FMCG Supply Chain Analytics: The Three Data Problems Every Distributor Has

The standard FMCG demand forecast runs on sales history — what was invoiced to the trade. For most distributors in the GCC and India, this means the forecast is based on what the distributor sold to retailers, not what retailers sold to consumers. In a market with seasonal patterns, promotions, and significant channel stocking behaviour, these two numbers diverge substantially. The distributor sells 10,000 cases of a product to retailers in March because a promotion is running. The retailer sells 4,000 cases to consumers. The forecast system sees 10,000 cases of March demand and projects 8,000 cases for June. The actual consumer demand, from which the real replenishment requirement should be calculated, was 4,000.

This is the demand signal problem: the data the distributor has (sell-in to trade) is not the data the forecast needs (sell-through to consumer). Solving it requires secondary sales data — data from the retail end of the chain, either from the distributor's own sales force reporting system, from the retailer's point-of-sale data (available from some modern trade retailers), or from a field force app that captures in-store visits and stock-on-shelf observations.

The analytics fix is building a demand sensing model that combines sell-in data (which is clean and complete in the ERP) with sell-through data (which is partial, lagged, and messy) and external signals — weather, promotional calendar, competitive pricing events. This is not a complex model, but it requires a data pipeline that brings all three data sources together in a governed layer, and a forecasting logic that weights them appropriately by SKU and channel.

Trade promotions are a significant investment for FMCG distributors and brand owners — typically 5-15% of net revenue depending on category and channel. The return on that investment is almost universally calculated in arrears: the promotion runs in March, the data is collected in April, the finance team analyses it in May, and the result lands in a slide deck in June. Three months after the promotion ended, you know whether it worked. By which point you have already committed to the next promotion cycle.

The data problem driving this lag is structural. Promotion costs sit in the trade spend system or in finance. Promotional lift data sits in the ERP (sales invoices during the promotion period). Baseline sales for the pre-promotion period sit in the same ERP. Consumer sell-through during the promotion sits in secondary sales data, which may or may not exist. Combining these four data sources, with proper time-period alignment and category-level attribution, is the analytical problem that most FMCG distributors have not solved.

The fix is a trade promotion analytics model: a data pipeline that pulls ERP sales data, trade spend data, and secondary sales data into a unified analytical layer, calculates uplift against a defined baseline methodology, and produces SKU-level ROI within two weeks of promotion end. Once the model is built and the data pipelines are reliable, the incremental cost of running it for every promotion is negligible.

Secondary sales data — what retailers and sub-distributors actually sell to end consumers — is the most valuable data in FMCG analytics and the hardest to collect reliably. In the GCC, where modern trade retailers (Carrefour, LuLu, Géant, Spinneys) have relatively mature data-sharing capabilities, secondary sales data is available but requires a retailer-specific integration for each chain. In India, the combination of modern trade, general trade, and kirana stores means that comprehensive secondary sales data simply does not exist for most mid-market FMCG companies.

What does exist is partial: field force visit data from a sales force app, stock-on-shelf observations from distributor merchandisers, order data from sub-distributors who may or may not be digitised, and occasionally retailer EDI data from larger modern trade accounts. Building a secondary sales view means aggregating these partial sources — each with its own format, lag, and coverage gaps — into a picture that is good enough to inform demand sensing and distribution planning even if it is not complete.

The analytics architecture for secondary sales is typically a data lake that accepts whatever format each source provides (SFA app data, retailer CSV exports, sub-distributor Excel submissions), applies standardisation and quality checks in the processing layer, and produces a visibility report that shows coverage (what percentage of outlets are reporting), lag (how many days behind each source is), and the best available sell-through estimate by SKU, region, and channel.

The root cause of all three problems is the same: FMCG distributors are running their analytics on ERP data alone. The ERP is excellent at recording what the distributor did — sell-in volumes, invoice values, inventory movements. It cannot record what happened at the retail end of the supply chain, what the consumer actually bought, or how a promotion performed against a counterfactual baseline. The data needed to answer those questions is outside the ERP — in field force apps, retailer systems, secondary data providers, and the sales team's collective memory.

The analytics investment that pays off in FMCG is not a better dashboard on ERP data — it is the integration of external data sources with the ERP to build a demand intelligence layer that the planning team can actually use. That integration does not require a significant technology budget. It requires a data engineering approach that connects external sources to the ERP, cleans and governs the combined dataset, and serves it to the forecasting and planning tools that already exist in the business.