Written by Jeff Coull, Director of Professional Services at Retalon
We’ve all heard the retail cliche, “we need to do it differently.”
Yet year after year, organizations plan to do the exact same thing as before.
Slight hyperbole, I know, but using last year (LY) actuals as your baseline for creating this year (TY) plans is doing just that.
Most organizations are creating financial plans that mirror their successes and, more importantly, their failures from the prior year. The worst part about this is how easily it can be avoided through a modest technology investment. A predictive analytic solution can provide LY’s level of demand, rather than simply actual sales, allowing planning teams to consistently get better and do things differently.
Demand Forecasting in an Unpredictable World
Retail used to be relatively stable – seasonal patterns, year-over-year trends, and historical sales were enough to build decent forecasts. But today, we’re living in a very different reality.
Between supply chain instability, inflation, economic uncertainty, shifting consumer behavior, and even viral moments on social media, demand can change overnight. What worked last year may be irrelevant this year, or even this month.
In fact, the Financial Times recently reported that U.S. shoppers are pulling back on discretionary spending due to inflation and economic uncertainty — contributing to lower foot traffic, weaker retail sales, and declining consumer sentiment (Financial Times, 2025). This kind of unpredictable consumer behavior makes it nearly impossible to rely on historical data alone when forecasting demand.
In a world this unpredictable, forecasting needs to evolve. Retailers can’t afford to rely on outdated data or static planning cycles. They need a way to understand what’s happening now and adapt in real time. That’s where modern demand forecasting comes in.
Forecasting Sales Isn’t Just Math — It’s Context
Predictive analytics help answer complex questions — and demand forecasting is full of them.
From a retail perspective, predictive analytics can tell you what you’re going to sell by calculating a unified demand forecast — or, more specifically, what you would have sold had the conditions resulting in missed opportunities not occurred.
Lost sales resulting from out-of-stock situations are a good example.
If you were out of stock for several weeks last year in key products, then using LY as your baseline will result in a low sales plan for those weeks TY, which drives lower purchases and therefore less inventory. You’re planning to repeat your poor inventory position, which will most likely lead to lost sales again.
Good demand forecasting requires asking deeper questions:
- Were we out of stock?
- Did we run a promotion?
- Did a competitor enter or leave the market?
- Did consumer preferences shift?
Relying only on last year’s sales data flattens all these variables into a single, misleading number. That’s why accurate sales forecasting requires a more contextual understanding of demand – not just what happened, but why it happened, and what should have happened instead.
The Case Against Sales-Based Demand Forecasting
Last year’s sales figures may feel like a safe starting point. They’re real. They’re familiar. But they’re also misleading.
Using them as the foundation for your forecasts assumes the conditions in the market are the same — which couldn’t be further from the truth. Every year is different, and forecast accuracy suffers when you don’t account for shifting consumer behavior, external disruptions, supply chain issues, or unexpected events.
Simply put, lost sales isn’t easy to calculate, and it’s a politically charged metric, so people naturally avoid using it.
It’s not as simple as looking at zero inventory situations and assuming missed opportunities. Calculating lost sales requires a forecasted level of demand that incorporates cannibalization factors, size range distribution, elasticity of demand, lead times and repeat availability, past promotional uplift relative to current/future promo activity, daily weights, nonannualized sales spikes, and so much more to deliver the accuracy required.
Therefore, the level of demand has to be compared to what actually happened to determine both the probability and magnitude of a lost sale.
For example, just because a store was out of stock, should a lost sale be calculated if all stores, even those with inventory, had zero sales as well?
The cost of not getting this right is significant. McKinsey reports that AI-driven forecasting can reduce supply chain errors by 20 to 50 percent — and cut lost sales and product unavailability by up to 65 percent (McKinsey, 2022). That level of improvement just isn’t possible when you’re working with static, backward-looking methods.
Different people are going to have very different opinions on the definition of a lost sale. But one thing is clear: it’s nearly impossible for your typical planning manager to do this kind of heavy lifting in Microsoft Excel – especially in today’s fast-paced, omnichannel retail world.
Forecast accuracy requires more than just historical sales data. Technology is the only answer, and only a fully integrated predictive analytics solution can deliver the level of accuracy retailers need today.
Doing Things Differently
Ultimately, there’s a lot of intense math and computer science needed to deliver an effective predictive analytic solution that supports retail planning and all its nuances.
That’s why modern retailers are turning to AI-powered demand forecasting solutions that move beyond static spreadsheets and historical averages. AI can analyze real-time data, identify hidden patterns, and adapt to change as it happens.
Instead of asking, “What happened last year?”, these systems ask, “What’s happening now, and what does that mean for future demand?” They consider everything from promotions and seasonality to weather changes and even viral social trends.
The best option is to choose a comprehensive, integrated solution, one that incorporates an extensive list of factors to deliver an accurate demand forecast — bonus points if the provider has a planning solution as well!
Using level of demand (not just last year’s actuals) as your planning baseline leads to more accurate plans, better inventory positioning, and stronger ROI year after year — as you truly start to do things differently.
It may seem daunting, but these technologies are here — and the benefits are real. Still unsure? Book a demo, and we’ll show you exactly how it works — no guesswork, no fluff. Just results.
About the Author
Alina Sayapov, M.Ed., PMP, is the Director of Marketing and resident translator of tech-to-human at Retalon. Alina uses her Masters in Education to turn retail AI jargon into plain English (and occasionally, puns). When she’s not strategizing content or fueling growth, she’s probably explaining to someone why machine learning isn’t actually scary. Alina’s real mission is to make smart technology less intimidating—and maybe even fun. She believes that if AI can predict what a retailer needs before they know it, marketers can at least try to explain it without a PowerPoint. 
