3 Steps for Better Forecasting Before Peak Season

3 Steps for Better Forecasting Before Peak Season


This is a guest post from Joannes Vermorel, CEO of Lokad, a software company that specializes in quantitative optimization of supply chains. In this post, Joannes discusses three steps to get your business in shape for peak season. 

Amazon has dramatically raised the bar for businesses that deliver goods ordered online. As illustrated by the recent acquisition of Whole Foods, even the more traditional segments of retail are now threatened by the Amazon behemoth.

Thus, more than ever, ensuring a high quality of service is critical for an online merchant to stay in business. In practice, for many businesses there is one period of specific importance for many lifestyle goods: the peak season, which typically happens before summer or before Christmas. In order to stay in business, follow these three steps to better forecast before peak season. 


Properly sizing the inventory before peak season is obviously critical because peak season is precisely the period when customer demand will be at its highest point during the year.

However, it’s also a very challenging exercise with hard-to-discern hurdles.

First, as suppliers may be on the other side of the world, for example in China, the peak season needs to be forecast well in advance. Actually, it is even recommended to treat the supplier lead time, that is the delay between placing the order and the delivery to your warehouse, as a forecast.

Indeed, lead times have their own seasonality, e.g., they get longer around the Chinese New Year in China because factories are closed for about a month. In practice, forecasting the supplier lead times start with properly monitoring them.

This implies that the online merchant should:

  • pay extra attention to its purchasing process
  • the date of ordering should truly reflect the date when the original purchase order is placed
  • this date should not be altered afterward.


Also, the suppliers themselves might be under pressure just before the peak season. Thus, sometimes it’s not sufficient to take the supplier’s lead times into account:

One should also pay attention to the supplier’s reliability. Indeed, the products that are most in-demand are likely to be the first to run out-of-stock on the supplier’s side. Merchants who successfully identify those strong-demand/weak-supply products can position themselves as de facto exclusive distributors, just because they are the only merchants who happen to have inventory left. Analyzing supplier reliability, just like supplier lead times, start by making proper measurements.

For example: if your company is ordering 100 units, but the supplier manages to ship only 25 units, then this “loss” should be recorded; later on, it will become part of the foundation of your supplier’s reliability analysis.



Second, preparing for the peak season demand also means properly anticipating demand even for new products that have never been sold before.

Indeed, in many markets such as fashion or consumer electronics, it’s the norm that the best seller for the next season is a product that has never been sold before. Yet, most practitioners have already observed that the majority of the forecasting solutions work poorly in such situations. Most forecasting toolkits work through moving averages of some kind, which obviously cannot capture the “product launch” effect and cannot produce any kind of pattern associated with the product lifecycle. Facing such limitations, many practitioners have come to believe that forecasting demand for the top sellers of the peak season is forever beyond the grasp of statistics.

Nevertheless, forecasting the demand for new products, i.e,. products never sold before, is not science fiction.

It boils down to two ingredients: good data and a good forecasting engine.

Since the products of interest are new to the market, there is no point in saying that we want better sales data. What we need instead is better “product” data: any characteristic that could hint at why a particular product is likely to be next season’s top-seller. In practice, this can be information as mundane as the style, the color, the price-point,but more refined information, such as whether the product makes it to the front page of the supplier’s catalog – or website – can also dramatically help to identify the candidates most likely to rise to the top of next season’s best sellers.

Uncovering the top sellers demands using a forecasting engine that is capable of leveraging product-level attributes in order to identify future best sellers based on the history of past top-sellers and their associated attributes. Lokad happens to feature a forecasting engine with this ability: width-oriented rather than depth-oriented; instead of expecting a long demand history for each product, the forecasting engine relies on many products, from which it draws similarities, which are then leveraged for next season’s forecast.

For smaller merchants who don’t have the resources to fully categorize their product catalog, another nice feature to have in a forecasting engine is a text-mining ability. It allows to leverage raw plain-text descriptions of products to compensate for that lack.


Third, readiness for the peak season should go beyond the pure demand forecast and take into account the economic drivers of your business.

For example, when considering two products with the same expected future demand, the final purchase order decision can be very different if one product is expected to be long-lived, while the other is perishable; the respective demand levels might be equal but the respective inventory risks aren’t.


The economic drivers represent all the forces that generate opportunities or costs for your business. Beyond carrying costs, another obvious economic driver is the profitability of the product. Again, when considering two products with the same expected future demand, if one of the product is vastly more profitable than the other one, then it makes more sense to purchase more of the profitable product.

Once more, optimizing the decisions for the peak season takes preparation. In particular, it means assessing the costs and opportunities associated with the the various products and keeping all this data organized so that the the final pre-peak season purchase orders will be made in a timely fashion. This also involves leveraging the latest sales data, which contributes to making your forecast more accurate.

All in all, forecasting the peak season is mostly a matter of (data) preparation and proper statistical tooling. The goal is not to eliminate ordering mistakes but rather to minimize costly errors while facing unavoidable uncertainty




Grow your eCommerce business with SkuVault

Schedule a Demo

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.