English · Türkçe | Live app: fatihaydost.github.io/MCDM-Workbench | Methods catalogue: /methods
A client-side workbench for multi-criteria decision analysis (MCDM / MCDA): weight criteria with 11 objective and subjective methods (CRITIC, Entropy, SD, MEREC, LOPCOW, CILOS, Equal, BWM, AHP, ROC, SWARA or your own weights), rank alternatives with 15 ranking methods (TOPSIS, SAW, VIKOR, EDAS, WASPAS, COPRAS, MARCOS, CoCoSo, MABAC, MOORA, ARAS, CODAS, PROMETHEE II, ELECTRE I, ELECTRE III), test robustness with weight sweeps and Monte Carlo simulation, compare all rankings side by side, and inspect every intermediate matrix and formula in English or Turkish.
- 1. Data: Enter a decision matrix in a spreadsheet grid, paste directly from Excel, import
.xlsx/.csv, or load one of the 25 published reference examples. Mark each criterion as benefit (↑) or cost (↓). - 2. Weights (11 methods + manual): Compute objective weights from the matrix (CRITIC, Entropy, SD, MEREC, LOPCOW, CILOS, Equal weights), elicit subjective weights from structured judgements (BWM, AHP, ROC / RS / RR, SWARA), or enter weights by hand.
- 3. Ranking (15 methods): Rank alternatives with distance-based (TOPSIS, VIKOR, EDAS, MABAC, CODAS), utility-based (SAW, COPRAS, ARAS, WASPAS, CoCoSo, MARCOS), ratio-based (MOORA), or outranking (PROMETHEE II, ELECTRE I, ELECTRE III) methods, each with an interactive geometric diagram of your own numbers.
- 4. Results & worked calculation: Inspect the final ranking, criterion weights, inter-criterion correlation heatmap, and every step of the calculation with KaTeX formulas and labelled matrices. Copy any step as TSV or LaTeX, or export the entire multi-sheet calculation as
.xlsxor.csv. - 5. Robustness & sensitivity: Sweep any criterion's weight from
0to1to see exact rank-reversal crossover points, perturb all weights (±5%,±10%,±20%), run 1,000-trial Dirichlet Monte Carlo simulations for rank acceptability, and test leave-one-criterion-out stability in a background Web Worker. - 6. Multi-method comparison: Run all compatible ranking methods on your matrix and weights simultaneously, inspect Spearman ρ and Kendall τ pairwise rank correlations, Kendall's W concordance, Borda consensus ranks, and method-by-method explanations of why two alternatives swap places.
- 100% private & in-browser: No sign-up, no server, no telemetry; your data never leaves your browser.
Compare how all 15 ranking methods order your alternatives, where rank lines cross, and why two methods disagree on a pair:
Stress-test the winning alternative against weight shifts, uniform perturbations, Monte Carlo weight sampling, and criterion removal:
All 26 methods compute in the browser and have a dedicated reference page with an interactive step-by-step illustration, formulas, pitfalls, compatible method pairs, and a decision tree guide at /methods:
| Family | Method | Origin | What it computes |
|---|---|---|---|
| Objective weighting | CRITIC | Diakoulaki et al. (1995) | Weights from min-max contrast ( |
| Objective weighting | Entropy | Shannon (1948) / Zeleny (1982) | Weights from Shannon entropy diversification |
| Objective weighting | SD | Diakoulaki et al. (1995) | Weights proportional to the sample standard deviation of min-max normalized columns |
| Objective weighting | MEREC | Keshavarz-Ghorabaee et al. (2021) | Weights from the logarithmic performance drop when each criterion is removed |
| Objective weighting | LOPCOW | Ecer & Pamucar (2022) | Weights from the log ratio of root-mean-square to sample standard deviation |
| Objective weighting | CILOS | Zavadskas & Podvezko (2016) | Weights solving the homogeneous impact-loss linear system |
| Objective weighting | Equal weights | Laplace | Baseline |
| Subjective weighting | BWM | Rezaei (2015, 2016) | Closed-form Linear Best-Worst Method from |
