A browser-based Statistical Process Control (SPC) tool built with Streamlit. Upload a CSV file, configure your analysis, and instantly generate control charts with full process capability reporting — no software installation required.
Supports X-bar/R, X-bar/S, and Individuals/Moving Range (I-MR) charts with automatic chart type selection, configurable spec limits, and one-click Excel export.
https://spcanalysistool.streamlit.app
Open in any browser — no Python, no Jupyter, no installation needed.
- Three SPC chart types — X-bar/R, X-bar/S, and I-MR; auto-selected based on subgroup size
- Dual-panel chart — X-bar (or Individuals) chart on top, R/S/MR companion chart below
- Control limit zones — green shading within control limits, yellow shading in warning zones
- Capability indices — Cp, Cpk (short-term) and Pp, Ppk (long-term) calculated automatically
- Anderson-Darling normality test — A² statistic + p-value banner; Cp/Cpk hidden when data is non-normal (only Pp/Ppk shown), since Cp/Cpk assume normal distribution
- Process inferences — automated 5-rule diagnostics (capability, centering, stability, ST-vs-LT consistency, normality) with bullet verdicts and narrative summary
- Process statistics — mean, short- and long-term sigma, min, max, and observation count
- Configurable spec limits — USL and LSL inputs with sensible defaults (mean ± 3σ)
- Excel + PDF export — formatted
.xlsxreport with embedded chart and data table;.pdfreport with chart, stats, and inferences - Graceful error handling — clear messages for malformed files, insufficient data, or invalid inputs
| Chart | Description | Auto-selected when |
|---|---|---|
| X-bar / R | Subgroup means and ranges | n = 2–8 |
| X-bar / S | Subgroup means and standard deviations | n ≥ 9 |
| I-MR | Individual values and moving ranges | Manual selection only |
- Open the app at https://spcanalysistool.streamlit.app
- Upload a file using the sidebar — supports CSV (comma, semicolon, or tab delimited) and Excel (
.xlsx/.xls). For multi-sheet Excel files, a sheet selector appears automatically - Select the column to analyze from the dropdown (numeric columns only)
- Set subgroup size using the slider — chart type auto-selects based on n
- Adjust USL and LSL if the defaults (mean ± 3σ) do not match your specification
- The chart and statistics update instantly — Download Excel Report to save results
| Metric | Description |
|---|---|
| Cp | Process capability — spread relative to tolerance (short-term sigma) |
| Cpk | Process capability accounting for mean centering (short-term sigma) |
| Pp | Process performance — spread relative to tolerance (long-term sigma) |
| Ppk | Process performance accounting for mean centering (long-term sigma) |
Note: Cp/Cpk are hidden when the Anderson-Darling test reports non-normal data (p < 0.05), since these indices assume a normal distribution. Pp/Ppk remain shown.
