A top-10 formulation manufacturer operating four sites was releasing finished batches in an average of fourteen calendar days. Manufacturing was not the constraint. The constraint was a paper batch record that travelled physically between production, laboratory and quality assurance, accumulating queries at every desk.
14 → 5
days average batch release
-82%
documentation errors at first review
-64%
QA review hours per batch
4
sites live within nine months
The starting position
Each batch generated roughly 180 pages of executed record. Quality assurance reviewed all of it, line by line, because there was no reliable way to know which entries deserved attention. Around one in three batches was returned to production for a correction, and each return cost between one and three days.
- Paper batch records physically routed across three departments.
- Laboratory results transcribed manually into the release package.
- No visibility of where a batch record sat at any given moment.
- Deviations raised days after the event that triggered them.
Designing for review by exception
The programme did not begin with software configuration. It began with a line-by-line assessment of the batch record to classify every entry as critical, supporting or informational. Only critical entries would require review; everything else would be reviewed only when the system flagged a deviation from its specification.
Redesigned release flow
Execute
Operators record steps on the line with in-context guidance and forced sequencing.
Validate
Every entry is checked against specification at the moment of capture.
Flag
Out-of-specification and out-of-sequence events raise an exception automatically.
Review
QA reviews exceptions and critical steps rather than the full record.
Release
Electronic signature completes release with a full, immutable audit trail.
Implementation approach
- 1Digitised two representative products first, covering the widest range of process steps.
- 2Ran paper and electronic records in parallel for two full campaigns.
- 3Integrated LIMS results directly into the release package, removing transcription.
- 4Trained operators on the line rather than in a classroom, using their own products.
- 5Rolled out site by site with a shared master template and local variants controlled centrally.
Results after twelve months
| Measure | Before | After |
|---|---|---|
| Average release time | 14 days | 5 days |
| Batches returned for correction | 31% | 4% |
| QA review hours per batch | 11.5 | 4.1 |
| Deviations raised within shift | 22% | 91% |
| Right-first-time documentation | 69% | 98% |
We stopped reading batch records and started reading exceptions. That single change gave us back nine days.
What made it work
- Executive sponsorship that protected the parallel-run period from schedule pressure.
- Operator involvement in template design, which removed resistance at go-live.
- A single master template preventing site-level divergence from returning.
- Integration with LIMS and equipment data so the record assembled itself.