BlueeBlack
← All case studies

Data & Analytics · Consumer Goods & Retail · 4 min read

When the Numbers Lie: Building a Data Platform That Tells the Truth

From 72 hours to real-time

Reporting cycle transformation across all business units

A consumer goods company whose leadership team was making decisions on 3-day-old, frequently incorrect reports now runs on a real-time analytics platform, with one version of the truth shared across every business unit.

Overview

Data isn't valuable when it's old, siloed, or wrong. For a mid-market consumer goods company operating across manufacturing, distribution, and retail, all three problems applied simultaneously. Monthly performance reviews were exercises in reconciling conflicting spreadsheets. Sales reported one number, finance reported another, and operations had a third. Decisions were slow, and sometimes made on information that was demonstrably false by the time it was acted on. BlueeBlack was engaged to build the data infrastructure this business needed: a clean, unified, real-time platform that everyone trusted and everyone used.

The Challenge

The company's data landscape was the product of 12 years of organic growth and point-solution procurement. Sales data lived in a CRM. Inventory in a standalone ERP module. Financial data in an accounting system. Production metrics in spreadsheets maintained by supervisors. These systems didn't talk to each other. Every report required a human to pull data from each source, paste it into a master spreadsheet, and reconcile the discrepancies, a process that took 2–3 days and still produced questionable accuracy. Leadership had no real-time view of the business. Weekly review meetings regularly descended into arguments about whose numbers were correct rather than discussions about what to do.

  • 01Data distributed across 6 disconnected systems with no integration
  • 02Reporting cycle took 72+ hours, making data stale before it reached decision-makers
  • 03No single definition of core metrics (revenue, margin, inventory turns), each team had their own formula
  • 04Finance's and sales's revenue numbers differed by up to 11% in a given month
  • 05No self-service capability, every report request went to a single overworked analyst
  • 06Historical data was inconsistently stored, no reliable basis for trend analysis or forecasting

The Approach

BlueeBlack began with a Data Audit & Metric Alignment Workshop, a structured 2-week process where finance, sales, operations, and leadership agreed on precise, shared definitions for 22 core business metrics. This resolved the political problem before touching a single data source. The technical build followed a modern data stack architecture: source system connectors feeding a cloud data lake, transformation via dbt, a data warehouse as the single source of truth, and a BI layer surfacing governed metrics in business-user-friendly dashboards.

What we built

  • Data Ingestion PipelineAutomated connectors extracting data from 6 source systems (CRM, ERP, accounting, point-of-sale, production floor, logistics) on configurable schedules, with incremental load and schema change handling
  • Cloud Data WarehouseCentralized warehouse on AWS Redshift serving as the single source of truth, with version-controlled data models built in dbt
  • Transformation & Metric Layer22 core business metrics defined, documented, and computed in dbt models, ensuring every dashboard pulls from the same calculation, not a local formula
  • Real-Time Operational FeedsKey operational metrics (stock levels, production output, daily orders) updated every 15 minutes via streaming pipeline
  • Self-Service BI LayerMetabase dashboards organized by function (Sales, Finance, Operations, Supply Chain), designed for non-technical users with drill-down, filter, and export capabilities
  • Executive ScorecardOne-page daily brief auto-generated each morning with KPIs for each business unit, delivered via email and a pinned dashboard

The Outcome

The first full month after go-live, finance and sales reported the same revenue number for the first time anyone in the company could remember. The 72-hour reporting cycle was replaced by dashboards that refreshed every 15 minutes for operational data and daily for financial summaries. The analytics team's ad hoc report queue, which had regularly built up to a 3-week backlog, dropped to near zero as business users started serving themselves. Leadership meetings shifted from reconciling numbers to acting on them.

Services — Data Engineering · Analytics Engineering · BI & Visualization · Data Strategy & Governance

Stack — AWS Redshift · Apache Airflow · dbt · Python · Metabase · Kafka (streaming) · PostgreSQL · REST APIs for source connectors

Impact at a glance

Reporting cycle
72 hoursReal-time / 15-min refresh
Revenue figure variance (Finance vs Sales)
Up to 11%0% (single source)
Ad hoc report queue
3-week backlogNear zero
Analyst time on report production
70% of capacity20%
Dashboards available self-service
234
Historical data depth for forecasting
18 months (patchy)5 years (clean)

Next case study

Always On, Always Right: Deploying an Agentic AI Assistant at Scale

Be the nextwrite-up.

Free consultation · Replies within 24–48 hours