Skip to Content
Business Trends
Author's profile photo Steve Christos

How to REALLY use predictive when planning


I’ve spent the majority of my career in FP&A roles and have seen more than enough presentations from vendors touting their predictive capabilities when it comes to forecasting.  While this should absolutely be a critical component of a modern planning process, I’ve often been perplexed at the use cases provided.  More often than not, the situation honed in on is predicting revenue for the upcoming year where a prediction is generated with a confidence level of +/- 20%.  Now, I don’t know about you, but if I told senior management that revenue will be $xxxxxx +/- 20%, I would be unceremoniously kicked out of the board room.  People lose their jobs when companies miss targets by a fraction of a percent, let alone a gap that big.  Thus, this is a poor example of where predictive capabilities can be beneficial in today’s economy.

Let’s discuss some areas whereas predictive forecasting really drives insight to action and minimizes the nonvalue tasks that eat up the bulk of our time:

  1. Create a baseline – Often when planning for different scenarios, it’s helpful to have a data driven baseline that extrapolates what could happen if current trends continue.  The gap between what you are planning and what the data trends reveal helps monetize the business drivers that need to be explained to stakeholders.
  2. Compare possible scenarios – Speaking of scenarios, a modern planning platform will allow for the easy creation and maintaining of multiple scenarios that contain different assumptions.  Normally, the analysis for these is done as of a point in time.  What if we freeze open headcounts?  What if we penetrate a new market?  Bring a new offering to market?  Predictive forecasting allows us to expand these assumptions beyond the fixed time period to inform which avenues to pursue.
  3. Plan for lower priority SG&A expenses – Possibly the quickest component of the planning process to streamline in order to save valuable time and reduce complexity are the non-critical operating expenses.  Think stationary & supplies, printing fees, etc.  These are the backoffice line items that are not driving the business forward but are a necessary component for cost control and guidance.  These are prime line items to allow predictive algorithms to forecast out in aggregate and cascade down to cost centers and business units.
  4. Forecast leading indicators – As I mentioned in my preamble, just running a forecasted scenario on something like revenue is often not a realistic exercise.  Another reason for this is that revenue is a lagging indicator.  It’s the result of other drivers and variables.  Rather, utilize predictive for leading indicators.  For example, a forecasted market response rate for a subscription service firm would help surmise the number of new subscribers.  This forecast in addition to an average subscriber revenue can formulate a more accurate prediction of revenue going forward.

Certainly, this is just a small subset of real-world predictive forecasting scenarios that you can leverage with SAP Analytics Cloud.  The ultimate goal is to get better insights, more efficiently.  Let me know others uses you may have!



Assigned Tags

      You must be Logged on to comment or reply to a post.
      Author's profile photo Huong Dao
      Huong Dao

      Hi Steve,

      Thanks for an insightful blog, which reveal the thought that FP&A experts really think about.

      For the first point, do you mean that we should keep a version of comparing between what has been predicted and what are the actual value? Keep it for a period of time and then review?

      A lot of my customer is asking how they can actually measure the performance of those forecast, love to hear your thought on this.



      Author's profile photo Steve Christos
      Steve Christos
      Blog Post Author

      Hi Huong,

      That's absolutely one way to look at it.  Let's look at another example.  Let's say your forecasting volumes sold for a particular product.  You've come up with a forecast of 10 million units.  When you run a predictive forecast based on the historical sales of the same product, it returns a result of 8 million units.  Now you have a 2 million dollar variance that will require an explanation (additional markets, sales promotion, etc)