Analytics Approach

I start with the business decision, then work backwards into the data, model, and validation. My priority is correctness and decision usefulness — not flashy visuals.

Principles

The rules I follow to keep dashboards accurate, trusted, and usable under pressure.

1) Define the KPI in plain English
Before writing DAX or building visuals, I write the KPI definition as a sentence anyone can agree on.
Prevents: “looks right” metrics • Enables: alignment + easier debugging
2) Source-of-truth reconciliation
I validate totals against the system that runs the business (ERP/CRM). If totals don’t match, I stop and fix the pipeline.
Prevents: leadership mistrust • Enables: KPI auditability
3) Model to avoid double-counting
I separate the “things that happen” (transactions) from the “ways we slice” (vendor, item, location, date) so filters behave safely.
Prevents: inflated totals • Enables: reliable slicing by any dimension
4) Build for action, not just reporting
I design outputs around “what should we do next?”—priorities, exceptions, owners, and drill paths that support execution.
Prevents: passive dashboards • Enables: faster decisions + follow-through

My workflow

A repeatable process that works across operational, inventory, finance, and automation reporting.

1) Define the decision

What decision will someone make from this? Who uses it, how often, and what “good” looks like.

2) Map the data to the question

Identify source tables/fields and clarify edge cases early (partials, cancels, returns, backorders, adjustments).

3) Shape + model

Clean the grain, standardize keys/dates, and build a model that behaves correctly under slicing and drill-through.

4) Build measures & UX

Implement measures with clear naming, then design visuals around scanning, prioritizing, and drilling to root cause.

5) Validate, ship, and monitor

Reconcile totals, spot-check records, confirm definitions with stakeholders, then monitor for drift (new data patterns, process changes).

Quality checks I always run

These checks prevent silent errors and keep dashboards trusted by ops and leadership.

Totals & reconciliation
  • Compare totals to ERP/CRM views
  • Validate date window logic (cutoffs/time zones)
  • Check the grain (one row = one thing)
Model safety
  • Look for duplicate keys / many-to-many traps
  • Confirm filters flow as intended
  • Test slicing by vendor/item/location/date
KPI sanity
  • Spot-check real records end-to-end
  • Validate edge cases (partials/returns/voids)
  • Confirm definitions with stakeholders

Where this shows up in my work

These projects apply the same approach: clear definitions, reconciliation, and decision-first reporting.

Inventory Velocity
→

Clear definitions for fast/slow/dead stock, with a model that stays stable under slicing and drill-down.

AI Vendor Follow Up
→

Automated data capture + system updates to keep operational fields (like ETAs) consistently maintained.

Purchase Order Health
→

KPI-driven risk view built for follow-up prioritization, validated against Business Central totals.

Daily Sales Order Report
→

Daily operational snapshot with reconciliation and KPI logging for consistent leadership reporting.