The problem
Sheet 02 · Rev A
Most business data is never used. Not because it's missing.
Every system keeps its own records. Not one field connects them.
IDC · Rethink Data · 2020 · 1,500 respondents · pp. 5, 28
68%
of the data a business already holds is never used.
Manufacturing ranked lowest of all industries surveyed in data automation. The top barrier was not budget or security. It was making the data usable.
Already paid for
The machines that made the data are paid for, and the people who typed it in are on payroll.
The server that holds it is backed up every night.
The money is already spent. The data is already there.
Almost nobody sells the fix, so almost nobody buys it.
The records
Five files and five key fields. No join.
The ERP knows the job number. The machine logs cycle time and tool number. Inspection lives in the quality system, on scanned forms, and in v3-FINAL-use-this-one.xlsx.
- erp_jobs.csvkey: job_no
- machine_log_2024.txtkey: Job
- quality_results.dbkey: JobNumber
- inspection_form.pdfkey: (none)
- v3-FINAL-use-this-one.xlsxkey: Job #
- 5 files · 5 key names · 0 shared
A Tuesday question
“Do we scrap more on this alloy, on machine four, on second shift, right after a tool change?”
You probably can't answer it.
You have all of it.
Every number is in the building, and no two files share a key.
- 01The dashboard nobody opens.
- 02The consultant's one model, one invoice.
Both started downstream of the problem.
Two clocks
Nothing dramatic happens. That's the trap.
Scrap is absorbed into margin, rework into hours, rush freight into the invoice. Each is small enough to swallow, so none is measured.
Clock 01 · The veteran
- · which machine drifts in July
- · which jobs never run at night
- · when the tool needs changing early
- · which supplier's lot runs hot
- · what the notebook means
- Storage location: his head · backup: none
He knows which machine drifts in July. He is 58. None of it is written down.
Clock 02 · The history
Every unstructured quarter is a bigger cleanup later. Prediction runs on history, so starting late moves the finish.
The warning is usually already in your records.
Five buildings
The shop is one version. Run it through your building.
Manufacturing
- The system
- ERP or MES
- The scrap
- Scrap batch
- The veteran
- Floor lead
Contract manufacturing
- The system
- ERP, plus customer portals
- The scrap
- Late lot, rework
- The veteran
- Scheduler who knows which job slips
Dental / Medical
- The system
- Practice management
- The scrap
- No-show
- The veteran
- Front desk manager
Distribution
- The system
- WMS or ERP
- The scrap
- Dead SKU, stockout
- The veteran
- Buyer who reorders in October
Law
- The system
- Practice management
- The scrap
- Over-budget matter
- The veteran
- Partner who knows which case settles
Same three facts.
- 01The data exists.
- 02Nothing joins it.
- 03Decisions run without it.
Next
Start with the exploratory phase.
One dataset, one model trained on your history. $12,500, credited in full.