Forecasting Logic For Aligned Omnichannel Inventory: Solving the Quiet Failures

Forecasting rarely fails in obvious ways. The tools are there, the models are available, and the teams are often experienced. And yet, it’s not uncommon to see inventory that is misaligned, stores and channels that are over- or under-served, and a growing sense of frustration when the numbers don’t reflect what’s really happening on the ground.

In my work helping companies optimize forecasting and inventory planning, I’ve found that the issue isn’t usually the math; it’s the logic behind the math. Small assumptions, unnoticed by those deep in the systems, can lead to persistent gaps between what the model says and what the business needs. These aren’t dramatic failures; they’re subtle oversights that quietly erode accuracy.

One of the most common issues involves how organizations calculate forecast components like seasonality or promotional uplift. Often, these metrics are calculated at the corporate level, aggregating data from both e-commerce and brick-and-mortar channels. But these channels behave differently. For instance, e-commerce peak earlier during Black Friday due to digital campaigns and many customers avoiding busy stores at that time, while brick-and-mortar sees a surge closer to Christmas as shoppers avoid shipping delays. Moreover, one channel usually dominates the sales data at corporate level. As a result, its behavior overshadows the other channel, resulting in distorted forecasts and misaligned inventory allocation.

Another frequent issue lies in how product hierarchies are defined. For example, calculating seasonality at the “t-shirt” level often flattens the curve, because it combines short-sleeve and long-sleeve styles; one peaking in summer, the other in winter. When grouped together, those distinct patterns cancel each other out and distort the input.

I’ve seen how small, well-targeted changes, like adjusting seasonality logic by channel or using more precise product groupings, can dramatically shift forecasting accuracy. In one case, these adjustments led to multi-fold improvement in e-commerce forecast accuracy during Black Friday. In another, better inventory allocation cut lead time for online fulfillment by half.

And finally, there is the difference between forecasting sales versus demand. Sales tell us what was sold. Demand includes what could have been sold. If a product sells out and enough lost sales aren’t estimated, future forecast end up even lower. It’s a self-fulfilling cycle that scauses understocking to persist.

Note that estimating lost sales doesn’t have to be complex. For example, if one store sells out completely while other similar stores (based on traffic, size, region, or weather), still have inventory and continue selling, that is a strong indication that demand exceeded recorded sales. Even simple rules like this can help close the gap between what actually happened and what the system captured.

These types of improvements are well within reach, even without advanced systems. Remember, forecasting doesn’t need to be flawless, but it does need to reflect your business reality. If you want to empower your team to identify gaps like these, place them at the heart of your business. The strongest forecasting improvements I’ve seen happen when planning teams stay close to the business, work alongside IT, merchandising, store and DC operations, and remain aware of how executive strategy and the business environment evolve over time.

Stay tuned for my next article, where I’ll share practical insights on inventory strategy.

Thanks for reading!