Statistical Analysis Platform
ODAH AI
Auditable statistical analysis, from raw data to publication
Data
- Upload
- Quality Check
- Analysis Scope
Research
- Research Aim
Measurement
- Scale Detection
- Scale Scoring
Analysis
- Analysis Plan
- Results
Output
- Report
Results
10Dataset Read
customer_survey.csvQuality Control
17 records excludedExcluded records were submitted for your approval
Analysis Scope
Corporate and individual customer groups
Research Aim
3 hypothesesThe effect of service quality and response time on satisfaction
Scale Detection
Subdimensions separated, reverse-coded items flipped
Reliability · Cronbach's α
3 scalesAssumption Checks
Shapiro–Wilk p = .003Normality not met — switched to a non-parametric alternative.
Group Comparison · Mann-Whitney U
p = .004Correlation · Spearman's ρ
| Satisfaction | Loyalty | |
|---|---|---|
| Service Quality | .62** | .54** |
| Response Time | −.41** | −.33* |
** p < .01 · * p < .05
Multiple Linear Regression · Satisfaction
R² = .48| Variable | β | SE | p |
|---|---|---|---|
| setup_time_days | −0.42 | 0.06 | <.001 |
| first_response_hr | −0.28 | 0.07 | <.001 |
| training_score | 0.19 | 0.08 | .021 |
Effect sizes and 95% confidence intervals are in the report
ODAH reads your dataset's measurement structure, runs it through quality control, selects the appropriate test and returns the result as a Word report with the methods section already written. The real difference is architectural: decision, computation and audit run as separate layers, and no statistical value ever passes through a language model. Every result is computed deterministically, assumptions are independently verified, and your raw data cells never reach the decision layer.
Key features
How it works
Nine steps from data to report.
Four phases, each ending with your approval. No guesswork in between.
- Rows, columns and measurement levels are identified
- Quality control: consent, attention checks, straight-lining, duplicates
- The records entering the analysis are submitted for your approval
- The measurement structure is resolved and confirmed by you
- Subdimensions are separated, reverse-coded items are flipped
- Reliability (Cronbach's α) is computed
- Assumptions are verified
- If normality fails, an alternative test is used
- Effect size and confidence intervals are computed
- Findings and tables are placed
- The methods section is written
- A publication-ready Word document is produced
Every result is computed deterministically.
Decision Layer
Resolves the data structure, derives measurement decisions and submits them for approval. It never sees your raw data cells.
Computation Layer
Runs the tests deterministically. The same data always yields the same result.
Audit Layer
Tests assumptions and independently audits contradictions. It reports assumption violations even when the result looks significant.
In ODAH, no statistical value ever passes through a language model.
Who it's for
- Graduate students
- Academics and researchers
- Data analysts
- Finance and accounting professionals
- Companies and SMEs
- Health and social science researchers
Data security
Your cells never reach the model
The decision layer does not see raw data
Never used for machine learning
Your data remains yours
Built in Türkiye
Turkish scale knowledge and reporting
Publication-ready report
Methods, tables and interpretation together
Products
Software we build ourselves.
Let's design your project together.
In a 30-minute discovery call we listen to your processes and systems, and tell you plainly which AI solution will actually work and how long the first prototype will take.