9 Demand forecasting:

Demand pattern:

• Horizontal: fluctuation around a constant mean.

• Trend: consistent upward or downward movement overtime

• seasonal: regular periodic fluctuation w/in a yr.

• cyclical: long term economic cycle over multiple years

Forecasting process:

• Steps

  1. Update history: keep historical data up to date.

  2. practical initial forecast: suing suitable forecasting methods.

  3. consensus meeting: involving marketing, sales, operation and finance teams

  4. revise forecasts: incorporate feedback and refine

  5. finalize and communicate: review by the operating committee, then formal communication

Inventory-Related Costs Steps for Building and Using a forecasting model:

        1. decide what to forecast (product, time, intervals, unit/dollar)

            2. collect and analyze data to detect patterns

            3. select and test forecasting models

            4. generate forecasts

            5. monitor forecast accuracy regularly

1. 1. 1. Key forecasting decisions

         • What to foreccast:. productor individual products.

         • Forecasting frequecy: weekly, monthly, quarterly, yearly.

         • Inventory Position Unit of measurement: Units sold or dollar sales.

         • Software tools: (SAS, forecast pro forecast master.

         • Forecasting method: depend on time horizon, data availability, accuracy needs, budget and personal expertise.

         Physical Inventory Forecasting models

         • quantitive methods: (judgemental and subjective)

         • executive opinion: senior exec collective judgement

         • salesforce essential: based on salesperson insight

         • market research: surveys and interviews to predict demand.

         • Panel consensus: group meeting of diverse department representatives.

         • dephli method: anonymous, iterative expert survey process to reach consensus.

         • casual method:

           • linear regression: predicts demand based on related variable

         • forecasting specific patterns:

           • horizontal patterns: use naive, moving, exponential smoothing

           • trend patterns: use trend adjusted exponential smoothing or linear regression (time as an indep variable)

           • seasonal patterns: use multipl or additive seasonal methods

         • forecast accuracy measurement:

           • forecast error:

           • common accuracy metrics:

             • mean absolute deviation: average absolute forecast errors

         • mad or standard deviation:

           • forecast error will lie b/w - 3MAD and +3MAD app of the time

           • market research: surveys and interviews to predict demand.

           • panel consensus: group meeting of diverse department representatives.

           • dephli method: anonymous, iterative expert survey process to reach consensus

           • tracking signal: CFE/MAD

           • used to monitor and detect bias in forecasting