| Subjective weighting | AHP | Saaty (1977, 1980) | Principal eigenvector & geometric mean priorities from pairwise ratios with Saaty's CR |
| Subjective weighting | ROC | Barron & Barrett (1996) | Cardinal weights from criterion importance ranks (ROC centroid, Rank Sum, Rank Reciprocal) |
| Subjective weighting | SWARA | Keršulienė et al. (2010) | Step-wise weight assessment chaining comparative importance ratios |
| Distance-based ranking | TOPSIS | Hwang & Yoon (1981) | Relative closeness |
| Distance-based ranking | VIKOR | Opricovic (1998) | Compromise index |
| Distance-based ranking | EDAS | Keshavarz Ghorabaee et al. (2015) | Appraisal score from weighted positive ( |
| Distance-based ranking | MABAC | Pamučar & Ćirović (2015) | Signed distance |
| Distance-based ranking | CODAS | Keshavarz Ghorabaee et al. (2016) | Pairwise Euclidean distance |
| Utility-based ranking | SAW | Churchman & Ackoff (1954) | Weighted sum |
| Utility-based ranking | WASPAS | Zavadskas et al. (2012) |
|
| Utility-based ranking | COPRAS | Zavadskas et al. (1994) | Benefit share |
| Utility-based ranking | MARCOS | Stević et al. (2020) | Utility function |
| Utility-based ranking | CoCoSo | Yazdani et al. (2019) | Combined compromise score |
| Utility-based ranking | ARAS | Zavadskas & Turskis (2010) | Optimality ratio |
| Ratio-based ranking | MOORA | Brauers & Zavadskas (2006) | Vector-normalized benefit sum minus cost sum |
| Outranking | PROMETHEE II | Brans & Vincke (1985) | Unicriterion preference functions (Linear, Usual, Level, Gaussian) and net outranking flow |
| Outranking | ELECTRE I | Roy (1968) | Concordance majority |
| Outranking | ELECTRE III | Roy (1978) | Pseudo-criteria ( |
The automated test suite (57 test files, 1,732 unit tests) recomputes published examples from peer-reviewed papers and checks every printed digit:
| Method | Reference | What is compared | Tolerance |
|---|---|---|---|
| CRITIC | Krishnan et al. (2021), Symmetry 13(6):973, Tables 1, 2 and 5 | standard deviations and weights | ±0.00005 |
| Entropy | Zavadskas et al. (2017), Sustainability 9(5):702, Tables 4 and 8 (the paper prints the weights on permuted rows); min-max option: Krishnan et al. (2021), Tables 1, 2 and 5 | weights (min-max: also the normalized matrix) | ±0.0001 (min-max ±0.00005) |
| SD | Diakoulaki et al. (1995), Computers & OR 22(7):763, Table 1, Section 5 and Table 3 (the printed σ is the population SD); Mukhametzyanov (2021), DMAME 4(2):76, Tables 1, 2 and 4, four matrices with cost criteria | weights and standard deviations | ±0.0005 |
| MEREC | Keshavarz-Ghorabaee et al. (2021), Symmetry 13(4):525, Tables 2-4 and Steps 3-6 (illustrative) and Tables 5-6 (comparative); Dua et al. (2024), EUREKA: Physics and Engineering (2):131, Tables 8-9 and 12-13 | weights, normalized matrix, S, S′ and removal effects | ±0.00005 (comparative ±0.0005; S, S′, E ±0.005) |
| LOPCOW | Ecer and Pamucar (2022), Omega 112:102690; Keleş (2023), ÖHÜ İİBF Dergisi 16(3):727, Tables 6-8 (10 logistics firms, 8 criteria) | normalized matrix, RMS, sample SD, PV and weights; RMS >= σ identity and m = 2 collapse | ±0.0005 |
| CILOS | Zavadskas and Podvezko (2016), IJITDM 15(2):267; Zavadskas et al. (2017), Sustainability 9(5):702, Tables 4 and 10 (Social block) and Table 8 (Economic block); Pala (2021), Bingöl Univ. J. Econ. Admin. Sci. 5(2):277 | sum-normalized matrix, square matrix A, relative impact-loss matrix P, system matrix F and weights | ±0.00005 |
| ROC | Barron and Barrett (1996), Management Science 42(11):1515; Stillwell, Seaver and Edwards (1981), OBHP 28(1):62; Roszkowska (2013), Optimum. Studia Ekonomiczne 5(65):14, Table 4 and Tables 5-6 | Rank Order Centroid (ROC), Rank Sum (RS) and Rank Reciprocal (RR) weights and tie-averaged rank weights | ±0.00006 |
