HOW TO IMPROVE FORECAST ACCURACY BY AGGREGATING LIKE ITEMS
Supply Chain Expert
Share this blog:
It’s a given that having clean, reliable data is one of the most important parts of forecasting. But, almost as important is knowing how to use it correctly. At Slimstock, we often talk about the importance of having individual stocking strategies for each of your markets. So, it’s understandable that this same thinking could carry over to forecasting, too. After all, if you’re customizing your assortment for each market, wouldn’t it make sense to forecast each one individually as well? In this case the answer is no.
Let’s say that you have 5 distinct markets that you sell the same brand of nails in, in a few different package sizes. Instead of forecasting each market separately, you’ll get a more accurate forecast by adding up – or aggregating – the total number of nails sold in all markets. This is true regardless of package size, too – if you’re forecasting nails it doesn’t matter if they’re boxes of 50 or 5,000.
WHY AGGREGATION HELPS YOUR FORECAST
Aggregation is beneficial because of the law of large numbers. This law states that the larger a sample size you have, the more representative it is of the total population. This translates into more accurate forecasts because it gives you more data to work with. More data is helpful because it:
Provides more points to analyze for patterns, trends and seasonality.
Smooths out the random variation that can disrupt models at an item level.
The more specific the segment of data you’re aggregating, the higher chance it has for error as there are usually fewer data available. This can be seen in the below graphic, where a specific item has less overall data available than an entire brand, and so is more likely to have an error.
The benefit aggregation is that it removes outliers from data, however this averaging function can also be a negative depending the segment forecasted. For more specific segments that offer fewer data points, the outliers that are smoothed out may be patterns that deserve attention. For this reason, it’s best to only use aggregation for specific segments like item, location and customer when they have steady demand and predicable seasonality.
For this reason it is important to understand that aggregation isn’t right for ever product in your assortment, but instead should be decided on a case by case basis.
Using aggregation in your forecasts does require you to disaggregate your order quantity back down to an item level once it has been calculated. This means figuring out the total number of 50 count boxes of nails needed versus 5,000 count. There are many different methods for doing this, and our team of expert consultants at Slimstock would be happy to find the best option that works for your business.
Ultimately, disaggregation requires the two qualities mentioned at the beginning of this article – well defined markets and understanding customer behavior. Only by having a detailed knowledge of who is buying your products and why are you able to parse your inventory correctly after disaggregation.
Slimstock helps companies optimize their inventory by taking a complete look at your inventory strategy and adjusting stock levels according to factors like location, market differentiation, seasonality, promotions and even weather. Talk to one of our helpful inventory consultants today and see how we can save you money by lowering excess tock while increasing service levels.
Speak to an expert about your inventory challenges