Forecasting the Future

People say that sales is all about personality. Get the right salesperson to connect with the right prospect and watch the magic happen. But sales managers know that numbers rule the sales game. To win the game, the sales manager and every executive up the line needs good, solid numbers. Today’s numbers reflect the accuracy of yesterday’s predictions. So the forecasting game is afoot – every quarter of every year. The more accurate the forecasts, the more effective the sales plan.

Of the many ways a manager can “blow the numbers,” here are just a few: The forecast of monthly sales for your territory turns out to be wildly unrealistic – way too high for achievement. Or the forecast was so low that nobody budgeted for the support your customers needed, so you missed some valuable business. Or the forecast failed to take certain industry growth into account so your company missed out on territory growth from a lack of field support.

Forecasting mistakes can affect the company at many levels. Field reps experience the cost of bad sales forecasting at the micro level. When forecasts are off by even narrow margins, at the national level these costs can be huge. The CEO gets upset when sales don’t meet forecasts because the Chief Financial Officer must borrow money at higher interest rates. Marketing execs have tense arguments, upward with their CEOs and downward with their own staff. Product managers may get the bounce.

Sometimes the costs of forecasting errors are big, hard dollars. In the age of just-in-time manufacturing, bad sales forecasts lead to wasted hours and materials on the shop floor. Where products have limited shelf lives, you might as well throw away the money along with the old stock. And field reps always feel the effects of clumsy forecasting. Even when company-level predictions are accurate, bad forecasts for sales districts can cause stress and compensation problems.

Given the desperate need for accurate sales forecasts, experts have devised systems that take the guesswork (or at least most of it) out of the process. Of course, no one can guarantee a plan against acts of God or the Federal Reserve. But those aside, Art Cook, a professional sales forecaster for major pharmaceutical firms, offers some tips on turning the art of forecasting into a science.

The Basics
Specialists in the home office usually initiate major sales forecasts, Cook says. Product managers and market researchers try to come up with objective predictions of short-term sales. Senior managers review these forecasts to see how well they match up with the company’s goals and may adjust them.

“The numbers are then rolled out to the field,” says Cook. The overall sales forecasts are allocated to regional and district offices. This step should not be merely mechanical. “There are a lot of analytical choices in doing the roll-out,” Cook emphasizes.

Short-term forecasts come in two basic flavors: projecting revenues from existing products with a track record and predicting the sales of new products.

“For existing products, you usually start with the historical trend,” Cook says. “There are a number of software packages that will extrapolate from the historical trend.” He thinks a simple Excel spreadsheet is just as good as the specialized forecasting software in most respects. “The other tools package the data better, with nicer graphs and more statistical parameters. But Excel’s routine for finding the trend is usually just as good as the fancy software.”

So do you just double-click on the Calculate icon? Not quite yet. “Historical trends work well for well-behaved markets that are not disturbed by other factors,” Cook says. Unfortunately, even for some very well-known products, there are other factors.

“If a market is not well behaved, you must change the trend forecast,” Cook says. Adjustments must be made for factors such as 1) your own or competitors’ price changes, 2) new product introductions, 3) changes in the size of the sales force, or 4) any major changes in advertising or marketing strategies.

“These factors are the bane of product forecasts because you move into a judgement area,” Cook says. “And you can’t make the judgements unless you are out in the marketplace.” Often a market research firm can help, but in any case somebody should be looking closely at the market.

Once you gather market information, you must still make judgements. These calls are usually made at several levels within a company. “First, the product manager will give the forecast a reality check,” Cook says. “Then there is more judgement by the head of product marketing. Then the vice president of marketing makes a judgement call.”

There is a lot of very good business experience in these senior managers’ heads. That helps to avoid the big forecasting mistakes possible when trend lines are simply extrapolated by a spreadsheet. But, inevitably, other considerations may enter the judgement calls as well.

“They can take the high road and say simply, ‘I don’t believe this number,'” Cook says. “But some senior managers want to create incentives for performance by raising the forecast. Or executives are worried about projected earnings per share, so they deliberately set a higher goal. There are a lot of possible drivers as the forecast moves up through the approval process.”

To a professional forecaster some of these considerations are misplaced. Compensation plans and special sales goals can set incentives for superior performance, but the forecast itself should be as accurate a prediction as you can make of how much revenue you will receive in the coming period.

The problems of judgement and tapping a wide range of in-house and outside experience are even more intense in the second type of forecast: new product sales.

These forecasts are usually derived initially from a market forecast of the product’s potential revenue. “You look at a profile of the product’s consumers or of the buyers in a business-to-business market,” Cook says. “For example, you may survey buyers to learn their product preferences and likely response to your new offering.”

Most experienced managers know the necessity of survey research, as well as its potential weaknesses. Once you have estimated the potential market, you must still make a sales forecast over a specific period of time, usually a very short period. That means predicting how fast the market potential will be reached.

Every marketing textbook discusses one standard tool – the market penetration curve. This is usually an S-shaped curve that shows the cumulative share of a new product’s ultimate market reached over time.

