Ave, and S-CVI/UA so you can defend item retention decisions to reviewers.

Does ScaleMinds run the EFA / CFA itself?

No — analysis runs on real pilot data using R lavaan, psych, or your preferred software. ScaleMinds pre-specifies the model, fit thresholds (CFI ≥ 0.95, RMSEA ≤ 0.08, SRMR ≤ 0.08), and the deterministic verifier checks the proposed model is identifiable before you collect data.

Can I use this for a PROM (patient-reported outcome measure)?

Yes. FDA PRO Guidance 2009 and ISPOR Good Research Practices are integrated. The workflow covers conceptual model, content validity, recall period, and qualitative concept elicitation steps required for regulatory PRO submission.

How is this different from just asking ChatGPT?

COSMIN-anchored prompts, deterministic CFA fit verifier, expert panel form generation, item quality audit against a 15-point rubric, and parallel per-domain item-pool generation. Plus the synopsis is structured for journal publication, not chat replies.

Can I cancel anytime?

Yes — cancel from Settings. Access continues to the end of the paid period. Your instrument exports remain yours. 7-day money-back guarantee on every plan.

Build your scale in a weekend, not a semester.

1 free AI call. No credit card. Sign in with Google.

Build your scale