Your Golden Batch workflow, made simple.
Follow the arrows. Build your reference first, then compare or simulate.
End-to-End Workflow
Choose your starting point, then follow the same steps.
Upload Dataset
Start New → choose CSV or Excel → Upload Dataset.
Set Active
Continue Existing → choose a dataset → Set Active.
Review your data
Check the active file. Edit if needed, then choose Save and Continue.
Choose your quality variable (Y)
Select Target variable (Y) → Visualize Batch-wise.
Compare with your ideal
Upload Ideal Batch → View Similarity Matrix → select suitable dataset batches.
Select golden candidates
Choose batches on the plot, manually, or using a final-Y range.
Preview → refine → lock
Preview Selected → X-Variable Check → add or remove batches → Locked Batches.
Review your golden reference
Check trends, golden means, Y similarity and feature importance. Export golden batch data or the PDF report.
Selections stay connected: removing batches or using Clear Plot Selection in Identify + Define updates Statistics too.
Find deviations
Choose a batch or upload a matching CSV → review charts and deviation levels → download the PDF.
Back to Statistics returns to your reference.Test a process change
Select a trained PCA model → Load Scenario → Add range → Simulate.
Train Model if needed. Back to Compare returns to the comparison.Export the golden curve
Download Golden Envelope → use its mean, minimum and maximum curves in your Ferment IQ project.
CSV curves use 0–100% batch time.2) Advanced Envelope Explorer: What To Select, When, and Why
Use this selection flow across Statistics, Compare and Simulator.
Select Y
Your measured quality outcome. Set it in Identify + Define to assess batch quality.
Review relevant X
Check process measurements in Statistics. Use feature importance to prioritize related variables.
Compare or simulate
Locate deviations in Compare. Test a selected X change over a time range in Simulator.
Golden mean + 5th–95th percentile band. Outside points indicate deviation. Check the level, percentage and worst time together.
Mean + minimum–maximum bounds. The exported envelope uses these bounds on normalized 0–100% batch time.
For outside points, Threshold Ratio controls the level classification. Check Status for insufficient data.
Useful details before you continue
- Use numeric measurements as X, not batch IDs or row indexes. Feature importance shows association, not proof of cause.
- Statistics’ similarity table uses your target Y. Its dedicated Golden Batch Envelope plot requires the Penicillin concentration(P:g/L) column.
- Final-Y range selection requires at least 30 records per batch. A comparison CSV must match the active dataset’s columns and their order.
- Maximum Absolute Ratio measures distance outside the band relative to its width. Changing the threshold does not change the outside-point count.
- Simulator uses the reference saved with its PCA model. Retrain when the model needs to reflect a changed dataset or golden selection.
3) Quick SOP (Recommended)
Use this short checklist for every analysis.
Activate & review — choose the correct dataset and save any edits.
Select Y & batches — visualize quality and choose golden candidates.
Preview & lock — check Y and X curves, refine, then click Locked Batches.
Review & export — validate Statistics and download data, the report or the Ferment IQ envelope.
Compare — check deviations and export the PDF.
Simulate if needed — load a model, edit a range, simulate and review the changed curves.
Changing batches in Statistics? Click Update Golden Batch to apply your edits before continuing.