Comment choisir la meilleure méthode de prévision de la demande ?

Qualitative methods: Based on judgments, opinions, intuition, emotions, or personal experiences and are subjective in nature. They do not rely on any rigorous mathematical computations. QUALITATIVE FORECASTING METHODS Executive Opinion: Approach in which a group of managers meet and collectively develop a forecast. Market Survey: Approach that uses interviews and surveys to judge preferences of customer and to assess demand. Sales Force Composite: Approach in which each salesperson estimates sales in his or her region. Delphi Method: Approach in which consensus agreement is reached among a group of experts.

Demand forecasting can be retuned by analysing data from two sources _ one is as reverse order requirement from customer (MTO - Make to order) need to capture for a certain period & second from historical demand/ behaviour of demand / cyclic based demand need to capture. After comparison of this two sources of data eventually gives very close forecasting.

One note to add, is that it's very easy to take qualitative perspectives and quantify them, but it does not make a qualitative sample into a quantitative measure. As an example, customer sentiment vs. customer engagement, are two measures that are qualitative vs. quantitative. For example surveys and their subjective nature will have scores attached to them. Those scores while data points, allow you to apply math to it, and subject to simple math, but it's not a substitute for real quantitative measures.

Quantitative methods: Based on mathematical (quantitative) models, and are objective in nature. They rely heavily on mathematical computations. QUANTITATIVE FORECASTING METHODS * Time Series Models: • Assumes information needed to generate a forecast is contained in a time series of data. • Assumes the future will follow same patterns as the past. * Causal Models or Associative Models: • Explores cause-and-effect relationships. • Uses leading indicators to predict the future. • Housing starts and appliance sales.

For Capacity Planning,the quantitative methods would be the most suitable since it's about what are the Capacity requirements which shall be planned against the upcoming orders or the forecasts and information should be quite precise

The large disadvantage of quantifying any infrastructure needed for customer use is that there are typically multiple growth and optimization paths for a specific infrastructure. When we have a lot of data, we are typically over burdened by the sheer amount. Taking time to qualify the measures that will be in use for the process, goes a long way in forecasting for the future. Add baselines and record where you were and where you want to be and measure often enough but not too much. Accuracy in the data is less important than precision. Precision in metrics will allow you to later account for some accuracy problems in formulas, however, if the data is inconsistent it's unreliable.

TIME SERIES MODELS * Forecaster looks for data patterns as Data = historic pattern + random variation. * Historic pattern to be forecasted: • Level (long-term average) – data fluctuates around a constant mean • Trend – data exhibits an increasing or decreasing pattern. • Seasonality – any pattern that regularly repeats itself and is of a constant length. • Cycle – patterns created by economic fluctuations * Random Variation cannot be predicted.

In most cases, a wonderful way to test methods is to use multiple methods and see where they lead based on the numbers. Trim those down to ones that work for your business. Also, having multiple people engaged separately in building models allows for variations when millions of dollars are at stake. Demand models are essentially predicting the future and anytime you can use more ways to guess a trajectory of your needs the better the guess. Time is always the largest factor.

Capacity planning is usually done with peak usage/transaction load. For predicting peak load, usage pattern should be considered. Peak usage can be typically defined as percentage of total users using the resource. Also important thing to consider is that during peak usage there maybe acceptable degradation in performance.

To add to peak usage. Peak usage is a subjective measure based on persona, but from an infrastructure analyst perspective. Peak usage in my mind is the difference between your peak and your bottom usage in a given period, usually a day. So for example, it's not enough to say 30% of our subscribers use the site on a given day at peak. While we might see an active DAU at 30% at peak, the product experiences a load that goes from 1200 transactions per second at peak to 100 when off peak and the Avg load is 300 tps. With a few more variables, you can match the needs of your infrastructure and downstream dependencies, and with more reliability.