What level of forecasting is appropriate for your organization? The more trust the organization has in the forecasting function the more complex methods and tools can be leveraged. The less trust, the simpler the numbers need to be. While forecasting has existed in various forms for some time, in the regards to it as an established industry it is on the newer side in terms of business function. Oliver Wight and Gartner, both leaders in the education space were only founded in the late 70’s and technology has had a profound impact on how the function is completed. Microsoft Excel was first released in 1985 and probably not widely adopted, especially in smaller organizations until the late 90’s. The reason I call out Excel specifically is because it is ostensibly the first available software that could be used without a significant investment. Additionally, it still plays a profound role in ad hoc reporting and scenario planning in even the most sophisticated organization. The point being, there is still a lot of room for adoption of forecasting as a discipline, particularly at organizations that can’t afford to invest in a fully integrated planning and/or ERP solution and need to build something from the ground up. When the loudest voices in the space are primarily from SAAS companies, it can be a muddy situation to navigate. To be clear, there are several powerful software solutions in the market, but an organization will be much better served understanding what processes they are hoping to make more efficient before jumping into that commitment.

I think the most illustrative example will be to envision the least developed method and build from there, mostly because starting from the bottom leaves the most room for improvement and additional improvements are basically going to be variations built on these concepts.

What we can envision is an organization that is planning and not necessarily forecasting. Typically, this is going to be a smaller organization of less than 500 employees. In this case, for the most part purchase decisions are being made with a lagging indicator rolling average to determine what the future will look like. Maybe some revenue forecasting is done as a function of needing to understand financials. Probably Finance driven with minimal input from Sales with no connection to an item specific level operationally.

Your supply chain would need to be awfully flexible both from a delivery to customer and delivery from suppliers for this approach to not have a significant impact on profitability. I suppose if your sales had a very minimal variance from period to period, that might also be acceptable. I’m going to guess that is a very small subset of businesses. Most are going to face variable sales, seasonality and supply chain disruption. Further, our consumer mindset has likely increased customer expectations for faster delivery and on time performance.  

The best analogy I have heard as to why you want to move away from only looking at a historical average is that it is like driving while only looking in the rearview mirror. If operational and supply chain numbers are only looking at items level moving averages, they are going to miss key future customer events that they could otherwise prepare for. This means, spending cash on inventory that won’t be needed over the next cycle versus spending it on inventory that may have see a spike in the next cycle. One product being over inventoried is not the end of the world, that is going to happen no matter how good you are as an organization, but on the conservative side, say you have 1000 products and 30% are over forecasted at any given time. Even if they are all low-cost items…and they won’t all be…that can add up to a significant amount of money being wrapped up in inventory that you can’t move through. Meanwhile having not anticipated the spike in another product means at best delaying delivery or paying expedite cost to your suppliers, at worst missing out on a sale.

So, what do you do? 

I would start with 2 actions in this scenario. Additionally, I am assuming that your supply chain purchases are connected with the amount of inventory you have on hand. If not, that is a whole other subject that we will likely address at some point.

To keep it simple, still work with a moving average as an output for your forecast. The first change will be that you should assign some resources to making sure your history is “clean”. Don’t overcomplicate this. Figure out the easiest and fastest way to identify outliers and then peel those out and rerun the average. Maybe only focus on your most important items first. A note on this, don’t get too cute with identifying “important item”, you can, and I have spun out on trying to create an ABC hierarchy. Either go with the most expensive items or the items where you know you have seen the most trouble. Sometimes it could be focused on specific product lines or for a specific customer. Domain knowledge of the organization is probably the most important aspect of this over a math formula. This alone will correct most of your unforced error inventory challenges. It will also help with some under inventoried situations and just generally avoid time consuming mitigation actions.

The second thing will be to get a rough feel for what sales might look like over the next few periods. Start with your new rolling average and then try to get a rough feel for where the rolling average might need to be increased. If Finance is in fact already talking with the Sales department to come up with targets, it may be worth starting there. Finance in general is a very organized operation and will likely have a version that is more easily adaptable for a forecast that can disperse through Supply Chain and Operations. Sales will have more intimate knowledge, no doubt, but they may not have the rolled-up perspective that might be required. Additionally, it is usually better for an organization to keep those who are most skilled at bringing in new business focused on that task rather than communicating the same information multiple times throughout the organization. This should help address several of the oversells, but certainly not all of them. That is challenging for innumerable reasons, but it should help reduce the pain.

If your organization doesn’t have existing software or if the software is burdensome to use, do it through Excel or Google sheets. Depending on the size of the organization, there is a really good chance it isn’t even a full-time position. The process is the most important part. As the organization gets familiar with these processes, documentation and then improvement will be the next logical step, but there doesn’t need to be a race. Unless there is a really specific and well defined problem that is clearly solved by more accurate forecasting, just keep moving forward one bite at a time. 

If your organization is farther along on this path, the fundamentals still apply. Focus on clean history and improving understanding of forward looking events.

Thanks for reading.

Until next time,

Blake Andrea