Comment pouvez-vous utiliser les données pour prévoir la demande?

Historical data at best can serve as the starting point. However, it is important to continuously monitor the trends, other important events and occurrence that could lead to a surge or fall in demand. I would prefer a quick look at the data from the PEST model perspective and consider Political, Economic, Social or Technological factors that could influence the demand.

Historic data only gives you a trend and depicts where planning can be improved. There are other avenues as well that one access to gain information such as market analysis, customer analysis, micro / macro indicators impacting purchasing power or decisions. Historical data can be more helpful in cases where business is in the CASH COW stage (having constant value/volume) where business life cycle is in maturity stage, you have captured the market with your full potential and recurrence is highly predictable. For the rest, relying on historical data can end up in bullwhip.

Historical data might lead to movements across time, especially as you are able to see iterations of the same time structure. Trends of differing amounts of time along with the larger movements of the historical data will enable a better view of where your demand is going. Historical might be yearly, trends might be 3 and 1 month moves against the preceding same time structure. Best guess against all inputs will give best possible forecast.

Historical data is the best way to predict demand, however to forecast demand one has to look at the current r trends in the market and keep in the prevailing situation.

Historical data is what it is - history. It helps us based on data of statistical significance to make better decisions, but it can't consider internal facts. If a companies growth is simply off (e.g. from 6 Mio € Revenue to 80 Mio € Revenue in one year), then any historical data in the world would not be able to predict this - not even AI. Using internal data like financial plans helps us to combine finances and inventory (as it should be) to make better predictions. We also have to consider marketing events, a change in seasonality and other facts that may lead us to wrong assumptions. Therefore, historical data has to be used, but it also has to be understood. We need more than history to look into the future.

One of the most important ways for accuracy forecast demand is through analyzing the data from the history of the product sales, analyzing the competitors’ similar products, analyzing your market, studying your potential dealers and consumer behavior, and your inputs at the end.

For any Forecast to be of value, Historical data set is of paramount importance. Its also important that historical data is correct and analyzed Correctly. Data driven forecast is mostly close to accurate. Trend Patterns can then be postulated.

Data can play a crucial role in forecasting demand for products or services. By analyzing historical sales data, market trends, customer preferences, and other relevant factors, businesses can gain insights into future demand patterns. Data from market research, customer surveys, and competitor analysis can provide valuable insights into consumer behavior and preferences. This data can help businesses anticipate shifts in demand based on market trends and changing customer needs. Various external factors, such as economic indicators, population demographics, and government policies, can significantly impact demand. By incorporating relevant data on these factors, businesses can enhance the accuracy of their demand forecasts.

There are some factors that we need to pay attention regarding the demand forecast depending on past sales data. Unless the data is updated periodically, system fails to see the new information processed by new data, that is, the change in trend, level and seasonality. Moreover, while adding the sales data of each new period, the oldest ones should be removed. Otherwise the past casts a dark shadow over the forecast scheme, reducing the effects of short-term. So, what should be the parameters of those values? If we speak with numbers; • Past period data must be at least twice the period that we will be predicting. • Past period data must be updated in a time frame shorter than the leadtime of the product either purchased or produced.

Organizations can not ignore customer feedback. Organizations exist for the customers. If you Don't have customers then you are not in operation. Your user department or end users may hold the key to your edge over the competition through their feedback.

How can you demand without having data? About twenty years earlier, we used to forecast by checking our data in the registers and order form books maintained by order bookers. We have come a long way. Now we use different softwares, apps and other mediums collect data. The more detailed data we have, the better we can forecast. If our best seller article in a range of 24 colors is a beige color trouser in Oct'22, then we have to check before forecast if the same color, size, pattern and other details are still in trend or not. Which were the shops where it's been sold?

Focus on those customer relationships! The collaboration that can exist is worth gold … my biggest contributions have come from information that has been shared because of a good relationship. Feedback is critical.

Customer Feedback is an important piece in planning your inventory, especially when talking about new product launches. How to know how they're going to sell? Some surveys, even not statistically significant, can give us an idea what the situation might look like. And in the end it's also a business risk to make decisions without being able to see the future; it's part of the game.

Keep sharing your view with your customers - they will know best what is going on in their business and what is coming. Put the time in and you will be amazed at what you learn and can take back to your own business!

It is crucial to understand the phase that the business is going through whether growth or maturity. We need to understand our business as well as the market. A business in growth phase may forecast on higher side being optimistic while a mature business will aim for higher accuracy taking a more pragmatic approach. Forecast accuracy improves with more historical data but needs corrections or adjustments when current internal and external influencing factors are under observation

Artificial Neural Networks do not always come with outstanding results when used alone. Along with the past sales data, user should also have a big data set of factors affecting it. This is not practical in all cases; but it is possible only with a significant workforce and budget dedicated. The best for an average company is the combination of machine learning and classical mathematical models. ETS, ARIMA or dynamic regression models provide excellent results when there is quite little data given, perhaps only sales quantities. If the company keeps records of promotions and campaigns, or has information even non-comprehensive and about the related data, then it will be good to include the neural networks in the game.

Machine learning is a great way to challenge your own business's assumptions. If you have history, your business assumption and a forecasting algorithm as a view - you are well set up to have a demand forecast discussion.

The most critical trap that the data examine team falls into is making predictions with a straight linear line in their mind. That approach has nothing to say except increase/decrease. So, we end up with a long social chat without numbers. Instead, output of the demand forecasting models must be added to the data projected on the screen. It is also convenient to show the variance arising from the forecast in the visuals. So, the audience will not be enchanted by such a powerful software, will not be overconfident in their findings, and will become cautious.

The future of the demand forecasting models lies in here. The models will search for those factors in the vastness of internet, and match them with some numerical facts. For example, imagine we ask ourselves what affects our sales? Air temperature, ex-change rates, something else? So, it will be time to answer an unanswered question; why? Such as “Yes, we do have an estimation but why so ? Why do the sales increase?”... In the future, forecasting models will be capable of measuring on what factors the sales. This way, our sales data will be able to meet the factors affecting itself. Obviously, that will ensure our predictions to be more consistent and conscious.

Absolutely, analysing historical data with statistical methods or algorithms helps to forecast demand. Apart from this we also need to consider qualitative forecasting to have a clear view of demand trends, enabling businesses to make well informed decisions.

Data is a cornerstone in this era. It helps to build predictive models that anticipate future trends. It promote transparency and accountability which ultimately helps an individual and organisation to overcome crucial decisions.

Forecasting can be a daunting job. Try to focus on the 80/20 principle and automate as much of your data as you can. That way, you can focus on exceptions and the lines which really matter and are best suited for a business discussion.