Business Intelligence Developer — Supply Chain & Finance Reporting (Power BI)

The BI layer, end to end: semantic models, reports, refresh infrastructure, and the data-quality gates that sit around all of it.

That includes converting what our existing Python pipelines, carrier data ingestion and month-end build scripts currently produce into Power BI dashboards, taking those feeds and turning them into governed, refreshing, self-serve reporting instead of static workbooks.

• PostgreSQL (read-only): orders, shipments, inventory, purchase orders, freight invoices, product cost master
• E-commerce platform APIs across multiple storefronts, brands and regions
• Warehouse-management / 3PL platform (API plus operational exports)
• Marketplace APIs
• Carrier billing feeds, parcel invoices at tracking level, freight at consignment level (deliberately different grains)
• Google Sheets used as live operational trackers
• Vendor statement workbooks, one bespoke layout per supplier
• Regional carrier and 3PL rate cards, in multiple currencies

• →Model purchase orders at both PO header and PO line-item grain, with the relationship between them made explicit and documented
• →Build star-schema semantic models across shipments, carrier cost, orders and revenue, inventory and landed cost
• →Maintain conformed dimensions (Date, Brand, Warehouse, Carrier/Service, Region, Product, Supplier, Store), with an explicit documented fact grain per fact table. Shipment, order, line and parcel grains coexist here, and naive joins produce figures that are internally consistent and badly wrong
• →Write DAX for shipping cost as % of revenue, cost per order, cost per unit, cost per label, invoiced-vs-estimated mix, landed cost, gross margin and accrual movement

• Own the Power Query, dataflow and SQL view layers
• Run incremental refresh, the on-prem gateway, scheduled refresh, and refresh-failure alerting
• Implement merge-with-carry-forward refresh semantics and row-conservation checks, so an incomplete upstream pull can never silently shrink a published dataset
• Handle open dimensions: source systems introduce new carriers, services, warehouses and categories without notice, and models must surface unexpected members visibly rather than dropping them or falling over

• Treat row-level detail as a first-class output: paginated reports, Analyze in Excel, full export. Reviewers audit raw rows, not summaries
• No hidden rows, no Top-N truncation, no residual "Other" bucket. Any exclusion is counted, logged and auditable

• Model the synced spine, replace formula chains with relationships, implement write-back for the hand-entered fields (Power Apps/Dataverse, SharePoint list, or an app table), and rejoin it in the model

• Automated validation on every model: row conservation (inputs = outputs + logged exclusions), schema parity against the prior good version, and regression anchors re-asserting known headline totals on each refresh
• Reconcile to existing outputs and to Finance to the cent before publishing; stop and escalate on unexplained movement
• Maintain a data dictionary and lineage for every published measure — source, grain, known limitations
• Prove changes on a sample copy, present before/after for anything that moves a published figure, and promote only with explicit sign-off
• Manage RLS, workspace and app permissions, and enable business users

• 3+ years of production Power BI: Desktop, Service, workspaces, apps, deployment pipelines, gateways, scheduled and incremental refresh
• Strong DAX: filter context, CALCULATE semantics, time intelligence, grain-safe additive measures
• Strong Power Query / M over messy sources: multi-header spreadsheets, wide-to-long unpivots, inconsistent vendor layouts
• Strong SQL, ideally PostgreSQL: window functions, tuning, and the judgement to push transformation upstream. Power-Query-only candidates will not cope with these volumes
• Dimensional modelling: star schemas, conformed dimensions, fact grain, SCDs, bridge and many-to-many patterns
• Python (pandas) at read/run/modify level. You'll inherit feeds from existing pipelines; you're not expected to write new ones
• Advanced Excel. Business reviews happen in Excel, and your paginated and exportable output needs to hold up there
• Demonstrated finance-grade reconciliation: tying a report to a ledger, invoice file or supplier statement and explaining every residual
• Clear written communication: documented assumptions, disclosed limitations, reviewable change notes
• Supply chain, logistics, e-commerce or 3PL domain knowledge: carriers, zones, service levels, freight vs parcel, landed cost, duty and tariff codes, volumetric measures, accruals
• Carrier invoice reconciliation
• E-commerce, marketplace or WMS API integration
• Experience replacing a business-critical spreadsheet with a governed model, including write-back patterns
• Multi-currency and FX handling
• Paginated reports (Report Builder), Tabular Editor, DAX Studio, ALM Toolkit
• Git and source-controlled Power BI projects (PBIP)
• Microsoft Fabric, Azure Data Factory/Synapse, dbt, DuckDB — genuinely preferred, genuinely not required. We have no committed Fabric migration today