Lean thinking etched the concept of ‘PULL’ in the mind of supply chain professionals which in-turn popularized the concept of sales based Replenishment Planning. Don’t try to forecast sales. Hold inventory for the replenishment lead time and just replenish what was sold. Sales are what actual customer pull is.

Replenishment planning proposed to take out the stress of forecasting the demand and just refill what was actually sold. Replenishment based on visual triggers in the form of Kanban bins became the defacto standard for replenishment in the shop floor.

In absence of automated systems to detect the consumption pattern and do a system driven demand forecasting, doing it manually was the only option. Doing forecasting manually was as good as no planning as the planning accuracy never went above 50%. It therefore made little sense to forecast. Relying on a pure replenishment model to backfill sales was a better option.

In the last few decades, technology made huge progress and with better forecasting techniques, the forecasting accuracy improved. Industry also started experiencing huge volatility in demand. With shorter product life cycles, proliferation of new products and rampant promotions the volatility shot north.

With increased competition, the focus changed from a reactive approach to a proactive one. Since actual sales always lagged behind future projected sales, relying on this lag indicator was going to push the company to react rather than predict.  For products that had volatile sales and exhibited seasonal pattern, the inventory that was to be held for the replenishment lead-time was proportionately large. This is because, to avoid stock-outs, the maximum sales over a period of time was considered for calculating the inventory levels. This in-turn created the problem of huge inventory write-offs and broke the bank with mounting working capital cost. Any rationalization to this inventory holding quantity increased stock-outs and hence was a strict no no.

With better forecasting techniques that combined the power of Time Series, Machine Learning and Causal algorithms, a forecasting based fulfillment process made much more business sense. With newer techniques like demand sensing correcting the short term demand, the forecasting accuracy improved. This combined with statistically determined safety stock targets based on demand and supply variability ensured that the service levels are adhered to.

By predicting and not reacting, the right products were sent to the right locations at the right time in the right quantity. This ensured that the service levels are adhered to, inventory is rationalized, working capital is minimized and inventory write-off and stock-outs are eliminated.