How to improve seasonal forecast accuracy for better customer outcomes

Seasonality is a core consideration when you’re forecasting how much stock you’ll need to purchase to meet expected demand. For example, in the auto industry, the demand for batteries typically spikes in winter, while wiper blades see higher sales during rainy seasons. However, there are a number of seasonal considerations, not all related to weather, which can seriously impact the effectiveness of your forecast.

In this article, we’ll go over some key concepts:

Exploring the reliability and utility of weather forecasts in purchasing

Isolating promotional activities (e.g. Black Friday sales) from seasonal effects

Ensuring that seasonal purchasing plans take into account supplier lead time

Comparing year-over-year vs recency weighted (i.e. “momentum”) demand models

Exploring the reliability and utility of weather forecasts in purchasing

Weather is a significant factor in seasonal demand for many products. While any good forecast should account for sales changes during each season, we’ve seen some clients try to incorporate real time weather forecasts into their sales predictions. This can have utility in certain situations, but can also be more trouble than it’s worth.

Short-term accuracy

  • Weather forecasts are generally reliable for short-term planning (up to 10 days). For items with very short leadtimes, it may be possible to use near term weather forecasts to make adjustments to immediate inventory needs.

Long-term predictions

  • Long-term weather forecasts are less reliable. While patterns like El Niño can indicate general trends, they should be, at most, a small influence on your overall demand plans. The Farmer’s Almanac and other long term forecasts are notoriously unreliable.

Isolating promotional activities (e.g. black friday sales) from seasonal effects

Promotional activities can skew your seasonal forecasts if not properly accounted for. It’s crucial to isolate these effects to maintain forecast accuracy.

Historical adjustment

  • Adjust historical data to remove the impact of past promotions, so that your system doesn’t over-forecast coming out of a big promotion. At Hydrian, we have automated ways of doing this, but manual review of promotional sales periods is fine, too.

Future planning

  • Add expected promotional activities to your future forecasts to ensure adequate stock levels. This is the flipside of historic correction — if you don’t load in future increases in demand due to promotional activity, you will not have the inventory necessary to support the promotion. You might even be spending extra marketing dollars to send traffic to items that aren’t available.

Lead time cycles

  • For example, if your sales and marketing team doesn’t pass along estimates for promotional sales increases at least one supplier lead time cycle before the promotion starts, you will be unable to move extra inventory into your DC in time to support the promotion.

Ensuring that seasonal purchasing plans take into account supplier lead time

One critical aspect of seasonal forecasting is aligning your purchasing plans with supplier lead times. Seasonal demand spikes are predictable, but if you fail to account for the lead time required by your suppliers, you might find yourself with stockouts during peak demand periods. Order too much stock too early, however, and you’ll needlessly hold excess inventory.

Early seasonal orders

  • Place large seasonal buy orders roughly one lead time cycle ahead of expected demand, with a reasonable buffer.

Calculate lead time buffer

  • To estimate an adequate lead time buffer, you could take the 75th percentile of each supplier’s historic lead time and then add a month.

Consider holiday variances

  • Make sure to account for holidays and other exceptions that may change the lead time of a particular order.

Year-over-year vs recency-weighted (i.e. “momentum”) demand models

When forecasting seasonal items, it can be tempting to use the prior year’s peak demand in order to predict this year’s peak demand. For extreme seasonal items (e.g. ice scrapers, which may not sell at all during the off season) this can make good sense, especially before the high season begins. But for items with at least some sales during off-peak seasons, it’s important to incorporate that recent data into your forecast. 

In any event, revising your forecast during the peak season based on actual recent sales is critical. Using a seasonally normalized model, where recent demand is first adjusted for seasonal effects before passing into your forecasting model, is usually the best approach. 

Key strategies:

  • For extreme seasonal items, simply using last year’s sales as your baseline forecast can be a good starting point, especially before the start of the season.

  • For items with at least some off-peak sales, and for all items once the peak season begins, seasonally normalize historic data by dividing each past period’s sales by a “seasonal factor” (the expected sales during that period vs the annual per-period average). Run your chosen forecasting model on this data.

  • To “re-seasonalize” the data, multiply each forecast period’s demand in your output by that period’s seasonal factor.

Conclusion

Effective seasonal forecasting requires a nuanced approach that considers supplier lead times, the right mix of demand models, the utility of weather forecasts, and the isolation of promotional activities. By choosing the right forecast frequency and approach, you can significantly improve accuracy and ensure better customer outcomes. Implementing these strategies will help you stay ahead of demand fluctuations and maintain optimal inventory levels throughout the year.

You might also like

Navigating inventory management challenges for SMEs​

Running a small to medium-sized enterprise (SME) is no easy feat.

AutoZone’s push for more distribution centers: When does expanding your DC network have diminishing returns?

AutoZone has recently made headlines with its aggressive construction of new distribution centers (DCs) referred to as “mega hubs.”

Shelf life: How to avoid losing money to expired or obsolete parts

Managing shelf life and obsolescence is a significant challenge in the automotive parts industry.

Subscribe to our newsletter

Get updates on the latest news across all core inventory-related processes.

Subscribe now!

Subscribe

Your email is safe with us, we dont spam.

Want to see how your inventory management stacks up?

We’re so confident in our results, we offer a free performance assessment to all prospective clients. This isn’t a canned sales deck – it’s a bespoke presentation that takes 20 hours of our time. Whether we work together or not, we promise you’ll walk away with useful insights that will improve your business.