Forecasting Systems and Methods
Forecasting systems are broken up into three main types, based on the type of information that is going into them: Qualitative, Quantitative, and Reference Class Forecasting.
Quantitative forecasts are built using information that can be quantified. You might know from some objective report how many widgets you sold last quarter, or you might know how many hours were lost at your factory due to injury last month. This type of information would go toward building a quantitative forecast. On the other hand, you may not have any numbers to support your forecast. It's possible that it's built on the opinions of experts, or those of your sales staff. They are making subjective judgments based on experience about the quality of the factors that go into your forecast.
These judgments make up a qualitative forecast. For this reason, qualitative forecasts are often referred to as judgemental forecasts. Neither of these two methods is better or worse than the other, but there are times when one or the other will be more practical. The third type of forecasting systems, Reference Class Forecasting borrows a bit from both Qualitative and Quantitative forecasting.
A Reference Class Forecast is one in which the forecaster collects information on a group of prior similar situations and looks at their outcomes. That is the Quantitative, objective side of this method. After building this reference class, the forecaster must decide where his project lands compared to the reference class. This is the Qualitative, subjective side of this method. You will have to decide for yourself which methods work best for you in your business in different circumstances.
Simple Forecasting Method
The Naïve approach is the simplest and most cost-effective forecasting approach. It's called Naïve because it simply takes the most recent value and proposes that value for the next period. Did you sell 15 cars last week? Let’s forecast 15 cars for next week. It took 5 years to build that hotel? Let’s forecast that this next one will take 5 years as well. The Naïve approach can also be applied seasonally. If you know that last Easter, you sold 500 chocolate bunnies, then this Easter you forecast the sale of 500 chocolate bunnies. If your movie theater sold 50 bags of popcorn last Saturday between 5:00 pm and 6:00 pm then this Saturday from 5:00 pm to 6:00 pm you will forecast the sale of 50 bags of popcorn. There are many cases where the naïve forecast is really all you need, and any resources put toward a more detailed analysis are not necessary.
If you are looking at data with an upward or downward trend, you could use that information to compose a Naïve forecast as well. If last week your sales team sold 5 more widgets than the week before, you could forecast that this week your sales team will sell 5 more widgets than last week.
Use Naïve forecasts when increased accuracy gained by using another method does not justify the increased cost. Simply put, if the Naïve method gives you usable information that is reliable enough for your purposes, there's no reason to use a more detailed and resource consuming method.
Quantitative methods include objective inputs that are not biased or based on personal opinion; these methods use hard data that cannot be manipulated. Since they are highly immune to personal biases, including knowledgeable opinions, they may “contaminate” judgmental approaches.
There are two main categories of quantitative methods:
Time Series Models – Use specific patterns from your business’ past to project patterns for your business’ future. For example, knowing your sales figures for the past three months can be used to forecast sales figures for the next three months.
Associative Models – Make predictions about a specific situation based on a different, but related situation. For example, knowing the number of seats sold to a particular sporting event can be used to forecast the number of cheeseburgers that are likely to be sold by the vendors at the event.
Time Series Models
By monitoring a specific segment of your business at regular intervals, you develop a basic understanding of what your business will do in the future. Time series models require the forecaster to make decisions to identify the underlying behavior of the series. Plot the information gathered on a graph and have a look. You may notice one of the following patterns:
Jumps – These are short-term upward or downward movements in the data or a step change that indicates a new base level. There are infinite causes associated with jumps in your data. A thorough investigation may reveal the roots of the jump, and it may turn out to be something you can plan for in the future, or better yet, cause to happen again, or find a way to avoid, as the case may be.
Trend – This is a long-term upward or downward movement in the data. The longer the trend, the more reliable forecasts based on the information become.
Seasonality – These look like spikes on your graph. When the spikes recur like this, they can be caused by weather patterns or market conditions that affect consumer habits during particular seasons, months, days of the week, or in some businesses times of day (restaurants and bowling alleys for example).
Cycles – These look like waves on your graph. Cycles can be attributed to many external political, social, cultural, economic, or even agricultural circumstances, for example. It may be that sales of a particular product are up during the middle of the month, and lowest on each end because the general population has more discretionary income in the middle of the month when they are not focused on paying rent/mortgages/bills, etc.
Irregular Variations – These are unforeseen one-time events that could not be planned for, and most likely will not recur. They can be positives like a short-term spike in sales from a celebrity being seen using your product, or negatives like a loss of factory output due to a power outage. As much as possible, these should be discounted and ignored when you are graphing your data and looking for trends.