| SWARA | Keršulienė, Zavadskas and Turskis (2010), Journal of Business Economics and Management 11(2):243, Table 1 (6 dispute-resolution criteria); Adalı and Işık (2017), IREM 5(4):56, Table 5 | comparative importance ratios s_j, step coefficients k_j = 1 + s_j, recalculated weights q_j and normalized weights w_j | ±0.005 |
| BWM | Rezaei (2016), Omega 64:126, Examples 1-3 (Example 3 from Liang et al. (2020), Table 1); Ratandhara and Kumar, arXiv 2506.17666, Examples 1-4 (one per closed-form branch); Demir and Bircan (2020), Tables 5-7; Arsu and Uğuz Arsu (2021), Tables 4-5 | linear-model weights, ξ, input-based CR and the criterion that fits worst | ±0.00006 |
| AHP | Saaty (2008), IJSSci 1(1):83, Tables 2, 7 and 8 (job choice, drink consumption corrected for a printed reciprocal typo, flexibility); Doğuç (2022), Beykent Univ. J. Sci. Eng. 15(1):19, Table 2 (recruitment); Bozóki, Fülöp and Rónyai (2010), EJOR 206(2):355, Section 6 (incomplete 8×8 matrix, LLSM completion) | principal-eigenvector and geometric-mean weights, λ_max, CI, CR, worst-fitting pair and incomplete-graph LLSM weights | ±0.0005 |
| TOPSIS | Opricovic and Tzeng (2004), EJOR 156(2):445, Table 3, two unit variants | D+, D−, closeness and ranks | ±0.0005 |
| SAW | Podvezko (2011), Engineering Economics 22(2):134, Tables 1-3 (linear max; Table 2 misprints one Lithuania value, Table 3 follows the misprint) and 5-6 (sum); ranks of Keshavarz Ghorabaee et al. (2015) Table 7, Zavadskas et al. (2012) Table 2 and Vujičić et al. (2017) Tables 8a-8b | normalized matrices, scores and ranks | ±0.0005 |
| VIKOR | Opricovic and Tzeng (2004), EJOR 156(2):445, Table 3, two unit variants; ranks of Keshavarz Ghorabaee et al. (2015) Table 7, all seven weight sets (v = 0.5) | S, R, Q, ranks, C1, C2 and the compromise | ±0.0005 |
| EDAS | Keshavarz Ghorabaee et al. (2015), Informatica 26(3):435, Tables 1-3 (47 items; Table 2 is rounded to 3 and then 2 decimals, all 517 cells reproduced) and Table 7 (ranks, all seven weight sets); Aydın Ünal (2019) and Sarıhan and Aydın (2023), each with a printing error recorded | average, PDA, NDA, SP, SN, NSP, NSN, AS, classes and ranks | ±0.006 |
| WASPAS | Aytaç Adalı and Tuş Işık (2017), IREM 5(4):56, Tables 4-7 (supplier selection, all 55 scores of Table 7 over λ); Chakraborty and Zavadskas (2014), Informatica 25(1):1, Tables 13-15 (robots); Zavadskas et al. (2012) Table 2 (Table 1 misprints one a1 value, recorded) | normalized matrix, WSM, WPM, Q at every λ and ranks | ±0.0001 |
| COPRAS | Özcan and Çelik (2021), IJPME 9(2):81, Tables 1, 2, 10, 11 (cream separators, Entropy then COPRAS; the printed rank row swaps A2 and A3, the printed Q does not); Podvezko (2011) Tables 7-8 (two cost changes flip the ranks); Zavadskas et al. (2007) Tables 1-2 (S+ and S- only: the printed Q is S+ + S-, not the paper's own Eq. 5) | weighted shares, S+, S-, Q, N and ranks | ±0.0001 |
| MARCOS | Stević and Brković (2020), Logistics 4(1):4, Tables 2-5 (23 drivers, FUCOM weights as Table 4 applies them; the paper's Figure 2 swaps C3 and C4, which would move 9 ranks); Madenoğlu (2020), EBOR 2(2):99, Tables 3-7 (6 suppliers, 16 criteria, AHP weights) | S, K-, f(K), the reference totals and ranks; f(K) = c x SAW (linear max) on 300 random problems | ±0.001 |
| CoCoSo | Yazdani et al. (2019), Management Decision 57(9):2501, Tables I, III-VI (7 logistics providers, expert weights; the paper prints k rounded, 1.3 for 1.2987) and Tables X-XI (10 green suppliers); Akgül (2021), Ekonomi ve Finansal Araştırmalar Dergisi 3(2):71, Tables 4-12 (9 Turkish banks, CRITIC weights) | S, P, ka, kb, kc, k, ranks and the λ rank changes; shift and scale invariance, zero-weight and constant columns on 200 random problems | ±0.0015 |
