For modern businesses that carry inventory, it’s challenging to strike the right balance between supply and demand. Sales surge and slow, customer buying habits change, and global market factors (like trade tariffs) impact pricing. And when demand is not forecasted properly, it can lead to a ripple effect of issues, like slow-moving stock and crammed warehouses. But with the right combination of artificial intelligence (AI) and human expertise, businesses can reduce inflated targets, lower safety stock levels, and improve warehouse operations.
To explain how this works in the real world, we’re going to look at a case involving Meridian Wine Distribution, a leading business in the liquor distribution sector in South Africa. As peak season approached, Meridian found itself in a classic distribution dilemma: after a surge, demand had softened, leaving them with thousands of excess cases of wine. Here’s how they were able to combine AI inventory management tools with disciplined human review to reset safety stock levels, clean distorted demand history, and cut excess inventory by about 15%.
(Check out the case study for more specifics).
The Pain: A Warehouse Under Strain (Peak Season Reality)
Peak season at Meridian is normally intense, but predictable. Orders increase, the team moves fast, and the warehouse hums at full throttle. In 2022, however, that rhythm changed.
Here’s what happened:
- October looked promising. Sales targets were met, confidence was high, and planners did what responsible distributors do: they carried extra stock to protect service levels.
- By mid-November, demand began to slow. The warning signs were subtle at first, but by the end of the month, Meridian was stuck with around 33,000 cases of excess wine.
- The main warehouse climbed past 90% utilization, a threshold that led to costly overflow storage. Pallets even started to stack on the floor, leading to physical risks for warehouse staff. Industry studies show that overstocking costs businesses upwards of $758 billion annually.
On the surface, Meridian was ready to deliver to its customers. But in reality, its warehouse was stretched thin and costs were rising.
The Signal They Were Missing: Why “Unusual Sales” Matter
The problem Meridian faced was due to a critical (yet common) scenario: sales teams were setting unrealistic targets based on historical peaks without accounting for outlier events. Promotional peaks and unusually large orders were treated as normal demand. When that happened, three things followed:
- Sales teams built next year’s targets on top of those peaks, effectively compounding them.
- Planners translated those inflated targets into higher forecasts and safety stock levels.
- Procurement ordered more stock “just in case,” which eventually showed up as excess in the warehouse.
What’s the solution? Instead of treating every spike as real, repeatable demand, Meridian needed a way to distinguish consistent sales trends from exceptional events. Without that signal, Meridian was effectively “betting on lightning striking twice” – planning as if every peak would happen again.
That’s when they turned to StockIQ’s Unusual Sales detection tool, in combination with human oversight.
The Pivot: From Unrealistic Targets to Normalized Demand
Meridian’s turning point began with a shift in mindset. Instead of asking, “How much more do we need to hold to protect service?” the team started asking: “Which of our past sales actually deserve to shape the future?”
Here are steps you can follow to do the same:
1. Let AI quickly surface exceptions
The first change was to stop relying on optimistic targets as a “future signal.” Using StockIQ’s AI-powered Unusual Sales Detection, Meridian let the system scan its 770+ SKUs and automatically flag demand anomalies – unusually large orders, promotional spikes, or patterns that deviated from what was “typical” for that item, customer, and time period. Research from McKinsey shows that applying AI-driven forecasting to supply chain management can improve accuracy by up to 50%.
2. Add the human layer (where real decisions happen)
Flagging anomalies was only the start. The real work happened in the conversations that followed. For each flagged event, Meridian’s planners would:
- Attribute the spike. Was it a promotion, a trade event, or a one-off customer order?
- Talk to Sales to capture missing context.
- Create and document an “event” in StockIQ so the story lived with the data.
- Decide whether to normalize or exclude the spike so it wouldn’t inflate future forecasts.
3. Set realistic sales targets
By removing one-time spikes, Meridian “flattened” its historical data, before sales teams built the next year’s targets. This prevented the artificial inflation that previously led to over-ordering, and allowed their sales team to receive more realistic baseline demand data for planning purposes.
4. Make it a continuous clean-up
Lastly, Meridian started treating history cleaning as an ongoing discipline. They began to regularly track inventory spikes, set milestones to ensure they were continuously moving in the right direction, and embed Unusual Sales reviews into their team’s daily and weekly workflows.
Meridian’s results came from a repeatable, measurable system that involves both AI tools and careful human oversight. By using AI inventory management tools like StockIQ’s Unusual Sales feature, you can flag spikes, and create workflows that allow you to either explain them or exclude them from future forecasts.
StockIQ: Fueling Human-Centric AI Warehouses
Meridian’s story demonstrates what happens when a distributor understands their demand signals and pairs that data with the right human judgement. By using AI to surface unusual sales – and humans to interpret, document, and act on them – Meridian replaced optimistic assumptions with normalized demand.
That is what a human-centric AI warehouse looks like in practice. With StockIQ’s AI-powered demand forecasts, features such as Unusual Sales, and granular controls, you can easily avoid inflated sales targets and drive away bloated inventory.
Contact us today or request a StockIQ demo to learn more.
Frequently Asked Questions About Human-Centric AI Warehouses
1. Why can outlier sales throw off a forecast?
When a one-time promotion, trade event, or unusually large customer order stays in the data, the system reads it as repeatable demand. That inflates forecasts, safety stocks, and purchase orders – even though the spike is unlikely to happen again. Over time, those outliers can lead to overstocking.
2. How do AI tools like “Unusual Sales” help flag anomalies?
StockIQ automatically scans sales across SKUs, customers, and time periods to identify orders that look unusually large. Then, planners can mark them as Unusual Sales, and choose to omit them from future forecasts.
3. What role do humans play in normalizing demand forecasts?
Humans provide the context AI can’t reliably infer. Planners investigate flagged spikes, speak with Sales to understand what happened, create and document events in the system, and decide whether a spike should influence future forecasts or be normalized out.