In your business forecasting, you'll encounter times when there's limited or no real data on which to base your forecasts. There may be occasions where data would be available, but the cost or the time required to collect it makes it unfeasible. There will be certain situations where personal judgment will be very important, such as analyzing what a competitor is likely to do, or when you are designing a new product.
Qualitative techniques include soft information in the forecasting process such as human factors, personal opinions and intuition. As these factors are difficult or impossible to quantify, the quantitative approach omits them completely. Data for the factors used in judgmental methods are obtained from a different type of source than quantitative models. Where do qualitative forecasts come from?
Opinions of executives – You have hired the best and brightest group of executives available. You trust their judgment. Put a team of them together to draw on their experience and knowledge to create a forecast. This approach is often used for long-term planning and designing major strategic maneuvers. Be aware though, there is always a possibility that a strong personality will sway the group, or that less responsibility will be taken for the forecast because they are not personally responsible for the outcomes of the forecast.
Opinions of experts – Sometimes executives are telling you what you or your shareholders want to hear, and not focusing on what might actually be the case. There may be experts in a particular subject, both internal and external to your organization, whose opinions would form the basis of a good forecast. Your sales staff for example are often the closest to your customers and may be the most informed source of quality information on how many of a particular product you are likely to sell.
Opinions of your customers – speaking of your customers, you may have the opportunity to survey them directly to get their opinions on a new product or service. Keep in mind that there's often a bit of a disconnect between what people say they will spend money on, and what they actually spend their money on. Your Marketing manual deals specifically with interpreting the results of your surveys, focus groups and other marketing efforts.
Your Forecasting System
Now that we’ve looked at what a forecast is, and why accurate forecasts are useful for your business, let’s have a look at the basic steps involved in putting an accurate forecast together.
Begin with the end in mind. This is a primary principle that applies almost anytime you plan anything. You will make a plan to get where you want to go, if you know where that is in the first place.
Step One - Write out the main goals of your forecast. What information is it built to supply? Who will use it and how? How much detail is needed? How many resources will you need to put toward the forecast? Once you have written this out, compare it with the considerations above and see how it stacks up.
Step Two - Choose your technique. Now that you know what it is you're trying to accomplish, you need to decide on a technique that will get your desired results. Write down the forecasting method that you believe will work best for your particular situation and the reasons why.
Step Three - Determine the factors you will use to build your reference class. What information will you need to build your forecast? Make a list of the information you will need and where you will get it.
Step Four - Plot out the time-frame of your forecast.
Indicate who will rely on the information from the forecast and when they will need it. Put in writing dates to start information gathering, finish information gathering, and assessing the information once you have it. Write out one entire cycle of your forecast from beginning to end. Make sure your forecast provides you with timely information that will be useful to all involved.
Step Five - Collect and analyze your information.
This could mean conducting surveys of customers or experts, obtaining sales data from your sales department, researching market trends, etc. Collecting information can be time consuming, but it's important to do it thoroughly. Note any clusters of data, and do your best to find out what caused the groupings. Note any particularly low or high cases and try to determine if there are any easily identifiable causes.
Step Six - Build your chart. Use your experience and judgment to place your project on the chart – it is advisable not to stray too far from the mean. If you are forecasting how long it will take to develop a new product, and every other team developing similar products that are in your reference class has taken two years, it would be inadvisable to predict that your product would be developed in 3 months, unless you have a very pronounced advantage.
Step Seven - Make your prediction. You are as prepared as you could be. Go ahead and make some predictions based on the information you have collected. Write down which departments are going to need access to the forecast and by when and distribute it accordingly.
Step Eight - Monitor your results. This is probably the most important step in this process. In fact, don’t bother with any of the previous steps if you are not going to follow through and monitor your results. Was your forecast accurate?
Step Nine - Assess the reliability of your prediction. On a scale of 1 - 10, how accurate do you believe your forecast is? If this is the first time you have made a forecast for this particular situation or project, you will have to use your intuition. If this is a regular forecast, you should have data that tells you how accurate you are with your predictions.
Step Ten - Adjust your forecast. The higher the level of accuracy you have in making your forecasts, the less you will need to adjust this number. If, on the other hand, you feel that your forecast is not very reliable, you may need to adjust your forecast quite a bit.
Step Eleven - Review your results. Was your forecast accurate? Why or why not? What can you do to make it more accurate? Did the information reach the interested parties in the appropriate time frame? Is there anything you can do to make the information more user friendly? Did you use the best technique for this forecast? Did the forecast cost you more to implement than it saved you? Write down the answers to these questions and any others you can think of that could impact your next forecast, and then start the process again from the beginning once you have applied the changes.