One of the most basic keys to good decision-making is accurate forecasting of the future. In order to bring about the best outcomes, a company must correctly anticipate the most likely future states of the world. Yet despite its importance, companies not only routinely make basic forecasting mistakes, they also shoot themselves in the foot by applying procedures that make accurate predictions harder to achieve.

The future, to state the obvious, is uncertain. We may want to know precisely what the future will hold, but we realize that the best we can settle for is having some idea of the range of possible outcomes and how likely these outcomes are. Yet most companies seem to ignore this fact and ask employees to provide point predictions of what will happen—the exact price of a stock, the precise level of growth of a country’s GDP next year, or the estimated return, to the dollar, on an investment.

These precise, single-value estimates are poor decision aids. Suppose the forecaster’s best guess is that a project will be completed within a year, but the second most likely outcome is that, if the judge denies a zoning appeal, the project will take more than two years to complete. No single point prediction can provide the decision maker with the essential information about what to prepare for.

One way to counteract this problem is to ask for range forecasts, or confidence intervals. These ranges consist of two points, representing the reasonable “best case” and “worst case” scenarios. Range forecasts are more useful than point predictions. But they run the risk of, on one hand, being so wide, including everything from total catastrophe to glorious triumph, that they are not very informative. On the other hand, it happens even more often that the range is drawn too narrowly, missing the true value. Forecasters often struggle with this accuracy-informativeness tradeoff, and attempts to balance the two criteria typically result in overconfident forecasts. Research on these types of forecasts finds that 90% confidence intervals, which, by definition, should hit the mark 9 out of 10 times, tend to include the correct answer less than 50% of the time.

In our research, we looked for a forecasting approach that could provide both accuracy and informativeness: one that will protect the forecaster from the known traps of overconfidence and biased forecasting, and provide an informative forecast that includes all plausible future scenarios as well as an assessment of how likely each one of them is. We have developed a method called SPIES (Subjective Probability Interval EStimates), which computes a range forecasts from a series of probability estimates, rather than from two point predictions.

You can experiment with a version of SPIES aimed at forecasting temperature below:

( function() {

var func = function() {

var iframe = document.getElementById(‘wpcom-iframe-form-ed652c7b1d867b5c69a6ed4e56698af2-537a9bea80e88’);

if ( iframe ) { iframe.submit(); }

}

if (document.readyState === ‘complete’) { func.apply(); /* compat for infinite scroll */ }

else if ( document.addEventListener ) { document.addEventListener( ‘DOMContentLoaded’, func, false ); }

else if ( document.attachEvent ) { document.attachEvent( ‘onreadystatechange’, func ); }

} )();

The rationale for calculating range forecasts this way is based on the finding that, while people tend to be overconfident in forecasting confidence intervals, they are much more accurate in evaluating other people’s confidence intervals and estimating the likelihood that a particular forecast will be accurate. So the SPIES method divides the entire range into intervals, or bins, and asks the forecaster to consider all of these bins and estimate the likelihood that each one of them will include the true value. From these likelihood estimates, SPIES can estimate a range forecast of any confidence level the decision maker prefers.

The SPIES method provides a number of distinct advantages for forecasters and decision makers. First, it simply produces better range forecasts. Our studies consistently show that forecasts made using SPIES hit the correct answer more frequently than other forecasting methods. For example, in one study, participants used both confidence intervals and SPIES to estimate temperatures. While their 90% confidence intervals included the correct answer about 30% of the time, the hit-rate of intervals produced by the SPIES method was just shy of 74%. Another study included a quiz of the dates in which various historical events occurred. Participants who used 90% confidence intervals answered 54% of the questions correctly. The confidence intervals SPIES produced, however, resulted in accurate estimates 77% of the time. By making the forecaster consider the entire range of possibilities, SPIES minimizes the chance that certain values or scenarios will be overlooked. Second, this method gives the decision maker a sense of the full probability distribution, making it a rich, dynamic, planning tool. The decision maker now knows the best- and worst-case scenarios, but also how likely each scenario is, and the likelihood that the estimated value, be it production rate, costs or project completion time, will fall above or below an important threshold.

How can you use SPIES to your advantage? Consider a manager who must decide how many units to produce. A forecast made with SPIES estimates the odds of all possible scenarios and thus assists the manager in mitigating the different risks of over- or under-producing. Similarly, a contractor could use SPIES to forecast the likelihood of finishing current work in time to take on more work, as well as the likelihood of progress on current projects slowing down, making any additional work a strain on resources.

With so many strategic decisions for firms depending on predicting the future, forecasting accurately is enormously important. While traditional forecasting methods tend to produce poor results, we are happy to report real progress helping people make better forecasts. The SPIES method represents a big step forward, integrating insights from the latest research results on the psychology of forecasting. Managers who receive richer and more unbiased information make better decisions, and SPIES can provide it.

You can use the SPIES tool below to try forecasts in the context of your own business:

( function() {

var func = function() {

var iframe = document.getElementById(‘wpcom-iframe-form-ba2ce7e1539270ffcebd5d7d0fc657a6-537a9bea81195’);

if ( iframe ) { iframe.submit(); }

}

if (document.readyState === ‘complete’) { func.apply(); /* compat for infinite scroll */ }

else if ( document.addEventListener ) { document.addEventListener( ‘DOMContentLoaded’, func, false ); }

else if ( document.attachEvent ) { document.attachEvent( ‘onreadystatechange’, func ); }

} )();