USA · FMCG Manufacturing
Hollandia Dairy
“Every month, the finance team at Hollandia Dairy would start the same exhausting ritual.”
From 6-day manual FP&A to same-day insight - across Finance, Production, Sales & Inventory.
Key Results
Same day
FP&A reporting
Was 6–8 days of manual work
5–15%
Sales uplift
From optimised replenishment
−40%
Stockout reduction
AI demand signals 10–14 days ahead
−30%
Spoilage reduction
Better production alignment
Tech Stack
The Situation
Four teams - Finance, Production, Sales, and Inventory - would each export their own version of the data. Someone would spend days chasing updated files over email, copying numbers into a master spreadsheet, reconciling conflicts, and eventually producing a report that was already a week old before it reached leadership. When a decision needed to be made, it was always based on last week's reality. And if someone asked "why are our stockouts up this month?" - nobody could answer that quickly. This is not a technology problem. It's a structural one. When data lives in four different systems and nobody owns the connection between them, every reporting cycle becomes a manual operation. And manual operations don't scale.
If you run operations in FMCG or dairy manufacturing, this probably sounds familiar:
- ✓
Your FP&A cycle takes a week or more - and half that time is spent chasing updated files
- ✓
Finance has one number, Sales has another, Production has a third - and nobody knows which is right
- ✓
Stockouts and spoilage are reactive problems, not something you see coming
- ✓
Every "quick" data question takes a day or two to get answered
- ✓
Your dashboards are always one week behind actual performance
- ✓
You've invested in ERP and BI tools but still can't get a unified view across teams
If three or more of these describe your operation, you're looking at the right case study.
The Root Problem
- 1
FP&A reporting took 6–8 days per cycle - almost entirely manual extraction, reconciliation, and report assembly
- 2
Finance, Production, Sales, and Inventory each operated on separate systems with no shared data layer
- 3
The same KPI (e.g. margin, yield, stock level) showed different values in different reports - causing repeated disputes
- 4
Demand forecasting relied on gut feel and lagging sales data - leading to both stockouts and costly over-production
- 5
No self-serve data access - every business question required analyst intervention and a day or two of wait time
How We Fixed It
Audit the data landscape first - no assumptions
Before writing a single line of code, we mapped every data source across the four teams: what system it lived in, how often it updated, who owned it, and how it connected (or didn't) to adjacent systems. This revealed that three of the four systems had usable APIs - and one required a file-based export approach. Knowing this upfront shaped the entire architecture.
Build the Unified Data Layer on Databricks
We built a centralised Databricks lakehouse as the single destination for all four operational domains. Data from Finance (ERP exports), Production (MES feeds), Sales (CRM + POS), and Inventory (WMS) now flows into one governed layer on a defined schedule - with lineage tracked end-to-end. No more email file chains.
Define business logic once - enforce it everywhere
We built a semantic layer on top of the unified data that encoded every business definition - what "gross margin" means, how yield is calculated, which inventory counts are live vs. reserved. This layer became the single source of truth. When Finance and Production now look at yield, they're looking at the same number, calculated the same way.
Deploy Conversational BI for self-serve querying
Rather than building another set of dashboards that people would check once and ignore, we deployed a Conversational BI interface allowing business users to query the unified data in plain English. "What were our top 5 SKUs by margin last week?" - answered in seconds, not days. No analyst required.
Layer AI demand forecasting on top
With clean, unified data available, we layered AI-driven demand forecasting across SKU-location combinations. The model ingests historical sales, seasonal patterns, and production capacity constraints to generate replenishment signals 10–14 days ahead. Buyers no longer wait for stockouts to happen - they act before them.
Measured Outcomes
FP&A cycle time
6–8 days manual
Same-day button click
↑ Key win
Sales performance
Baseline
5–15% uplift
↑ Key win
Stockout incidents
Reactive, uncontrolled
20–40% reduction
Dairy spoilage
Reactive, uncontrolled
10–30% reduction
Data access
Analyst-dependent, 1–2 days
Self-serve NLQ in seconds
Metric consistency
Different numbers in every report
One definition, one number
What This Means For You
What this means if you're in FMCG or dairy manufacturing
The week-long FP&A cycle isn't just an inconvenience - it's costing you margin. Every day of delay is a day where stockouts are happening, spoilage is accumulating, and production is running on assumptions rather than actual demand signals. The technology to fix this exists and is not as complex or expensive to deploy as most companies assume. The first step is getting your data into one place with one agreed set of business definitions. Everything else - forecasting, NLQ, dashboards - becomes straightforward after that.
Next Step
Is this your situation?
Book a 30-minute call. No slides, no pitch. We'll look at your specific setup, tell you what's causing the problem, and what a realistic fix looks like - including timeline and cost range.