Manufacturing
How to inventory your data across ERP, MES, QMS and CMMS
By SourceX Editorial · Updated
Short answer
A manufacturing data inventory is a one-sheet map of what each system holds: record types, years held, linking keys, owner, export method and rights flags for your ERP, MES, QMS and CMMS. Build it from short interviews and record counts, not exports, and test the keys between systems first, because linked records carry most of the value.
Key takeaways
- Inventory systems, not files: one row per live or retired system, with its record types listed inside the row.
- Years held means years you can still open and export, not years since the company or the module started.
- The keys that join systems, such as job, lot, serial and asset IDs, decide whether records tell a complete story.
- Set rights flags on every row before anyone talks about pulling a sample.
- Spreadsheets and retired systems belong in the inventory, because they often hold the oldest history.
The inventory template, one row per system#
The inventory template uses one row per system, including retired systems and the spreadsheets that stand in for missing modules. Every row answers the same seven questions, which keeps even a sprawling plant on a single sheet and makes gaps obvious.
The example entries below are typical starting points to replace with your own answers. Where nobody knows, write unknown; an honest gap is cheaper than a wrong guess about history depth.
| System | Record types | Years held | Linked keys | Owner | Export method | Rights flags |
|---|---|---|---|---|---|---|
| ERP | Quotes, orders, jobs, routings, BOMs, purchasing, shipments, invoices | Since current go-live; earlier years in the prior ERP | Customer, part and revision, job, order, PO | ERP administrator | Report writer, saved queries, API | Customer pricing, customer part numbers, attachments |
| MES | Operation starts and stops, quantities, scrap, downtime, travelers | Check purge settings | Job, operation, machine, lot or serial | Manufacturing engineering | Database views, vendor export | Operator names and badge IDs |
| QMS | NCRs, CAPAs, inspections, audits, complaints, calibration | Often long; may be split with paper | NCR, lot, serial, job, part, supplier | Quality manager | Module exports, report builder | Customer drawings, photos of customer parts |
| CMMS | Work orders, failure codes, PM tasks, parts used, labor | Older years may sit in a prior CMMS | Asset ID, work order, part, location | Maintenance manager | File export, API | Technician names, outside service reports |
| Historian or IoT platform | Sensor tags, alarms, events | Set by compression and retention settings | Tag mapped to asset, timestamp | Controls engineer | Historian query tools | Equipment maker terms |
| Spreadsheets and shared drives | Schedules, expedite lists, inspection logs, cost models | Whatever survived on the drive | Job or part, typed by hand | Each file's keeper | File copy | Customer names, personal notes |
| Retired ERP or QMS | Pre-migration history | Earlier years; confirm it still opens | Old IDs that may need a crosswalk | IT | Database backup, report dumps | Original vendor license terms |
Step by step: running the inventory#
Running the inventory is a sequence of short conversations followed by a few counts, and nobody exports record content along the way. The order below starts with the people who know where decisions get recorded.
- Step 1: list every system, including retired ones and spreadsheets that replace a missing module.
- Step 2: interview each system owner using the seven columns, writing unknown where nobody knows.
- Step 3: have IT run record counts by year for the main record types in each system.
- Step 4: test the keys between systems with counts, such as how many NCRs carry a valid job number.
- Step 5: set rights flags with the quality manager, the controller and counsel.
- Step 6: mark each row keep, review or exclude, and date the sheet.
How to measure years held without fooling yourself#
Years held should measure history you can still open, query and export, not the age of the company or the date a module was bought. The two often differ by a migration or two.
Ask three questions for each system. When did the current system go live, and what history came across? Has a purge, an archiving job or a retention setting removed older records? Is the previous system still installed, backed up and openable, and who holds its license?
Historians and MES platforms need special care. Many keep full-resolution data for a limited period and compress or delete older data according to settings someone chose at installation. Record those settings in the row, because they decide whether older sensor history exists at all.
Which keys connect ERP, MES, QMS and CMMS?#
The keys that connect a plant's systems are the job or work order number, the part number and revision, the lot or serial number and the asset ID. Records that share these keys can be joined into a chain from order to outcome; records that lost them stand alone.