| MABAC | Pamučar and Ćirović (2015), Expert Systems with Applications 42(6):3016, Tables 5, 7-12 (7 forklifts, 10 criteria, DEMATEL weights; the two Likert criteria coded 1-5); Biswas and Das (2018), Management Science Letters 8:121, Tables 2-7 (9 hybrid vehicles, Entropy weights, all 9 printed digits); Çelik (2020), İslam Ekonomisi ve Finansı Dergisi 6(2):311, Tables 3-12 (6 Turkish participation banks, CRITIC weights computed here) | G, Q, S and ranks; S = min-max weighted sum - B on 300 random problems; G finite at 500 alternatives | ±0.0001 |
| MOORA | Brauers and Zavadskas (2010), Technological and Economic Development of Economy 16(1):5, Appendix C Table 2 (6 projects, 9 criteria); Brauers and Zavadskas (2006), Control and Cybernetics 35(2):445, Tables 1 and 5; Vujičić et al. (2017), Tehnički vjesnik 24(2):619, Tables 1, 7 and 8a-8b (aircraft selection with Entropy and CRITIC weights) | vector-normalized matrix, weighted matrix, S+, S-, y and ranks | ±0.0002 |
| ARAS | Zavadskas and Turskis (2010), Technological and Economic Development of Economy 16(2):159, Tables 1-4 (14 rooms, 6 criteria; Table 2 Room 8 illumination typo 390 vs 400 lx recorded); Bakır and Atalık (2018), Business and Economics Research Journal 9(1):179, Tables 3-6 (11 airlines, 8 criteria, Entropy weights) | optimal row, normalized and weighted matrices, S0, Si, utility degree Ki and ranks | ±0.0001 |
| CODAS | Keshavarz Ghorabaee et al. (2016), Economic Computation and Economic Cybernetics Studies and Research 50(3):25, Examples 1-2 (7 robots, 5 criteria and 14 rooms, 6 criteria; Section 4 sensitivity script < τ inversion recorded); Badi et al. (2018), DMAME 1(1):1, Tables 2-5 (6 steelmaking suppliers, AHP weights); Trung (2022), EUREKA: Physics and Engineering (5):112, Tables 12 and 14-16 | normalized and weighted matrices, negative-ideal point, Euclidean Ei, Taxicab Ti, relative assessment matrix and Hi | ±0.0005 |
| PROMETHEE II | Brans, Vincke and Mareschal (1986), EJOR 24(2):228, Tables 2-4 (6 hydroelectric power plants, 6 criteria, all 6 preference function types); Goswami (2020), Foundations of Management 12(1):93, Tables 4, 7, 13-18 (6 laptops, 7 criteria); Abdullah, Chan and Afshari (2019), IEEE Access 7:43472, Tables 5 and 9-12 | aggregated preference matrix Π, leaving flow φ+, entering flow φ−, net flow φ and PROMETHEE I incomparabilities | ±0.00001 |
| ELECTRE I | Roy (1968), RIRO 2(V1):57; Wang and Triantaphyllou (2008), Omega 36(1):45, Section 3.1 (Galway wastewater treatment plant, 5 sites, 7 criteria); Soner and Önüt (2006), J. Fac. Eng. Arch. Gazi Univ. 21(3):579 | concordance C, discordance D, outranking graph S, SCC-condensed kernel K = {A2, A3, A5} and net concordance / discordance | ±0.0005 |
| ELECTRE III | Roy (1978), Cahiers du CERO 20(1):3; Wang and Triantaphyllou (2008), Omega 36(1):45, Section 3.2, Tables 2-3 (11 waste incineration strategies, 11 criteria, 110 off-diagonal credibility cells) | pseudo-criterion concordance C, credibility σ(a, b), descending and ascending distillation pre-orders and incomparabilities | ±0.005 |
Alongside the published literature tables in tests/fixtures/, the suite verifies permutation, scale, and shift invariants, exact algebraic equivalences across methods, and boundary conditions across degenerate or extreme inputs.
- UI & state: React 19, TypeScript, Vite, Zustand, Wouter, Tailwind CSS 4, Radix UI
- Math, tables & i18n: Pure TypeScript core (
src/core/), KaTeX, SheetJS (.xlsximport/export), i18next (en/tr) - Verification: Vitest (1,867 unit, property, invariant, and literature fixture tests across 65 test files) and Playwright (end-to-end & axe-core accessibility tests)
- Deployment: Static build on GitHub Pages
Requires Node 24+ and pnpm.
pnpm i
pnpm dev # http://localhost:5173
pnpm test # Vitest unit & literature verification suite
pnpm e2e # Playwright end-to-end & accessibility tests
pnpm build # TypeScript check + production bundle