“Market penetration curves work well in some industries and horribly in others,” Cook says. “In many consumer product markets, these curves can give reasonable forecasts. But in other markets, for example pharmaceutical sales, these curves do not yield good results.”

Cook says this is one of the toughest problems in forecasting for many industrial markets. “You may have a good feel for the size of the ultimate market, but there is great uncertainty about the speed of the uptake. You can find the theoretical pace of market penetration. You can try to find a historical analogy in the sales of a previous product. But for every analogy that fits, you will find 10 exceptions that do not fit. Estimating the uptake pattern tends to be one of the most difficult things in forecasting.”

Meeting Deadlines
Most major corporations do their short-term forecasts at least quarterly and sometimes monthly. Once a forecasting process has been set up and everybody is comfortable with it, implementation should be very efficient. All the data assembly, extrapolation of trends, judgements, conferences, adjustments and roll-downs to field salespeople should happen very quickly. The organization needs these numbers by set deadlines to function well.

Cook says a good monthly forecast can be done in about one week. That includes review of the accuracy of past forecasts to improve future predictions. For each major product, three to four headquarters staffers must devote one week per month to coming up with the forecast. For key products or difficult forecasts, you may have to use more effort.

A quarterly forecast allows more time. Cook says final quarterly predictions often take three to four weeks. That means more man-hours per forecast, but about the same number of hours on an annual basis.

Much more time is required to set up or reengineer a forecasting process. Cook estimates this will take up to six months, even for a company that has been making forecasts regularly. “There are lots of people who have a stake in the process. The company is changing its whole way of doing the forecasts. So you have to get internal buy-in on the new process.”

Cook thinks you should also look at your competitors’ forecasting process to benchmark your own methods. In any case, reengineering requires looking at all the elements of your current process. “Sometimes you need new software, and sometimes you need new forecasting methods. Maybe you just need to change the trending routines in your software package. But you may have to change your business methods, for example, by changing the person who has to approve the forecasts. Or you may have a problem in allocating the national forecast to your regions.”

So reforms depend very much on locating the problems with the old forecasts accurately and getting everyone to agree on the new methods. Corporate sensitivities are often involved. “After that, it takes two to three months to implement the new process,” Cook estimates.

A frequent problem with many forecasts is allocating national goals to regions, districts and ultimately to the territories of field reps. “The company is making its numbers all right, so management is happy,” Cook summarizes. “But the balance may be off at the field level, and the sales force is pretty unhappy.”

The problem usually occurs simply because headquarters staff, who originate the company forecast, do not devote enough time and attention to the allocation or roll-down to the field. One solution is to develop forecasts “from the bottom up.” That is, you ask district offices to make their own sales forecasts and then sum them up at headquarters.

This approach has advantages and disadvantages. The advantage? District and territory forecasts are often more accurate. The disadvantage? It takes a lot more time by more people to generate forecasts from the bottom up.

One compromise solution is to generate bottom-up forecasts for only three to four key products, where accuracy is worth extra effort. The national staff generates forecasts for all the other products in the traditional ways: by trend, judgement and research on new product markets.

Another solution for unreliable regional numbers is simply to do a more thorough job of analysis at headquarters. Cook says there are often recurrent patterns: particular regions consistently beat or fall short of forecasts for all or certain product lines.

The pattern may be a problem with the sales force or sales management. On the other hand, the forecast method may be ignoring differences in regional markets. Whatever the reason, looking for these patterns should be part of the forecasters’ job. “You must mine the data by territory and region,” Cook says. “You look for different propensities to buy in each region.”

Accuracy
A lot of business goals depend on making forecasts as accurate as possible. Nevertheless, even the best forecasting process will frequently miss the mark. “If you ask top management what kind of accuracy they want, they will want it down to a penny per share of earnings,” Cook says. “In effect, that means 97 to 98 percent accuracy for the overall sales forecast.”

That is usually way beyond the accuracy possible for individual product forecasts, at least on a monthly basis. “Most of them end up with 60 to 65 percent accuracy at the product level,” Cook says. “And that is in well-behaved markets, without a lot of dynamics like advertising and new products.” The accuracy rate of product forecasts by region is usually worse. And Cook says that long-term forecasts are even less reliable.

Long term
There is another type of sales forecast. Usually once a year, a major company makes a strategic forecast of its business for the next seven to 10 years. The strategic effort is timed not to conflict with the regular short-term forecast, since some of the same people will be involved.

The long-term forecast will affect investment, new plant location – the entire shape and orientation of the company. Since it is not really a sales-planning tool, the strategic forecast is usually made only at the national level. It is not rolled down to regions and districts – these subunits may be entirely altered over the forecast period anyway.

The challenges in doing long-term forecasts are much greater than in making short-term predictions. “It takes a lot more judgement, because the market can change dramatically,” Cook says. Even when changes are expected, the timing of change can be very uncertain. For instance, in pharmaceuticals managers often know what new drugs their competitors are working on. But the pace of the federal government’s approval process for new drugs is extremely uncertain. Many industries have a similar uncertainty about the pace at which new technology will actually enter the market.