Test each key with counts rather than files. If nearly every NCR carries a valid job number, quality history connects to production; if few do, the inventory should say so plainly.
| Key | Created in | Should appear in | Common break |
|---|---|---|---|
| Job or work order | ERP | MES, QMS, labor records | Rework jobs opened without a link to the original job |
| Part and revision | ERP or PLM | MES, QMS, inspection plans | Revision kept as free text or dropped in the MES |
| Lot or serial | ERP or MES | QMS, shipping, returns, warranty | Serials written only on paper travelers |
| Asset ID | CMMS | MES machine list, historian tags | Different machine names in each system |
| NCR or CAPA number | QMS | ERP rework jobs, supplier returns | Cross-references typed into notes |
| Customer and supplier IDs | ERP | QMS complaints, supplier NCRs | Duplicate masters after acquisitions |
Rights flags to set on every row#
Rights flags mark anything in a row that could limit how its records are used or shared, and they belong in the first draft. A flag is not a decision; it tells the reviewer where to look.
- Customer-owned design: drawings, models, specifications or inspection plans supplied by customers, often stored as attachments.
- Export-controlled: parts, programs or technical data your compliance lead has classified, plus anything nobody is sure about.
- Personal data: operator names, badge IDs, technician notes and email addresses in comments.
- Confidential commercial terms: customer pricing, quality agreements and supplier cost data.
- Vendor terms: software or equipment maker terms that limit export or claim rights in data.
- Third-party documents: supplier certifications, outside lab reports and contractor service reports.
Illustrative: a metal stamping supplier maps four systems#
Illustrative: a fictional metal stamping company that supplies brackets and welded assemblies to appliance and truck makers runs a cloud ERP, an MES on its press lines, a QMS for NCRs and PPAP files, and a standalone CMMS. Its COO wants to know what the plant holds before an ERP upgrade.
The inventory shows that NCRs link to jobs reliably, but the CMMS names presses by location while the MES uses numbered machine IDs, so downtime and repairs cannot be joined. A maintenance planner's spreadsheet maps one to the other and becomes the crosswalk. Customer-owned die drawings sit as ERP attachments and are flagged.
The COO marks NCRs, press downtime and maintenance work orders as keep, marks PPAP files as review because they contain customer specifications, and asks IT to preserve the old ERP database before the upgrade starts.
Mistakes that make an inventory useless#
The most damaging inventory mistake is turning it into an export project. Once people start pulling samples, the work slows, sensitive files spread across laptops, and the questions that matter most, about keys and rights, get skipped.
Other mistakes are listing record families without the system they live in, recording years since founding instead of years held, leaving retired systems off because nobody logs in anymore, and treating spreadsheets as noise. A schedule board kept by one planner may be the only record of how the plant handled changes.
How SourceX uses a system inventory#
SourceX reads a system inventory as the starting point for a metadata-only fit check, and nothing is shared during that initial assessment. Rows carrying customer-design or export-control flags are set aside, and the rest are weighed with the SourceX Enterprise Data Value Framework, including uniqueness, human-generated signal, recency, data cleanliness and rights.
Systems that clear the fit check move through the SourceX five-step transaction of Supply, Rights, Preparation, Approval and Delivery, with the manufacturer signing off at each step. Large histories, such as historian archives, stay in the manufacturer's own storage or ship on encrypted drives.
Frequently asked questions
How detailed should each row be?
Detailed enough that a reviewer can tell what the records are, how far back they go and what limits their use, without opening a file. A short phrase per cell is usually right. If a cell needs a paragraph, split the row into two systems or two record types.
Should paper records be in the inventory?
Yes, as their own rows. Paper travelers, signed inspection sheets and binders of CAPAs may hold the only copy of older decisions. Note where they are stored, how they are organized and whether anyone has scanned them, then decide later whether they matter.
Does IT need to pull anything for the inventory?
Only counts. Record counts by year and simple key-match counts show how deep and how connected each system is. Those numbers can be produced without exporting customer names, prices or free text, and they answer most fit-check questions.
Who should own the inventory once it exists?
Usually the COO or the IT director, with each system owner updating their own row. Date the sheet and revisit it before any migration, system retirement, acquisition or data review, because those are the moments when history is lost or suddenly needed.
Is the inventory shared with buyers?
Not as it stands. It is an internal planning document with system names, owners and restriction notes. Parts of it inform a fit check, and if a package proceeds, a separate, approved description of the records is what a buyer sees.
Related resources
- QuestionTraining data vs evaluation data: what's the difference?
- QuestionDo AI labs buy images?
- InsightCan I license Epicor Kinetic data to AI companies?
- InsightCan roofing contractors sell their data to AI companies?
- InsightCan metal fabricators and welding shops license their data?
- SolutionEnterprise data: the records of how organizations actually work
See if your company qualifies
A short company assessment. No data uploads are needed.