Choose a Sign-in Option
Forecasting with confidence by finding the drivers of variation through verifiable data analysis.
What it is: Forecasting with big and small data is a specialized version of regression and correlation found in this library at: Using Regression to Find Correlation. Anyone can forecast without data, but it is, at best, a guess without any substantiation. Completing regression and correlation analysis to understand the drivers of change can create a situation where forecasts can be made with confidence. Forecasts can be based on regression and correlation analysis, which indicates either a time series forecasting model or a non-time-dependence analysis.
What it does: By identifying which drivers can assist in predicting future demands or marketplace reactions, a business can more effectively budget, plan, and make business decisions. There will be less waste and inefficiency caused by planning for future events that do not happen or being caught by surprise for events and activities that happen unexpectedly. Anyone can make great business decisions if they know everything that is going to happen, but it is much more difficult when there are a variety of things that could happen and you do not know which ones are most likely. That is why data-based forecasts are so effective in improving business decision outcomes.
How is it used: Forecasts are used to plan budgets, manufacturing production, pricing movements, new capital spending, and many other business decisions that require information to make an informed decision.
Where: Almost every aspect of a business relies on some type of forecasting to plan future decisions. Some of these forecasting situations include the following:
Capital expenditures: Being able to forecast future demand can allow a company to plan manufacturing and supply chain capital expenditures. It can also forecast the need for people in all areas of the business, which could require expenditures for buildings, labs, and other people-related capital expenses.
Forecasting manufacturing production levels: Manufacturing must plan raw materials, people work hours, equipment usage, product mix, expansions, and other factors based on forecasted sales demand for its products.
Sales forecasts: Marketing must plan for sales forecasts of the products it will sell, the mix of products, its advertising budgets, and other factors that are driven by forecasted sales demand.
Predicted impact of advertising: Being able to forecast the impact of advertising on sales will allow a company to plan the advertising level to create the optimum level of sales over the next quarter and will best leverage their available resources.
Forecasting inventory levels: Inventory that is too large has a huge impact on working capital and finance charges, while too little inventory can cause sales loss and create mistrust with customers to the point that you could lose their business entirely.
Economic drivers: Understanding the impact econometric indicators have on your business can allow you to make proactive moves in the face of changing market and economic conditions.
Forecasting workforce demand and employee attrition will allow an HR organization to make annual hiring planning. When you hire off of college campuses, you need to plan 9–18 months in advance to manage your campus recruiting. Knowing the number of people you will need a year in advance is critical to effective workforce planning.
Many other applications too numerous to name.
Why: Without insight into what will be happening to the important drivers in your market and industry, you are flying blind in making important business decisions
Where it shouldn't be used: When you do not have quality data or the data has major holes or is too limited for regression analysis, you should not use it for forecasts.
Any restrictions: Don't overstate the confidence of the data. Don't make assumptions the data does not support.
Warnings: Make sure you can explain why a trend is happening. One business in Europe was growing for more than a decade at 12% a year. The company needed to build a new plant for future demand, but the plant was very expensive and had very high fixed costs, so if it was not run at near capacity the business would lose a lot of money. Even though the sales had grown at 12%, the underlying market drivers could only justify a 3% growth per year. So why was the growth at 12% for more than a decade? The in-depth market analysis learned that a change from hand-laid plaster to mechanically blown plaster onto the wall under construction required significantly more of the company's product as the change was made. Country after country had been adopting the new technology over the last decade or so, and the last of the countries were just finishing the transition. That means that the sales growth would fall back to 3% a year. If they had built the plant, they would have lost millions of dollars a year for more than a decade because they built it too soon. So even though the time series forecast said growth was at 12%, an analysis of the business said that the 12% would end and change to 3%. Statistics don't lie, but they can be misleading if you do not do your due diligence (McCarty, Roger. Corporate Director of Strategy, Dow Chemical Company).
Gathering data: Collect data for dependent and independent variables over the relevant time frames for your analysis.
If possible try to find steady-state time frames where there are no major disruptions to how the market is acting.
If there is a step function change, then analyze the time before and the time after the big change separately to see if the variable relationships are the same.
Try and identify all of the factors that might be contributing to change over time.
Analysis of data: Complete regression analysis and determine correlation factors for all dependent variables. (Using Regression to Find Correlation.)
Utilize linear regression and exponential curves to fit the historical data to time series when trying to forecast.
Project trend lines and curves into the future to forecast dependent variables.
Interpretation of results: Use common sense and market data to try and explain variations and relationships with dependent variables.
Can you explain why the variable changed, and does it make appropriate sense?
Are there changes in the system (interaction of independent and dependent variables) that will occur during the forecast period that need to be accounted for?
Have there been other analyses that have been completed in the past that you could compare your analysis to and determine if there are consistencies in the results? If not, can you explain the reason for the variations?
Presentation of results: Present the outcome for the forecast projections (normally graph form), the data used to make the projections, and the reasoning used to confirm the projections.
Sales-Forecast-TemplateOutput Representation and Recommendations:
Normally shown in tables and graphs of historical and forecast lines and or columns.
Forecasting Examples for Business and Economics Using the SAS System
This content is provided to you freely by Ensign College.
Access it online or download it at https://ensign.edtechbooks.org/projectbased_internships/forecasting_with_big_data.
