standardize predictions

Many businesses fail to set realistic cost and financial targets, resulting in significant losses and cash flow problems. Profit leakage isn't just the result of an economic downturn, it's also a reflection of companies not investing the time and energy necessary to carry out the vital sales and demand forecasting activities.

Making good predictions isn't as hard as it sounds. Proper application of proven forecasting techniques can significantly improve profits. Specifically, using statistical methods to forecast under the guidance of rules and formulating a more dynamic overall enterprise plan can remove some of the guesswork in operational decision-making and return the focus of forecasting to optimizing profitability instead of waiting for things to happen. Go to "fighting fire" again. According to Gartner Group, businesses that successfully implement a unified forecasting process can expect at least 10 percent revenue growth.

Many top managers are skeptical of the accuracy of forecasts due to unpredictable competitors and market conditions. But many large corporations such as General Electric and Walmart have found reliable forecasting methods and applied them enterprise-wide.

Out of the misunderstanding of predictions

The first step is to eliminate ineffective technology. Some companies use very complex models without properly taking into account the company's network of suppliers and distributors, making the results unreliable.

At the other extreme, many businesses rely too heavily on the input of salespeople and managers to generate results. Evidence shows that, no matter how experienced they are, the opinions of these people can lead to inaccurate results. Because they tend to:

confuse goals (hope) and forecasts (reality); think their own personal judgments are more reliable than statistical forecasts; forecast conclusions serve their function and distrust forecasts from other departments; overestimate marketing campaigns and other Effects of revenue management actions.

Don't rely on forecast conclusions from individual parts of the company. For example, functions such as finance, manufacturing, and sales may each produce forecasts independently, but none monitor changes in other departments' forecasts and revise their forecasts to reflect those changes.

Businesses with these problems must address several technical issues in order to create reliable financial forecasts. The required data is often distributed across multiple systems, such as finance, production, sales, and supply chain. In most businesses, getting all this data is a near impossible task. But if information technology is structured to integrate these disparate systems, it will be relatively easier for employees to extract data.

Many companies also need to go further and solve process problems, which requires implementing a systematic, cross-functional forecasting approach. Corporate politics and trust issues should also be considered.

Six Best Forecasting Practices

Some companies are already taking steps to improve their forecasting capabilities. Many businesses in manufacturing and retail recognize that inadvertent forecasting standards lead to high inventories and low profits.

To address this puzzling problem, several manufacturers and retailers have jointly developed a set of standards for planning and forecasting throughout the supply chain. Using Collaborative Planning, Forecasting and Replenishment (CPFR), trading partners agree on common business goals and measures, jointly develop sales and operational plans, and collaborate electronically to update sales forecasts and replenishment plans. The participating parties also use standard methods to collect opinions from suppliers and customers, and use standard data formats to generate sales forecast conclusions.

Companies using CPFR include Eastman Kodak, merchandise retailer JC Penney, Kimberly-Clark, Kmart, Nabisco and Walmart. According to a survey by manufacturing consultancy Industry Directions, companies that adopted the CPFR standard improved their forecast accuracy by about 20 percent. For example, Walmart's "Online Retail Chain" system provides weekly forecast data to 3,500 of the company's more than 5,000 suppliers. The data gives insight into Walmart's retail activity, helping suppliers improve their forecasts and deliver products to retailers at the right time. However, most companies have poor or inadequate investment in forecasting processes.

Based on the experience of companies in retail and other industries, a set of 6 best forecasting practices has emerged:

● Standardized input

● Standardized forecasting method

● Forecast frequency

●Limit bias

●Measure performance

●Employ collaborative sales and operations planning

While circumstances vary, businesses that follow these practices will improve their predictive capabilities. The reason for success is much more than just buying the most polished software. The benefits come from better data, people, processes and tools.

The first step is to standardize the information entered into the sales forecasting process. If your sales reps all follow the same rules for categorizing opportunities, predictive models can run on similar data criteria every time.

Standardization is the categorization of sales opportunities with uniform rules. First, you define the stages of the sales cycle. Then, define what kind of progress needs to be made to move further up to the next stage. Finally, the odds of contracting are determined according to standard rules. In general, the information entered should be based on facts rather than subjective opinions.

The most common forecasting method is time series models, which are based on statistical comparisons of current conditions with similar periods in the past. It allows people to observe current activities and predict future outcomes.

