Automating stock control with alerts: finding out before the customer does

There's an unwritten rule in inventory management: if you rely on remembering to check stock manually, sooner or later you'll forget at the worst possible moment, right when that specific product starts selling more than usual. Stock alert automation solves a seemingly simple problem with a direct impact on sales: warning that a product is running low before it actually runs out, with enough time to restock without losing a single sale to unavailability.

Why manually checking stock doesn't scale

With a twenty-product catalogue, checking stock by hand every few days is feasible. With two hundred products and several suppliers with different delivery times, it becomes a task impossible to sustain consistently, and that's exactly where problems start appearing: products running out with nobody noticing until a customer tries to buy them, or the opposite problem, overstocking products that don't turn over while capital that could be generating more profit elsewhere sits idle.

The types of alert that make a difference

• Minimum stock alert. The most basic type: automatic warning when a product drops below a defined threshold, ideally calculated based on that specific product's sales velocity and its supplier's delivery time, not a fixed number the same across the whole catalogue.
• Accelerating sales trend alert. More sophisticated: detects when a product starts selling faster than usual (for example, after going viral on social media or entering a well-received promotion), and warns before the traditional minimum stock threshold would have triggered, giving more time to react.
• Stagnant stock alert. The reverse of the above: warns when a product has gone a long time with barely any turnover, a sign it might be worth a promotion, a price cut, or simply no longer restocking it.
• Restock deadline-at-risk alert. Combines current stock level with the supplier's known delivery time: if remaining stock won't last until the restock arrives, it warns with enough lead time to place the order in time.

Calculating the right threshold: it's not a fixed number

One of the most common mistakes when setting up stock alerts is using the same numeric threshold across the whole catalogue, without accounting for the fact that a product selling five units a day needs a very different warning threshold from one selling five units a month. The right threshold is calculated by combining that specific product's average sales velocity with its supplier's delivery time, plus a reasonable safety margin for unexpected demand spikes. A product with a fast supplier (2-3 day delivery) needs much less safety margin than one with a slow supplier (3-4 week delivery), even if both sell at the same rate.

A specific case: the store that stopped losing its best-seller every month

A personal care products store had a product that accounted for a disproportionately large share of its revenue, but that ran out fairly regularly because the configured restock threshold (the same across the whole catalogue) didn't reflect its extremely high real sales velocity. Every time it ran out, they lost several days of sales on their most profitable product while waiting for restock. After recalculating the threshold specifically for that product, based on its real sales velocity and its supplier's lead time, the alert began triggering with enough margin to place the order before running out, completely eliminating those recurring stockouts.

How AI improves on this beyond fixed rules

The more advanced inventory management systems use predictive models that go beyond a simple fixed-threshold rule: they analyse seasonal patterns (a seasonal product needs early restocking ahead of the predictable peak), recent trends (an acceleration in sales over recent weeks automatically adjusts the prediction of when it will run out), and even external factors when available (for example, if a specific ad campaign is about to launch and is expected to spike demand for a particular product). This allows anticipating stockouts far more precisely than a static rule that doesn't adjust to the business's changing circumstances.

Connecting alerts to action, not just notification

An alert that arrives and nobody acts on is of little use. The next level of automation, increasingly accessible even for small businesses, connects the alert directly to action: automatically generating a draft order to the supplier when the alert triggers, ready for someone to review and confirm instead of creating it from scratch, or even automatically sending the restock order with no human intervention for suppliers with an established, trusted relationship and pre-agreed pricing.

Alerts by product, not just by warehouse

Setting up alerts only at the overall business level, without breaking them down by location or sales channel, can create a false sense of security: total stock may look sufficient while it's actually poorly distributed, with a specific warehouse or channel already out and another with excess. Setting up specific alerts by product-and-location combination, not just a single global number, avoids this blind spot and allows reacting before the stockout becomes visible in the channel where it's actually happening.

How to avoid alert overload nobody ends up reviewing

A poorly calibrated alert system that constantly warns about everything ends up producing the opposite of the intended effect: the team gets used to ignoring notifications because there are too many, and the genuinely important alerts get lost in the noise. Prioritising alerts by real impact (high-turnover or high-margin products first) and adjusting sensitivity to reduce false positives is just as important as having the alert system in the first place; a system nobody actually looks at adds no value, however sophisticated it is.

Common mistakes when configuring stock alerts

The first frequent mistake, already touched on but worth expanding, is copying the same threshold from another business or a generic template found online, without adapting it to your own catalogue's real sales velocity. A threshold that works well for a fast-turnover fashion business can be completely unsuitable for a furniture business with much longer sales cycles, and applying the same criterion with no adjustment generates constant alerts in one case and insufficient ones in the other.

The second mistake is not accounting for the variability of the supplier's delivery time, treating it as a fixed number when in practice it varies (holidays, the supplier's own demand spikes, transport incidents). A threshold calculated on the supplier's "ideal" or "promised" delivery time, with no margin for the real variations that specific supplier has shown in the past, leaves the business exposed exactly when the supplier takes longer than usual, which often coincides with the demand spikes where running out of stock hurts most.

The third mistake is setting up alerts and never checking whether the thresholds are still correct as each product's sales velocity changes. A product that starts selling much faster after a successful ad campaign, or that slows down after losing seasonal relevance, needs a different threshold than it did six months ago, and businesses that configure alerts once and forget about them end up with a system increasingly out of step with the catalogue's current reality.

The fourth mistake is assuming a low-stock alert always means "restock now," without checking whether that product is still worth restocking. A product running low but whose demand has fallen steadily over recent months doesn't necessarily need automatic restocking at the same level as before; restocking out of inertia, just because the alert fired, without checking whether the product's conditions have changed, can end up tying up capital in a product that no longer has the same pull it had when the original threshold was calculated.

The fifth mistake, more about change management than technical setup, is not explaining to the team why this process has been automated and what's expected of them now. If someone used to eyeball stock every morning and that task now disappears with no explanation, it's common for them to keep manually checking "just in case," duplicating work with no extra benefit, or to stop paying attention to inventory altogether, losing the human judgement still needed to correctly interpret every alert the system generates.

The sixth mistake is not accounting for the fact that different products may need different people responsible for reacting to an alert. Routing every stock alert to a single person, regardless of product category or supplier involved, creates bottlenecks when that person is on holiday or overloaded, while splitting responsibility by specific categories or suppliers, with alerts routed to the right person from the start, speeds up reaction time and stops an important alert sitting in the inbox of someone who isn't the one meant to act on it.

A seventh mistake, simple to fix, is not occasionally checking whether alerts are still arriving through the right channel. An alert email that starts landing in the spam folder, or an app notification that stops sounding after a phone update, can leave the system technically working fine while nobody actually receives the warning anymore, a silent failure only discovered once the problem the alert was meant to prevent has already happened.

Frequently asked questions