Andrés D. Ramos Luna · Planning and plant-data consultant
What I learned reconciling production plans against the ERP, week after week.
Every assumption about your data is rebuilt with two independent queries, never against a hard-coded value, and at least one is tied to the raw source.
I saw checks stay green for eight weeks while data was being lost. Since then, none of them is compared against itself.
When leveling to finite capacity, what doesn't fit is pulled forward up to ten working days or pushed out with a risk flag. Nothing is redistributed silently.
A plan that fits on paper but not on the machine is a broken promise with a due date.
Every week a snapshot of what was shown is kept. The ERP overwrites its history; your planning shouldn't lose it.
If you can't open what you saw on Monday, you can't learn from Monday.
Real cases in metal manufacturing. Turn on audit mode to see the evidence for each one.
ProblemDated demand by resource only existed as a manual export from a closed ERP procedure.
MethodCascading netting along the routing, split by resource and back-scheduled in working days.
EvidenceReconciled against the export: 98.5% of parts identical line by line; 3 gaps closed with root cause.
ProblemThe ERP releases the plan at infinite capacity: it fits on paper, not on the machine.
MethodDay-by-day leveling against each resource's real ceiling, in its own unit, emitted in the format that is uploaded by hand today.
EvidencePilot with 6 resources and 12 weeks; quantified, per resource, how much of the plan doesn't fit.
ProblemMeasuring output against the plan without inflating the figure.
MethodOutput is capped to plan per part; the surplus is “extra” and unplanned output is “off-plan”. Nothing is discarded.
EvidenceBoard certified by 27 checks; weekly closes that are never rewritten.
ProblemFour nearly duplicate recovery plans, one per customer.
MethodOne shared engine that levels in three tiers, respects routing and multi-level BOM precedence, and iterates to a fixed point.
EvidenceA/B test with zero differences for two customers; conservation verified on every run.
ProblemComparing raw demand snapshots overstated growth.
MethodMaturation curve by lead time and a firm window measured over 12 months of weekly snapshots.
EvidenceRaw and adjusted views at times pointed in opposite directions; the historical snapshot reconciles to 0.03%.
Three ways to start. All of them end with something your team can run again without me.
I review the sources feeding your plan and give each one a verdict.
Netting, material coverage and finite-capacity leveling on top of your ERP.
A monitor that shows how each resource is doing against plan and what is holding it back.
Before modeling plans, I drew the cells that execute them. Layouts made in AutoCAD during my automation years.
From the robotic cell to the data warehouse.
| Rev | Date | Description | Approved |
|---|---|---|---|
| A | 2022 | B.Sc. Mechatronics Engineering, UANL FIME. | andcom |
| B | 2022 | Automation: robotic cells, Siemens and Allen-Bradley PLCs, Yaskawa, OTC and Fanuc robots. | andcom |
| C | 2025 | Move into BI and plant-data analysis on the ERP. | andcom |
| D | Mar 2026 | Data core: source knowledge and the first checks. | andcom |
| E | May 2026 | BOM traceability with work-in-process by station. | andcom |
| F | Jul 2026 | Plan compliance monitor, every 15 minutes. | andcom |
| G | Aug 2026 | Azure SQL data warehouse with versioned migrations. | andcom |
| H | Nov 2026 | Independent production-planning consulting. | andcom |
Before the data, the cell. I programmed PLCs, integrated robots with presses and drew the layouts. So when a plan says a machine can, I know to ask whether it really can.
Siemens and Allen-Bradley PLCs · Yaskawa, OTC and Fanuc robots · AutoCAD · B.Sc. Mechatronics, UANL FIME 2022Netting engine reconciled against the ERP export, line by line per part.
Count of active checks in the data core, November 2026.
A/B comparison of the shared engine against two customers' previous plans.
Scheduled frequency of the plan load and the compliance monitor.
Thirty minutes to review how you plan today and where the figure breaks between the ERP and the floor. If there's nothing to fix, I'll tell you.