With a time series model, you can adjust forecast conclusions based on the accuracy or bias of previous forecasts, making current forecasts more reliable. For example, if you make a forecast 5 weeks before the start of a new quarter, you review the forecast made 5 weeks before the start of the previous quarter, compare the forecast with the actual result, and then adjust the new forecast accordingly.

Embracing data over intuition

Many businesses have applications that collect data points and run predictive models. Includes financial and budgeting modules in ERP and "Supply Chain and Demand Planning" software.

Time series models are best suited for short-term forecasting. If mid- to long-term (i.e. more than 6 months) forecasts are required, time series methods cannot reveal long-term trends that must be considered.

Most businesses generate forecast conclusions on a quarterly or monthly basis. This is often insufficient if the required level of precision and response speed is high. In some cases, companies using real-time enterprise forecasting models make forecasts on a weekly basis. The sooner you update your forecasts, the sooner you will know which areas need help, what adjustments need to be made at the production level, and how the budget should be adjusted.

Infrequent forecasts can lead to a significant disconnect between budget and actual results. Of course, you don't want to stop business and just make predictions. But generally speaking, the higher the forecast frequency, the more accurate the conclusions.

Despite all the tools and studies showing the opposite results, many business people still believe that their judgments are more accurate than well-established statistical techniques. These biases are difficult to predict because different people are involved in the forecasting process at different times. Incorporating personal bias into the forecasting process undermines the significance of comparing historical data.

A best practice is to limit individual subjective predictions. In general, trust data and statistical models rather than intuition. If you must add personal judgment, wait until the statistical model has produced a prediction, not before.

To continuously improve any initiative, performance measurement is critical, and the forecasting process is no exception. You need a systematic approach to reviewing forecast errors and continually improving your company's processes and models. The best way is to form task teams and review past forecasts. It is best to have a committee of senior managers whose mission is to improve weaknesses in data, processes, and execution.

Effectively implement forecasting processes

Once better forecasts are made using consistent inputs and standard models, a collaborative planning process begins operating across the company and across all suppliers. This process, called "Sales and Operations Planning" (S&OP), is usually conducted on a weekly basis and covers all major business areas. It combines baseline forecasts with supply chain constraints, resource availability, marketing-promotion information, and other relevant operational information.

The result is a complete supply and demand plan and financial plan. These programs are used to balance supply and demand, address areas of weak demand, and take advantage of opportunities. Make sure all areas of business are pursuing the same goals. Here are a few tips for implementing an effective forecasting process:

Benchmark your current operations. Compare your forecasting methodology with industry and other company best practices. Survey forecasters within your company and compare the results to industry surveys.

Educate forecasters and users. Since most business schools don't teach forecasting, many managers don't understand forecasting. Educate managers in methods, procedures and applications. For the potential for improvement, explain why it is worth the time to do it.

Form cross-functional task teams. Teams of mid-level forecasters and business managers survey the enterprise's benchmarks and then determine the desired approach and outcomes. Such discussions can evolve into standing task teams to implement the required enterprise-wide changes.

Assess data quality. Is your opportunity management process consistent across the company? Is information consistently captured in every area for every product? Are employees utilizing the system in a timely manner so that the data is of high quality? You are likely to find many inconsistencies, and the key is to identify and resolve those inconsistencies. Involve a control committee composed of representatives from different parts of the company.

Implement a communication process. A poorly designed process can lead to a rift between forecasters and users. This is a common pitfall in forecasting. To avoid this, be sure to carefully plan your own process for communicating your predictions. Make sure forecasters and users agree on the format and volume of data.

Redesign the company planning process. Accurate forecasting should be a key driver of a company's planning process, which requires most businesses to completely change the way they plan. Instead of just deriving plans from each area using different assumptions each quarter, plans should be more dynamic and adaptable in the short term based on changes in forecasts.

To achieve this, standards are introduced so that all areas of the business can align and synchronize their plans. Consider implementing a rolling plan that adapts to changing needs and enables efficient use of corporate resources.

All of these initiatives require you to change your business processes and get senior-level buy-in. You need comprehensive planning that takes into account change management, process, strategy and technology.

Original article from the March 2003 issue of Optimize magazine with permission. The magazine's publisher, CMP Media LLC, registers the copyright. Translated by Lian Qingsong.

Philip Bligh is Inforte's CEO, Darius Vaskelis is the company's vice president of research and development, and John Kelleher is the company's consultant.