Power Platform in logistics — What I learned going from operations to IT
I started in operations — loading docks, schedules, the daily chaos of freight movement. When I moved into IT, I knew the problems intimately. That made the tools easy to choose.
This is what I built, how I built it, and what I’d tell someone starting from the same place.
Power BI — dashboards that get used
The fleet runs on Excel. The problem isn’t data — it’s making sense of it across a dozen spreadsheets with different formats, naming conventions, and levels of completeness.
Shipments by Volume — monthly view: Volume trends per client over 3 years, with windowed analysis by year, quarter, month, and week. Filters by origin, destination, and zone. The semantic model runs on clean SQL views — no transformations in Power BI, just well-structured data from the start.
Performance — on-time vs delayed: Weekly breakdown showing on-time delivery rate per client and location. The metric that actually matters to operations. Updated weekly via a Power Automate flow that queries the database, builds an HTML table, and emails the ranking to stakeholders.
Delay reason analysis: A line chart showing delay trends over time with a stacked ranking of root causes. Turned “why are we always late” from a gut feeling into a data question with answers.
I use DBeaver to build the semantic model. Named conventions, documented columns, incremental refresh on large tables. The SQL layer is the most important part — no visualization fixes bad data.
Power Automate — 9 flows in production
These are the patterns that do the actual work:
Bookings pipeline (create/update/cancel): Three flows triggered by calendar events. A new booking logs to Excel and SharePoint, sends a confirmation email, and creates a calendar event. Updates and cancellations follow the same pattern with appropriate notifications. The triad keeps the dataset in sync automatically.
HR onboarding (ALTA): The biggest flow. Triggered by a new SharePoint item, it runs through approval stages (manager → HR), then auto-generates 16 document types using Word templates, creates a folder structure, converts documents to PDF, sends onboarding emails, and creates calendar events. Multiple approval stages with conditional routing based on employee type.
Weekly performance pulse: SQL query → HTML table with color-coded rankings → branded email with company signature. No dashboards to check — the data arrives in the inbox.
Monitor flows: Running daily — insurance expiry tracking (multiple insurers with different renewal cycles), HR dates (birthdays, passports, medical checks, CAP certificates, licenses, A1 forms, tachograph expirations). Each has empty-result handling so no alert means nothing to report.
Training authorization: Takes a Word template, populates it from SharePoint data, converts to PDF, and emails it to the driver. Replaced a manual document generation process that took 15 minutes per driver.
Power Apps — RBConnectHub
A container scanning app I built before QRFleet. Drivers scan a QR code on a container, the app shows vehicle info and lets them start the reporting process. It works within the Microsoft ecosystem but has clear limits:
- UI customization is constrained
- Image handling requires workarounds
- Performance degrades as the dataset grows
That experience directly led me to build QRFleet with FastAPI + Next.js for full control.
What I’d tell someone starting
Know the business problem before choosing the tool. I built better dashboards because I lived the process. Power BI answered questions I already knew were worth asking.
Power Automate is deceptively simple. A flow that runs fine at 10 executions a day will break at 200. Think about scale — API limits, concurrency, error handling, logging. Every flow should log its executions somewhere you can debug later.
SQL is the bottleneck. If your semantic model is weak, no visualization will save it. Invest time in clean views, documented columns, and consistent naming. It pays for itself in the first week.
Use AI as a force multiplier. I use language models for DAX expressions, HTTP request templates, regex patterns, and debugging flows. Not as a replacement — as a tool that removes friction from the repetitive parts.
Your operations background is an advantage. You know what questions matter. Most people learn the tools and then try to find problems. You already know the problems.
What’s next
A unified view that combines BI (what happened), real-time ops (what’s happening now), and alerts (what needs attention) in one interface. The pieces are all there — Power BI, Power Automate, SQL, and maybe n8n for the integration layer. It’s about connecting them cleanly.
