Comparing DEA Analysis Professional (ex-KonSi) to Other DEA Tools

How to Use DEA Analysis Professional (formerly KonSi DEA) for Efficiency MeasurementData Envelopment Analysis (DEA) is a non-parametric method used to evaluate the relative efficiency of decision-making units (DMUs) — such as firms, hospitals, schools, branches, or production lines — that consume multiple inputs to produce multiple outputs. DEA Analysis Professional (formerly KonSi Data Envelopment Analysis DEA) is a dedicated software tool that implements DEA models, offers data management, allows model selection and customization, and produces detailed reports and visualizations for efficiency analysis.

This article explains — step by step — how to use DEA Analysis Professional to measure efficiency, choose appropriate DEA models, prepare and import data, run analyses, interpret outputs, and apply results to practical decision-making. It also covers common pitfalls, advanced features, and best practices.

Overview: What DEA Analysis Professional Does

DEA Analysis Professional provides:

Model selection: CCR, BCC, input- or output-oriented models, super-efficiency, and others.
Data management and validation tools.
Efficient frontier computation and projection of inefficient DMUs.
Slack and sensitivity analysis.
Statistical outputs and graphical visualizations (efficiency scores, peer groups, target recommendations).
Exportable reports for presentations and decision support.

When to use DEA

Use DEA when:

You need to compare units that use multiple heterogeneous inputs and produce multiple outputs.
There is no reliable price information to build a parametric production function.
You want relative efficiency measures (frontier-based) rather than average or regression-based metrics.
Sample size is reasonable: a common rule of thumb is at least 3×(number of inputs + outputs) DMUs for stable results.

Key limitation: DEA is deterministic and sensitive to outliers and measurement error; complement DEA with sensitivity analysis or bootstrap methods where possible.

Preparing your analysis

1) Define the decision-making units (DMUs)

Identify comparable units (same objectives, similar operating environment).
Ensure units are homogeneous in function; do not mix dissimilar operations.

2) Select inputs and outputs

Inputs: resources consumed (labor hours, cost, machines, beds, etc.).
Outputs: desirable results (sales, patients treated, graduates, units produced).
Avoid mixing inputs and outputs incorrectly; each variable should be clearly input or output.
Keep the number of variables moderate relative to DMU count to avoid rank deficiency and too many units scoring 1 (efficient).

3) Data collection and cleaning

Use the same measurement units across DMUs.
Check for missing values, outliers, or zero values where infeasible.
Normalize or scale data if appropriate (DEA is scale-invariant for ratio data in many models but inconsistent scaling can affect numerical stability).

Getting started in DEA Analysis Professional

Installation and setup

Install the software following the vendor’s instructions.
Create a new project and set a descriptive project name and metadata (date, analyst, domain).
Configure workspace options such as default model types, precision, and output folders.

Importing data

DEA Analysis Professional typically supports CSV, Excel, and direct database connections.
Prepare a tabular file: first column DMU identifiers; subsequent columns inputs and outputs. Include a header row with variable names.
Use the import wizard to map columns to inputs/outputs and to set variable types (input/output, ordinal/continuous).

Example data layout:

DMU, LaborHours, CapitalCost, Outputs_Sales, UnitsProduced BranchA, 1200, 25000, 500000, 1200 BranchB, 900, 18000, 420000, 950 ...

Data validation inside the tool

Run the built-in validation to locate missing values, negative or zero inputs/outputs (if inappropriate), and dominate variables.
Use descriptive statistics and histograms provided by the tool to visualize distributions and spot outliers.

Choosing a DEA model

DEA Analysis Professional offers several common models:

CCR (Charnes–Cooper–Rhodes): Assumes constant returns to scale (CRS). Use when DMUs operate at an optimal scale.
BCC (Banker–Charnes–Cooper): Assumes variable returns to scale (VRS). Use when scale inefficiencies are likely.
Input-oriented vs Output-oriented: Choose input-oriented when the goal is to minimize inputs for a given output; choose output-oriented when maximizing outputs with given inputs is the goal.
Additive, SBM (slack-based measure), and non-radial models: Use for direct treatment of slacks and when radial measures obscure inefficiencies.
Super-efficiency models: Rank efficient DMUs beyond the standard efficiency score of 1.

