The bottom line
Running seven systems that don't talk to each other doesn't mean you have data - it means you have seven versions of the truth. FMCG businesses that break this pattern and unify into a single governed layer stop arguing about numbers and start acting on them.
In This Article
The Seven-to-Twelve System Problem
A typical mid-size FMCG company runs between 7 and 12 operational systems that each hold a partial view of the business: an ERP for procurement and financials, a Distributor Management System for secondary sales, a WMS for warehouse management, a TMS for logistics, a trade promotion management system, a retail execution tool, and a collection of regional Excel files that bridge the gaps between all of them.
Each of these systems has its own SKU master, its own customer hierarchy, and its own definition of key metrics. Inventory on hand means something different in the WMS, the ERP, and the DMS. A customer is identified by different codes across sales, logistics, and finance. These are not data quality problems - they are data architecture problems, and no amount of data cleansing resolves them without structural integration.
The result is an organisation where every cross-functional question requires a manual data pull and a reconciliation exercise. S&OP meetings spend the first 30 minutes debating whose numbers are right rather than making decisions. That is not a people problem. That is what seven-to-twelve disconnected systems produce.
When your S&OP meetings start with debating the numbers rather than making decisions, the meeting is not the problem - the data architecture underneath it is.
What the Silos Actually Cost
The cost of data silos in FMCG shows up in three places. First, forecast accuracy below 70% - not because the forecasting methodology is wrong, but because the demand signals coming from DMS, POS, and ERP are inconsistent and arrive at different frequencies. The model is forecasting noise as much as demand.
Second, stockouts that coexist with overstock in adjacent warehouses. This is the operational paradox that siloed inventory data creates: the replenishment system sees a stockout at the depot level, orders more, and the new stock arrives next to inventory that was simply invisible to the system. Inventory is in the right place. The data is not.
Third, missed promotional uplifts. When the trade promotion management system is not connected to the demand plan, promotional periods are not factored into the forecast. The sales team beats plan during a promotion. The supply chain runs out of stock during the peak. The supply chain team gets blamed for not planning properly - when the data to plan with was never shared.
Building a Single Source of Truth
A focused data integration programme connecting the top four to five operational systems - ERP, DMS, WMS, and trade promotion data - into a single governed platform typically takes 3–6 months to reach reliable, decision-grade data quality. The foundation is master data: a single SKU master, a single customer hierarchy, and agreed definitions of the five to ten metrics that drive operational decisions.
Once master data is governed and the integration layer is in place, the S&OP meeting changes character. The first 30 minutes stop being about reconciling numbers and start being about deciding what to do with them. That shift - from reconciliation to decision - is the most immediate and measurable benefit of breaking the silos.
The technology to do this is not exotic. Microsoft Fabric, Databricks, or Snowflake connected to the source systems via CDC or API integration, with a governed semantic layer on top. The hard part is not the technology - it is the master data rationalisation and the organisational agreement on definitions that precedes it.
What One Truth Looks Like on Microsoft Fabric
Concretely, the bridge above the silos is one governed store the systems feed into, not a master system they all migrate onto. We land the ERP (SAP S/4HANA, SAP ByDesign, or Microsoft Dynamics 365), the DMS, the WMS, and the trade-promotion data into OneLake on Microsoft Fabric — change data capture and API connectors handling the integration, Azure Data Factory orchestrating the batch feeds. The source systems keep doing their jobs; what changes is that one copy of the data now exists where every function can read it.
The master data work is the spine: a single SKU master, one customer hierarchy, agreed definitions of the five to ten metrics that drive decisions. Those definitions live in a Power BI Direct Lake semantic model, so inventory on hand, forecast accuracy, OTIF, and fill rate mean the same thing in every report. When a number is questioned, lineage shows which system and which record produced it — the reconciliation argument ends because the answer is on the screen.
That single foundation is also what makes the rest of the supply chain analytics estate possible. The same governed data that settles the S&OP debate feeds demand sensing, replenishment, and the control-tower view of OTIF by lane. You build the bridge once and everything downstream runs on it, rather than rebuilding a new integration for each initiative.
Where This Still Breaks
The technology is the easy 30%. The hard 70% is master data and definitions, and that is organisational, not technical. Every silo has an owner who built it to serve a real need, and asking them to accept a shared SKU master or customer hierarchy can feel like losing control. Without someone empowered to settle those disputes, the integration lands but the definitions never converge — and you end up with connected systems that still disagree.
This is where a single accountable owner matters. The Fractional CDO model exists precisely for this: embedded senior data leadership that owns the master data and the definitions across functions, a few days a week, without a full-time hire. Where that ownership is absent, silo-breaking programmes stall in committee.
And the honest limit: a unified platform will not fix incentives that reward each function for its own number. If sales, finance, and supply chain are each measured on metrics that pull in different directions, they will keep maintaining the spreadsheets that protect their version. The single source of truth makes the disagreement visible; leadership still has to resolve it.
Industrial businesses don't lack data. They lack a single version they can trust.
Sequencing the Integration So It Lands
Breaking silos fails most often not on the technology but on trying to do everything at once. The sequence that lands starts with master data — agreeing one SKU master and one customer hierarchy — because every downstream join depends on it. Skip this and you connect the systems only to discover they still disagree, which is worse than before because now the disagreement is automated. The unglamorous fortnight spent rationalising codes is what makes the rest work.
Then connect the highest-value systems first, not all of them. For most FMCG businesses that is the ERP and the DMS — procurement, financials, and secondary sales — landed into OneLake on Microsoft Fabric via change data capture, with Azure Data Factory orchestrating the feeds. The WMS and trade-promotion data follow. Each system added has to map to the agreed master data on the way in, so the governed layer stays consistent rather than accumulating new variants.
The semantic layer is the payoff. A Power BI Direct Lake model exposes one definition of inventory, OTIF, fill rate, and demand that every function reads from, with lineage back to the source record. From that point the S&OP debate is settled by the screen, and the same governed foundation carries the supply chain analytics that come next — control-tower visibility, demand sensing, automated replenishment — without a fresh integration each time.
Three to six months gets a mid-market FMCG business to decision-grade data across the top four or five systems on this sequence. The ones that try to boil the ocean — every system, every region, at once — are still reconciling spreadsheets a year later. Order and restraint are the whole trick.
What Changes for the Operations Leader
The measurable shift is the S&OP meeting itself. The first half-hour stops being a reconciliation exercise and becomes a decision: which SKUs to build, where to move stock, how to cover the promotion. That change — from arguing about the numbers to acting on them — is the most immediate return on breaking the silos, and it shows up in the first cycle after the foundation lands.
It also does not require a multi-year programme to start. A six-week Discover and Foundation build connects the top four to five systems into a governed Microsoft Fabric lakehouse with a first set of Power BI reports the planning team uses daily — first value in 6 weeks, with the remaining systems folded in as the estate matures.
Most FMCG operations leaders already know their seven systems hold seven answers. The foundation turns that into one answer with the lineage attached — and the stockouts-next-to-overstock paradox, the missed promotions, and the sub-70% forecast accuracy start to resolve because they were symptoms of the same architecture problem all along.
Data silos are a political problem as much as a technical one. Every silo has an owner who built it to serve a legitimate need. The solution isn't to tear them down - it's to build the bridge above them that gives every function a shared view without anyone losing control of their data. That's what a governed data platform actually does.
Free Assessment
Where does your operation sit on the data maturity curve?
8 questions. 3 minutes. You get a scored breakdown across data infrastructure, analytics readiness, and automation potential — with a specific next step for your industry.