Charts
CPV / LOD, %
LOD, %UCL 2.5
24-117 · 2.62 %
What changes
Fewer documentation-driven deviations, a shorter review cycle, drift flagged before it crosses a limit.
Cost and risk
Fewer documentation-driven deviations.
Each section is checked as its department finishes, so the shift that made an error can fix it before it becomes a deviation.
Documentation-driven deviations per batchSite operations
BR-24-125 · checked as each section finishes
- 1 DispensingWarehouseChecked at hand-off
- 2 GranulationManufacturing6.3 recorded before 6.2Corrected on the same shift
- 3 CompressionManufacturingChecked at hand-off
- 4 PackagingPackagingIn progress
QA reviews and signs the whole record
Documentation-driven deviations per batchSite operations
BR-24-125 · checked as each section finishes
- 1 DispensingWarehouseChecked at hand-off
- 2 GranulationManufacturing6.3 recorded before 6.2Corrected on the same shift
- 3 CompressionManufacturingChecked at hand-off
- 4 PackagingPackagingIn progress
QA reviews and signs the whole record
Cost and risk
More batches per reviewer.
The same team keeps pace as volume grows.
Review effort per batchQA operations
BR-24-117 · 412 pages
410 pages, nothing flagged2 pages to a reviewer
Review effort per batchQA operations
BR-24-117 · 412 pages
410 pages, nothing flagged2 pages to a reviewer
Cost and risk
Drift flagged before it crosses a limit.
Nelson and WECO rules flag a trend in process and equipment data as it comes in.
Out-of-trend signalsMSAT
Out-of-trend signalsMSAT
Time
Shorter batch review cycle time.
Reviewers start from the checks that failed and the entries that couldn't be read, with the rest of the record still in view.
Batch review cycle timeQA operations
BR-24-117 · where review starts
- p. 37 · 6.5 Weight, kg18.40Read with low confidence, 0.71
- p. 112 · Entry 9.209:14Timed before 9.1
410 pages, nothing flagged, still in view
JOJ. Okafor verifies and signs
Batch review cycle timeQA operations
BR-24-117 · where review starts
- p. 37 · 6.5 Weight, kg18.40Read with low confidence, 0.71
- p. 112 · Entry 9.209:14Timed before 9.1
410 pages, nothing flagged, still in view
JOJ. Okafor verifies and signs
Time
Less time gathering records for an investigation.
The batch records are already in one place to search, with their process data.
Time gathering before analysisMSAT
lot 22-0433
- ERPLot 22-0433 · MCC · supplier 2
- Batch records24-117 · 24-118 · 24-121
- Process dataInlet air temperature, all three batches
Time gathering before analysisMSAT
lot 22-0433
- ERPLot 22-0433 · MCC · supplier 2
- Batch records24-117 · 24-118 · 24-121
- Process dataInlet air temperature, all three batches
Time
Less time to retrieve a record for an inspector.
Executed records, paper ones included, can be searched once they're digitized.
Record retrieval time in an inspectionCompliance
BR-22-031 line clearance
BR-22-031 · page 8 of 396Executed 2022Paper
7.1 Line clearance verified✓ RK 14.03.22
Record retrieval time in an inspectionCompliance
BR-22-031 line clearance
BR-22-031 · page 8 of 396Executed 2022Paper
7.1 Line clearance verified✓ RK 14.03.22
What it's used for
From review to inspection, one platform.
Every use runs on the same batch data, so an investigation later traces the same record QA reviewed.
Batch review. Batch Review Copilot reads and checks every page of an executed batch record, and QA starts from what it flagged.
Impact assessment. A balance is found out of tolerance. With your equipment records in the genealogy, you have every batch weighed on it in one list.
Inspections. Asked where a lot went, or who changed a control limit and when, you answer from the genealogy and the limit's change history.
CPV. Each parameter you trend for continued process verification is charted against its limits by the same calculation every cycle.
-
Deviations
When a deviation puts other batches in question, the genealogy shows which ones share its lot. The deviation itself stays in your QMS.
-
Investigations
Work back from the batch to the lots and process data behind it, already linked when the investigation opens.
-
Traceability
A supplier flags a lot. Every batch it went into comes back from one query.
-
Annual product review
The annual product review's batch list, deviations and charts come from the systems you connect.
-
Moving off paper
When a product moves off paper, its past executed records come with it, read into the same data as the new batches, each value linked to its page.
The ways in
Whatever you run today, and whatever you run next.
Start with one way in, two or all three, and add another when you're ready. Each feeds the same record, so a product's paper batches and its electronic batches share one history.
Paper
The site stays on paper. The data does not.
Every page is read and every field is scored for confidence, and anything read with low confidence goes to a person who verifies it and signs.
Paper Record Digitization, with Batch Review Copilot
Beta
Execution
There is no form to design.
EBR builds the electronic record from the paper master batch record you've already approved. AI drafts the conversion, your people approve it, and operators work from screens that look like the pages they know.
EBRSystems
Connect what you run. Keep what you run.
Synapse links the batches, lots and parameters across the systems you connect, by algorithm rather than machine learning, so the same data always gives the same links.
Integration, with SynapseThe platform
A module for each step in a batch's life.
From executing the batch record to the annual product review.
01Run
EBR Beta
An electronic batch record for processes that keep changing.
02Digitize
Paper Record Digitization
Batch records, certificates of analysis and equipment logbooks, read into data the rest of the platform uses.
03Review
Batch Review Copilot
Review by exception for paper batch records, your own and your contract manufacturer's.