Choose model based on:

Economic/operational question (minimize inputs vs maximize outputs).
Scale assumptions (CRS vs VRS).
Need to rank efficient DMUs.

Running the analysis

Select DMUs and variables.
Choose model type and orientation.
Set solution options:
- Precision and numerical tolerance.
- Return-to-scale restrictions (CRS/VRS).
- Constraints (weight restrictions, assurance regions) if you want to reflect prior knowledge or limit unrealistic weightings.
Run the calculation.

DEA Analysis Professional computes:

Efficiency scores (θ for input-orientation; φ for output-orientation).
Reference sets/peer groups for each DMU.
Target input/output levels and slack values.
Dual variables (weights) and lambda values (intensity vectors).

Interpreting results

Core outputs

Efficiency score: Values ≤ 1 for input-oriented models (1 = efficient; = inefficient). For output-oriented scores, values ≥ 1 indicate efficiency (1 = efficient).
Peers/reference set: Efficient DMUs that form the convex combination that projects the inefficient DMU onto the frontier.
Targets: Suggested proportional reductions in inputs (input-oriented) or expansions in outputs (output-oriented) to reach the frontier.
Slacks: Non-proportional adjustments needed after radial projection.

Example interpretation

Branch X has input-oriented efficiency 0.78: it should reduce inputs by 22% (in radial terms) to reach the frontier; additional slacks indicate further specific input cuts.
Branch Y is efficient (score 1) and appears in several peers lists, indicating best-practice status.

Advanced diagnostics & robustness

Sensitivity and influence analysis

Jackknife or leave-one-out analysis: Check how removal of single DMUs affects others’ efficiency.
Bootstrapping (if supported): Obtain confidence intervals for efficiency scores to assess statistical significance.

Weight restrictions and assurance regions

If DEA assigns unrealistic zero weights to important outputs, impose restrictions to reflect managerial priorities or economic logic.

Super-efficiency and ranking

Use super-efficiency models to rank efficient DMUs and to perform outlier detection; be cautious as super-efficiency can be unstable if data contain outliers.

Visualizing and exporting results

Use graphics (efficiency histograms, frontier plots, target arrows) to communicate findings.
Export tables and charts to Excel, CSV, or PDF for reporting.
DEA Analysis Professional often provides peer network diagrams and efficiency decomposition charts — use these for stakeholder presentations.

Common pitfalls and how to avoid them

Too many variables relative to DMUs: reduces discriminatory power — follow the rule-of-thumb minimum DMU count.
Mixing heterogeneous units: compare only comparable DMUs.
Ignoring outliers: extreme DMUs can distort the frontier. Identify and decide whether to exclude, Winsorize, or analyze separately.
Overreliance on DEA alone: combine results with qualitative assessment and other quantitative methods (stochastic frontier analysis, regression) when possible.

Practical example (brief)

Problem: Evaluate 30 bank branches using 3 inputs (staff hours, operating cost, branch area) and 2 outputs (new accounts, loan volume).
Model: BCC input-oriented because branches vary in scale and the goal is to reduce resource use.
Run DEA Analysis Professional: import data, validate, choose BCC input-oriented, run analysis.
Results: 8 branches efficient (score 1); inefficient branches have radial reductions between 10–45%. Peers identified for each inefficient branch with target input mixes and slack adjustments.
Action: Managers use targets to set resource reduction plans and investigate operational differences with peer branches.

Best practices checklist

Ensure DMU homogeneity.
Keep input/output count reasonable given sample size.
Clean and validate data thoroughly.
Choose model orientation that matches managerial objectives.
Use weight restrictions when economic logic requires.
Run sensitivity/robustness checks (leave-one-out, bootstrap).
Visualize results and translate targets into actionable steps.

Final notes

DEA Analysis Professional (formerly KonSi DEA) is a powerful tool when used with care: define the problem clearly, prepare data appropriately, select the right model, and interpret results in context. Combine DEA outputs with managerial insights and robustness checks to drive meaningful efficiency improvements.