Data Analytics in Singapore

Food & Beverage Manufacturing

A Singapore-based F&B company manufacturing sauces, seasonings, and ready-to-eat products sold across 6 ASEAN countries from 3 factories (Singapore, Malaysia, Indonesia) had no unified supply chain visibility. Each factory ran a separate MES (Manufacturing Execution System), each country's sales were tracked in separate ERP instances (SAP in Singapore, Oracle in Malaysia, a local system in Indonesia), and logistics between factories, distribution centers, and retailers was coordinated through spreadsheets and WhatsApp. The problems: (1) Demand-supply mismatch: the sales team in Thailand would request 50,000 units of a popular sauce, the Singapore factory would schedule production, but by the time the product shipped (3-4 weeks), Thai retailers had already sourced from a local competitor. Conversely, the Indonesia factory would overproduce based on optimistic forecasts, creating SGD 200,000+ in excess inventory quarterly. (2) No cost transparency: the true cost of delivering a product to a Thai supermarket shelf included: raw material costs (sourced from 5 countries), manufacturing costs (different per factory), logistics costs (factory to DC to retailer — spanning multiple countries), import duties, and distributor margins. Nobody could calculate the landed cost accurately — pricing decisions were based on estimated costs that were 15-30% different from reality. (3) Quality traceability: when a customer complaint arose about a batch of product, tracing the batch from the retail shelf back through distribution, manufacturing, and raw material sourcing took 3-5 days of manual investigation — unacceptable for food safety and potentially catastrophic if a recall was needed.

We built a supply chain analytics platform over 12 weeks: (1) Unified data pipeline: automated extraction from 3 factory MES systems, 3 country ERP instances, 2 warehouse management systems, and the logistics provider's tracking system. Data normalized into a single supply chain model: consistent product codes (each factory used different SKU numbering — mapped to unified product master), standardized units (one factory tracked in kilograms, another in pieces, another in cases — all converted to a base unit for analytics), and common time zones (all timestamps converted to Singapore time for cross-factory comparison, with local time available for factory-level analysis). (2) Demand sensing dashboard: real-time sell-through data from key retailers (POS data feeds from major supermarket chains in Singapore, Malaysia, and Thailand — negotiated as part of the supply agreement). The dashboard shows: current week sell-through vs. forecast (by product, by retailer, by country), inventory at each point in the supply chain (factory → DC → retailer — total pipeline inventory per product), and demand trend alerts (when actual sell-through deviates from forecast by more than 20%, the production planning team is alerted — enabling production adjustment before the mismatch becomes costly). (3) Landed cost calculator: automated calculation of the fully-loaded cost of delivering each product to each market. The model includes: raw material cost (weighted average across suppliers, updated with each purchase order), manufacturing cost (allocated based on production time, energy, labor, and overhead per factory), logistics cost (actual freight rates from the logistics provider API, not estimated rates), import duties and taxes (per-product, per-country — configured from customs classification), and distributor margin. The result: accurate landed cost per product per market — updated daily. Product managers see actual margins by market, enabling: pricing optimization (products priced below target margin are flagged), sourcing decisions (which factory should produce which product for which market — based on total landed cost, not just manufacturing cost), and discontinuation decisions (products that are margin-negative after full cost allocation are identified for review). (4) Quality traceability: every production batch tagged with: raw material lot numbers (linked to supplier delivery records), production parameters (temperature, time, line speed — from MES), quality test results (lab analysis at production and at DC), and distribution path (which DC, which logistics shipment, which retailer). Batch traceability query: enter a product barcode or batch number, and the system shows the complete history from raw material receipt to retail shelf — in under 30 seconds (vs. 3-5 days previously). (5) Predictive demand forecasting: ML model trained on 3 years of sell-through data, incorporating: seasonal patterns per country (Ramadan affects demand differently in Malaysia vs. Singapore vs. Indonesia), promotional lift effects (quantifying the incremental demand from specific promotion types), and leading indicators (weather data — sauce consumption correlates with cooking frequency which varies with weather; economic indicators — GDP growth rates correlate with premium product demand). The model produces 8-week forward demand forecasts by product and country, with confidence intervals.

Demand-supply alignment: overproduction waste reduced by 32% in the first two quarters. The demand sensing dashboard enabled production adjustments 2-3 weeks earlier than the previous monthly planning cycle. Annual waste reduction value: approximately SGD 640,000. Landed cost accuracy: the company discovered that 4 products were margin-negative in the Thai market when fully-loaded costs were calculated (they had appeared profitable under the estimated cost model). Two were repriced (restoring margin), two were discontinued in Thailand (reducing complexity and freeing production capacity). Net margin improvement: 2.3 percentage points on the Thai portfolio. Quality traceability: batch tracing time reduced from 3-5 days to 30 seconds. During the first 6 months, 2 quality incidents occurred — both were traced to root cause within hours (vs. days previously), and the scope of affected product was identified precisely (20 cartons vs. a precautionary recall of 2,000 cartons under the old system). Estimated recall cost avoidance: SGD 180,000. Inventory optimization: pipeline inventory visibility reduced total inventory days from 45 to 34 — freeing SGD 1.8 million in working capital. The company reinvested the freed capital into expanding distribution in Vietnam (a new market entry that the previous cash-constrained operation could not fund). Development cost: SGD 210,000. First-year quantifiable value: SGD 640K (waste) + SGD 380K (margin improvement) + SGD 180K (recall avoidance) + SGD 1.8M (working capital) = approximately SGD 3.0 million. Platform ROI: 14x.