04Trace
Batch Genealogy with Synapse
Any lot forward, any batch back, and how they were linked on any past date an inspector asks about.
05Trend
Charts
Continued process verification, kept as a controlled record.
06Annual review
Reports
The annual product review, set up once and run again each year.
How it is built
Every result can be inspected.
The checks and relationship rules are deterministic. Every extracted field has a confidence score, every limit has a change history, and every AI output is logged.
Evidence
ERPGoods issueLot 22-0433MCC, to order 4471
MESOrder 4471Batch 24-117step 3.2, charge
CoASupplier 2Lot 22-0433certificate on file
LIMSAssay resultBatch 24-11799.4 %, from LIMS
A relationship
22‑0433 → 24‑117
material lot to batch
How this was found
- Found by
- Relationship rule, deterministic
- Records
- ERP, MES, supplier CoA
- Version
- Every version kept
- If no rule fits
- Sent to an expert
For the expertOpen the genealogy
Evidence
LOD, % · batches 24-101 to 24-124 · sites 1 and 2
2023 Q4 · version 4UCL 2.5 %
2023 Q3 · version 3UCL 2.8 %, kept
A control limit
UCL 2.5 %
version 4
History
- Change
- Under change control
- Before
- Version 3, kept
- Controlled path
- Kept apart from exploratory analysis
For the expertOpen the chart
Evidence
BR-24-117 · page 37 of 412 · executed
6.4Blend, min
6.5Weight, kg
6.6LOD, %
6.7Mesh
6.8Inlet temp, °C
An extracted field
18.40 kg
confidence 0.71
How this was found
- Read by
- Field extraction, page 37
- Confidence
- 0.71, low
- Result
- Routed to a reviewer
- Recorded in
- Part 11 audit trail
JOJ. Okafor verifies and signs
The person decides. The system puts what needs attention in front of an expert first. It never approves a record, releases a batch or decides a disposition.
Validation
Validation included.
Mareana validates the platform and its AI capabilities under GAMP 5 and hands over the validation package, with the OQ evidence. Your team qualifies the platform for your intended use.
- The protocols
- IQ, OQ and PQ, all written by Mareana
- Your part
- Run the IQ and PQ, review the OQ
- AI capabilities
- Each classified by risk and validated in proportion
- Revalidation
- Triggered by a change to a model, its prompts or its data sources
- Part 11 and Annex 11
- Shown through validation and control evidence, not a certificate
- For your audit
- The validation records, open to you under NDA
- Agents
- Gather and prepare. None changes a GxP record without a person's review and approval.
- Exploratory analysis
- Kept apart from the validated path
Build or buy
Why not build it ourselves?
Building on the AI assistant you already have, or on your own data platform, is a real option, and often the first one proposed. The difference is in what your team would have to build, validate and keep up themselves.
Built in-house
- ValidationYour team writes and maintains the validation for the AI and for how records are linked, from the start.
- ScopeUsually one workflow, and it can be a strong one.
- PaperGeneral document tools struggle with handwriting and batch record layouts.
- Review and auditBuilt, and validated, again for each new workflow.
Mareana
- ValidationDone by Mareana under GAMP 5 and handed over, for your team to qualify.
- ScopeOne record across paper, the systems you connect, sites and partners, and every use draws on it.
- PaperBuilt to read executed batch records, handwriting, layouts and most languages included.
- Review and auditCome with the platform: checks a reviewer can inspect, routing to a person, an audit trail and e-signature.
Questions
The questions
that come first.
What about Part 11?
When a reviewer verifies a field or signs off a batch review, the signature is electronic and the step is recorded on a 21 CFR Part 11 audit trail.
Is every module available today?
Yes, all six modules are available today, EBR in beta. The platform is live with customers.
What changes first?
That depends where you start. With batch review, the reviewer's week changes first, with less of it spent checking entries one by one. Where release is waiting on the batch record, that's the wait Mareana is designed to shorten. Where release is waiting on testing or an investigation, the record isn't what's holding it up, and we'll tell you so. We don't put a figure on these changes yet. A proof of concept on your own records is where the change is measured, with the hours and the errors counted separately.
Where would we start?
With the problem in front of you: batch review, an investigation, CPV or the annual product review. The start can be as small as one product at one site.
What does a first step look like?
A call to see whether there's a fit, then a demo on Mareana's own data. Once a confidentiality agreement is signed, we run a small proof of concept on one of your own batch records.
Tell us where you'd start.
We'll show you that part of the platform on records like yours, and how the record it builds carries into the next problem.
Book a DemoBook a Demo- 01Batch review
- 02Deviations
- 03Investigations
- 04Traceability
- 05Impact assessment
- 06CPV
- 07Annual product review
- 08Inspections
- 09Moving off paper
From the blog
Latest from the blog
Practical reading on batch review, traceability and digital records in pharma manufacturing.
Digital Transformation in Pharma Manufacturing: The 2026 Guide to Data, AI and GxP Compliance
Digital transformation in pharma has moved past the question of whether to digitize. Most manufacturers already run an MES, a LIMS, a QMS and an ERP. The work in 2026…
Mareana Deep Dive Vol 10
Welcome to the 10th edition of Mareana Deep Dive. This edition includes our updates from August and September 2026 focusing majorly on inspection readiness. Inspection readiness depends on how quickly…
Pharmaceutical Inspection Readiness with Mareana
Pharmaceutical inspection readiness depends on data infrastructure as much as procedural discipline. If the data proving operational control is fragmented, no amount of mock auditing compensates. • Paper batch records…
