When you realize your organization has a forecasting problem, what do you do to solve it? In particular, if you realize you need new forecasting software, how do you begin to find it?
All too often, the first step in a software selection process is the Request for Proposal (RFP) (or its cousin, the Request for Information (RFI)). But in a May 1 post on her Supply Chain Shaman website, Lora Cecere rightfully questions the value of the RFP for forecasting software selection.
In "Forecasting: Let's Start by Asking the Right Questions," Lora makes the bold (but I believe, correct) assertion that "RFPs and PowerPoint presentations are the worst way to buy decision support technologies like supply chain planning." In her experience, "most companies ask the wrong questions and consequently have the wrong discussions." So what are the right questions?
Lora argues that the software selection journey begins by understanding the problem, and this is a key point. "Bad forecasting" is a good excuse for all kinds of organizational performance ills. But how do you know that forecasting is the true cause of your business problems? Or even if bad forecasting is the cause, how do you know that the forecasting problem can be solved? Instead of a scatter-shot RFP that wastes everyone's time, Lora suggests you conduct a data-driven discovery.
The Questions to Ask
Lora's list of nine questions covers the fundamentals that management needs to understand about their business before finding an appropriate vendor. A couple of examples:
1. How forecastable are your products?
Suppose your customer demand behavior is so erratic that 50% accuracy is the best you can expect to achieve. Then maybe your problem isn't so much bad forecasting (which can never be made better than 50% accurate), but how to manage an organization in the face of such demand uncertainty.
3. What is your Forecast Value Added by product segment?
Is your current forecasting process "adding value" by making the forecast more accurate compared to alternative forecasting methods (like using a naive forecast?). You may find, as many companies do, that a significant percentage of your forecasts are less accurate than a simple "no change" model. As mortifying as this may be to those responsible for the negative-value-adding forecasts, at least knowing this can be a blessing. Steve Morlidge of CatchBull provides excellence guidance on addressing this issue -- you can find links to articles and other materials on his website.
The point of the discovery exercise is better identify your organization's specific requirements, rather than just cut & pasting generic questions from RFP templates. This should allow you to have more intelligent discussion with vendor candidates, and better differentiate their capabilities and appropriateness for your situation. The typical RFP just demonstrates to the vendor how little the organization understands their true business problems, and how little thought they've put into to solving them.
The BFD on RFPs
In Part 2 of this series, I'll share my own commentary on RFPs from The Business Forecasting Deal (the book). I'm very much in line with Lora Cecere's sentiments on this matter.
