# Martin Bartovsky

> Independent ERP and BI consultant near Paris. Numbers you can trust, from the warehouse to the board report.

Contact: martin@bartovsky.com
LinkedIn: https://www.linkedin.com/in/bartovsky

## Pages
- [About](https://bartovsky.com/about)
- [Contact](https://bartovsky.com/contact)

## Work
- [Eight dashboards, eight decisions](https://bartovsky.com/work/decision-dashboards): Representative Power BI dashboards - receivables, liquidity, P&L against budget, production delays, labour hours, stock, margin and machine effectiveness. Each answers one management question and ends with a read-out - what changed, why, and what to do about it. Built from scratch with synthetic data.
- [ERP migration: the work around the software](https://bartovsky.com/work/erp-migration-mid-sized-manufacturer): Supervising a move from a legacy ERP to a new system - infrastructure, a role catalogue that denies by default, and master data cut down to what is actually used.
- [One route for every delivery date: mapping a purchasing process](https://bartovsky.com/work/process-map-delivery-dates): Production stopped for missing material that was, on paper, ordered and on its way. A process map in BPMN showed four handovers where the delivery date got lost - and turned each into a rule the ERP had to support.
- [PDM and ERP: which system owns which part](https://bartovsky.com/work/pdm-erp-ownership-split): Engineering works in a PDM vault, everyone else in the ERP. Splitting ownership - purchased parts mastered in the ERP, made parts and structures in the vault - and the harder question of what happens to the manufacturing BOM on the next import.
- [Receivables risk across a group of companies](https://bartovsky.com/work/receivables-risk-group): Which receivables are at risk, where, and how the risk develops - a Power BI model over regular ERP snapshots, with comments that stay valid only as long as they should.
- [Shortages and late orders, seen before the order is late](https://bartovsky.com/work/production-shortages-delays): Power BI for production planners - which components will run short, which operations are already behind - and why one large "shortage" was really work in progress the ERP could not see.
- [One repository for Power BI, several suppliers](https://bartovsky.com/work/power-bi-shared-repository): An internal team and external suppliers building reports in the same tenant. Version control owned by the client, access through groups, nothing written straight to production - now in live use.

## Notes
- [I run AI agents on my own server - and client data never goes near them](https://bartovsky.com/notes/ai-agents-two-zones-researcher-critic): Autonomous agents are useful precisely because they work while nobody watches. That's also the risk. How I split data into two zones, and why the most valuable agent in my setup is the one that checks the other.
- [Azure DevOps for Power BI with external developers: the set-up, step by step](https://bartovsky.com/notes/azure-devops-power-bi-external-developers-setup): Entra security groups, B2B guests, group rules that hand out licences, a protected main branch and a Fabric workspace that follows it - the PowerShell and Azure CLI to set it up, and what to check on the way.
- [BOM health: the numbers that show the process still works](https://bartovsky.com/notes/bom-health-kpis): Rules for BOMs and engineering changes can be measured - and without measuring, they fall apart within a year. Eight indicators, the three to start with, and where the data sits in SAP S/4HANA for a Power BI report.
- [Ten tests for your BOM import before go-live](https://bartovsky.com/notes/bom-import-ten-tests): The most important test of a BOM import isn't the first load - it's the second. Ten scenarios to run with the implementation partner on one real assembly, the clean-up rules before migration, and a pilot plan with success criteria.
- [From Excel to PLM: six levels of EBOM-to-MBOM tooling](https://bartovsky.com/notes/ebom-mbom-tooling-six-levels): Getting engineering data into the ERP and turning it into a manufacturing BOM are two different problems. Six levels of tooling ordered by entry investment, which one to start with, and how the levels map to SAP S/4HANA.
- [Engineering changes: one number, revision or new part, and where-used first](https://bartovsky.com/notes/engineering-change-one-number): Most BOM errors aren't wrong changes - they're changes made in one product that silently apply, or don't, in others. One change number with an impact type, the form-fit-function rule, phase-out as an approved step, and a where-used list on every change.
- [Your ERP will import BOMs from Excel. The second import is the problem.](https://bartovsky.com/notes/erp-bom-import-second-import): The engineering BOM says what a product is made of; the manufacturing BOM says how it is made. Moving one into the other is not an export and an import - it's a compare-and-merge, and the rules you set at go-live will hold for years.
- [Claude Code and Codex on a locked-down Azure VM: set-up and lessons learned](https://bartovsky.com/notes/headless-ai-agents-azure-vm-lessons): A private server with no public inbound port, agents without admin rights, a Telegram gateway and a permission model that looked tighter than it was. The build, the commands and the mistakes worth avoiding.
- [When does an intermediate deserve its own item number? Seven criteria](https://bartovsky.com/notes/intermediate-item-seven-criteria): Every stock record for a semi-finished part costs an order, an issue, a receipt and maintenance on every engineering change. A rule for deciding, per part family, whether a blank or a weldment gets its own item - or stays an operation on the production order.
- [Running a local LLM on a laptop: Ollama, Mistral and Open WebUI in Docker](https://bartovsky.com/notes/local-llm-ollama-open-webui-docker): A private ChatGPT-style assistant that never sends a prompt to the cloud - set up step by step on Windows, with an honest look at what it can and can't do on an ordinary business laptop.
- [Running an ERP project in Microsoft Planner: buckets, labels and rules for four companies](https://bartovsky.com/notes/planner-erp-project-basic-plan): A client, an ERP vendor, an MES vendor and an e-shop agency in one plan - using only what a Planner basic plan offers. How to set up buckets, the 25 labels, task naming, checklists and the weekly rhythm, with PowerShell to build and report on the plan, and what an MCP server could add.
- [More than one Power BI supplier? The repo belongs in the client's tenant](https://bartovsky.com/notes/power-bi-multiple-suppliers-repo-ownership): When internal analysts and external suppliers build reports together, versions live on laptops and knowledge leaves with whoever leaves. Four rules for putting Power BI and Fabric under version control that the company - not the supplier - owns.
- [The quote is the first version of the BOM](https://bartovsky.com/notes/quote-is-first-bom): In a quote, the estimator already decides the blank, the operations, the tool and the purchased parts - before any item exists. The cheapest big saving isn't a new tool, it's costing in the same work centres, operations and rates as the ERP.
- [A free SAP system on your laptop - and reading it from Power BI](https://bartovsky.com/notes/sap-abap-trial-docker-power-bi): SAP's ABAP Cloud Developer Trial runs in Docker on Windows. Set-up, a first custom table filled from SAP demo data, and the bridge to Power BI - without touching anyone's production system.