Forecasting is a thankless job. It’s a lot like being a referee or umpire in your favorite sport; the only time a game official is noticed is when they do something wrong! Similarly, a forecaster’s primary aim is too stay out of the “news”.
Make no mistake, forecasting is a very important function in any business. In the software business, your whole business plan could be riding on meeting the forecast to fund growth and product development. In a hardware business, it’s an more even critical issue–you have to worry about creating too much or too little inventory–either of which can create a huge problem for your business.
Forecasting is hard in the best of circumstances
It’s bad enough when you are trying to forecast an existing mature product in a mature industry segment. This is a difficult and complex task, using well known techniques such as smoothing, trending and seasonality to fine tune the next monthly or annual forecast.
Early in my career at Hewlett Packard I spend 4 months in a special assignment dedicated solely to improving forecast accuracy. The marketing department was engaged in an ongoing “discussion” (OK-argument!) with the manufacturing department over inventory levels. Not surprisingly, manufacturing wanted the inventory levels to be lean, while marketing favored a more robust number. This was because manufacturing was being graded on their costs and at that time “owned” the inventory; while marketing was graded on revenue–and low inventory levels often lead to missed sales opportunities.
I became a Lotus spreadsheet guru and we used everything we could find to try to improve our forecast accuracy. Keep in mind that these were high tech products (computer printers), but successful product lines with significant historical data available. Try as we might, the best we could ever do was to get within +/-25% of the eventual unit sales number.
New technology products are the worst case scenario for forecasters
The main message here is that forecasting any product in high tech industries is very difficult from an accuracy perspective. Forecasting accurately the performance of NEW PRODUCTS with no historical data in technology markets is nearly impossible, imo. When you add in brutal competition, a tight market research budget, vague notions of market size, an early stage on the user acceptance curve and often the reality of an unknown brand — forecasters of new technology products better be able to handle stress well to make sure they don’t end up in substance abuse clinics. But of course, even though it’s hard– it’s still VERY important. So what’s a forecaster to do?
There are two basic methodologies that I typically utilize when attempting to forecast sales for a brand new technology product:
Top down forecasting method
The first approach that I usually employ is the “top down” method. You might also call this the “macro” approach. This is an exercise of defining the size of your total addressable market via existing market research or by estimating the total number of potential users, as well as also estimating what a reasonable market share will be for your product — given the various attributes of your market position. To establish your share consider everything you can in your analysis: your marketing budget, brand strength, an unbiased view of how your product stacks up vs. the competition, etc. It may be helpful to put it all in a spreadsheet, and quantify the various important attributes of your company/product vs. your competition. Be careful about assigning too much precision to these numbers, however; remember that garbage in equals’ garbage out. But if you go through this exercise thoughtfully it can be very helpful in analyzing your relative market position. In this case, obtaining your top down forecast is then as easy as multiplying the market share % you think you can obtain times the market size that you came up via research.
Bottoms up forecasting method
After I’ve done the top down or Macro forecast, I like to also use a “bottoms up” or “micro” approach as a sanity check. To do this, you want to gather information on what you think you can sell by canvassing individual stakeholders in the sales area: direct field sales reps, Online/Web store, dealers, international distributors, etc. It’s helpful to gather info from any channel that will be a significant contributor to sales for this new product. Usually it’s impractical to do a complete survey of everyone that may be involved in the sales effort. What’s important is to obtain a representative sample that is both broad enough and deep enough that the data you gather has at least some statistical significance. At that point, you can “normalize” the data. For example, say you were able to gather data from a broad cross-section of sales points, totaling approximately 10% of the total sales infrastructure (10% of dealers, 10% of direct sales reps, etc). You would then multiply the total number of units/dollars you obtained from your sales entities times 10, to reach a bottoms up forecast totaling 100%.
Do you have forecast convergence?
The key to this exercise is to discover whether your two views of the market are close enough that they appear to be focusing on the same topic! If they do, you may be in pretty good shape with your forecast. If they are off by an order of magnitude–or two , it’s probably time to reconsider some of your assumptions.
So there’s my advice on how to approach the unenviable task of forecasting a brand new technology product. It’s a high risk, high return activity under the best of circumstances–and ideal conditions are seldom found in this activity in the technology space. But if you are able to construct both a top down and a bottoms up forecast and the two numbers at least fall in the same ballpark, you’re probably on the right track.
Give it a shot yourself next time you’re faced with a daunting new product forecast. Feel free to shoot me an email with your questions, or leave a comment to extend this discussion.
If you liked this post please share it with you colleagues using the “share” buttons below: