From the Magazine: The inventory playbook for the AI age

Take baby steps when entering the world of predictive analytics

Inventory is the lifeblood of the aftermarket — but also one of its biggest risks. Every part sitting too long on a shelf ties up capital. Every part missing from that same shelf costs a sale, and possibly a customer. For warehouse distributors and retailers in the aftermarket, the old approach of “buy, stock, and hope it sells” no longer cuts it.

Predictive analytics and artificial intelligence are reshaping how companies decide what to stock, where to stock it and when to move it. In the near future, these tools will no longer belong just to global tech firms; they are increasingly practical for aftermarket businesses of every size.

From guesswork to foresight

Predictive analytics simply means using data — your own and external sources — to estimate what customers will need next. AI builds on that by spotting subtle patterns and learning from every new piece of information.

In practical terms, it means replacing instinct and spreadsheets with insight. Instead of relying on last year’s sales alone, AI systems can consider hundreds of variables: Regional vehicle mix, weather patterns, supplier reliability, economic trends and even the timing of new vehicle launches.

For distributors and retailers juggling thousands of SKUs, the payoff is accuracy: Knowing which parts are truly likely to move, and which are quietly eating working capital.

For distributors and retailers juggling thousands of SKUs, the payoff is accuracy: Knowing which parts are truly likely to move, and which are quietly eating working capital.

AI in inventory action

The shift isn’t theoretical. Several companies, both within and beyond the automotive industry, are showing how smarter forecasting works in the real world.

Uni-Select, one of Canada’s largest aftermarket distributors, partnered with supply chain management company Manhattan Associates to overhaul its seasonal demand forecasting at a Montreal distribution centre. The system used machine learning to adjust replenishment automatically as demand patterns shift. According to Manhattan’s case study, Uni-Select achieved better forecast accuracy, reduced exceptions and improved inventory productivity without sacrificing service levels.

Walmart Canada offers another lesson. It piloted computer-vision technology from Focal Systems in about 70 stores to monitor shelves in real time. Cameras detect when products are running low and trigger restock alerts automatically. The retailer later expanded the rollout across the network, citing more accurate shelf visibility and faster replenishment.

And in a study on the automotive aftermarket published in the journal Data Science and Management, researchers found that combining internal sales history with external signals — such as regional demand trends and supplier performance — can improve forecast accuracy for slow-moving parts by more than 20 per cent.

Each of these examples highlights a key point: Predictive systems thrive when data is broad, clean and constantly refreshed.

The Canadian challenge

Canada presents unique challenges for inventory planning. Our geography and climate produce wildly different vehicle needs from coast to coast. Trucks dominate in Alberta; imports in Ontario and Quebec; older cars persist longer in the Maritimes. Add tariffs, cross-border shipping delays, and regional weather swings and it’s easy to see why national averages rarely help.

Predictive analytics allows companies to localize their planning. A distributor in Calgary might factor in winter-tire seasonality and vehicle age, while one in Montreal focuses on hybrid-component demand. AI can handle these micro-patterns simultaneously, something even the best human planners can’t do manually.

It’s also a hedge against volatility. Tariff changes, shipping costs and global supply hiccups can alter lead times overnight. Companies using predictive systems can see early warning signals — and reorder, substitute, or rebalance inventory before competitors even notice a shift.

The shift isn’t theoretical. Several companies, both within and beyond the automotive industry, are showing how smarter forecasting works in the real world.

So, where do you begin?

Pick a focus area. Choose a product family where availability really matters — brake parts, filters or sensors.

Clean your data. Get this wrong and the rest won’t matter. Make sure historical sales, returns and supplier lead times are accurate. Garbage in, garbage out still applies.

Add external data. Weather, vehicle-registration trends, or even local economic indicators can make forecasts more realistic.

Test and measure. Start with one region or category and compare results: Fill rates, stockouts, obsolescence and cash tied in inventory.

Automate exceptions. Use AI not just for forecasts but for alerts, such as flagging which SKUs are drifting off forecast or which suppliers are causing slowdowns.

What the future looks like

As computing power grows cheaper, predictive analytics will move closer to the edge, embedded in warehouse systems, store-level dashboards and even supplier portals. The next generation of platforms will integrate forecasting with live supply-chain signals, helping distributors adjust purchase orders automatically as real-world data changes.

Meanwhile, AI models are learning to interpret unstructured information, such as technician search data, service-bay trends and parts-lookup queries, offering yet another source of demand insight. For the aftermarket, that means forecasting not just what sold last month, but what customers looked for and didn’t find.

At its heart, predictive analytics isn’t about algorithms — it’s about visibility and timing. The distributors and retailers that can sense demand shifts early and adjust before others will outperform those relying on gut feel.

Inventory used to be about control. Now it’s about action. Get in on it.

---

Kumar Saha is Vice President (U.S.)/managing director (Canada) of global automotive data firm Eucon. He has been advising the North American automotive industry for over a decade and is a frequent conference speaker and media commentator. He is based out of Toronto.

This article originally appeared in the November 2025 issue of Jobber News