| Field | Description |
|---|---|
| A² | Anderson-Darling statistic |
| p-value | Stephens (1986) approximation; verdict at α = 0.05 |
| Verdict | NORMAL (p ≥ 0.05) → all capability indices shown; NON-NORMAL (p < 0.05) → only Pp/Ppk shown |
Five-rule diagnostic block printed below the metrics:
| Rule | Checks |
|---|---|
| 1. Capability | Cpk vs 1.33 / 1.0 thresholds |
| 2. Centering | Cp vs Cpk gap (off-center detection) |
| 3. Stability | Out-of-control points on the X-bar chart |
| 4. Drift | Cpk vs Ppk consistency |
| 5. Normality | Anderson-Darling verdict |
| Metric | Description |
|---|---|
| Mean (X̄) | Grand mean of all subgroup averages |
| Sigma ST | Short-term sigma estimated from within-subgroup variation |
| Sigma LT | Long-term sigma — overall standard deviation of all individual values |
| Min / Max | Minimum and maximum values in the dataset |
| Observations | Total number of data points used in the analysis |
- Top panel: X-bar (or Individuals) chart with UCL, LCL, USL, LSL, and a stats annotation box
- Bottom panel: R, S, or MR companion chart with its own UCL and LCL
The downloaded .xlsx report contains two sheets:
| Sheet | Contents |
|---|---|
| SPC Report | Stats summary table (chart type, control limits, spec limits, capability indices) + embedded chart image |
| Data | Per-subgroup X-bar and R/S/MR values |
The downloaded .pdf report contains a single A4 page with:
- Report title, column name, chart type, and date header
- Embedded control chart (X-bar + companion panel)
- Three-column stats table — Capability, Process, Control Limits
- Process Inferences section (5-rule bullets + narrative summary)
pip install -r requirements.txt| Package | Version | Purpose |
|---|---|---|
streamlit |
≥ 1.33.0 | Web app framework |
pandas |
≥ 2.0.0 | CSV loading and data manipulation |
numpy |
≥ 1.26.0 | Subgroup calculations and control limits |
scipy |
≥ 1.11.0 | Anderson-Darling normality test |
matplotlib |
≥ 3.8.0 | Chart rendering |
openpyxl |
≥ 3.1.0 | Excel file export |
fpdf2 |
≥ 2.7.0 | PDF report export |
pytest |
≥ 8.0.0 | Unit tests |
streamlit run app.pyIf
streamlitis not on your PATH, usepython -m streamlit run app.py.
pytest tests/ -v- Fork or push this repo to your GitHub account
- Go to share.streamlit.io
- Connect your GitHub account and select this repository
- Set Main file path to
app.py - Click Deploy — the app is live instantly at a public URL
spc-control-chart-tool/
├── app.py # Streamlit app — UI, chart rendering, Excel/PDF export
├── spc_utils.py # SPC calculation logic — chart stats, capability, Anderson-Darling
├── spc_inferences.py # 5-rule diagnostic engine (capability, centering, stability, drift, normality)
├── report_utils.py # Excel + PDF report builders
├── requirements.txt # Python dependencies
├── spc_tool.ipynb # Original Jupyter notebook version
├── sample_spc_report.xlsx # Example Excel export output
├── .streamlit/
│ └── config.toml # Streamlit theme configuration
├── tests/
│ ├── test_spc_utils.py # Tests for SPC calculations + Anderson-Darling
│ ├── test_spc_inferences.py # Tests for 5-rule inference engine
│ └── test_report_utils.py # Tests for Excel/PDF builders
├── screenshots/
│ └── chart_type.png # Chart type selection reference
└── docs/
└── plans/ # Implementation planning documents
- Short-term sigma (Cp/Cpk) is estimated from within-subgroup variation using R-bar/d₂ (X-bar/R) or S-bar/c₄ (X-bar/S) — this reflects the process's inherent capability when operating in a stable state.
- Long-term sigma (Pp/Ppk) is the overall standard deviation of all individual measurements — this reflects actual process performance including shifts and drifts over time.
- Subgroup truncation — data is truncated to the nearest complete subgroup; incomplete trailing observations are excluded.
- SPC constants (A2, A3, D3, D4, B3, B4, d2, c4) are tabulated for subgroup sizes n = 2 through 25 per AIAG/Shewhart reference tables.
- Anderson-Darling normality — uses
scipy.stats.andersonwith the Stephens (1986) small-sample correction (A²* = A² · (1 + 0.75/n + 2.25/n²)) and piecewise p-value approximation. Minimum sample size: 8. When p < 0.05, Cp/Cpk are suppressed since their normal-distribution assumption no longer holds.
The included demo uses the UCI Wine dataset (178 observations, 13 chemical attributes). Replace with your own process or manufacturing data for real analysis.
Aeberhard, S. & Forina, M. (1992). Wine [Dataset]. UCI Machine Learning Repository. https://doi.org/10.24432/C5PC7J.
MIT — free to use, modify, and distribute.