A drop in revenue, a product that suddenly stops selling or an average basket that changes sharply isn't necessarily an anomaly. Sales naturally vary with the day, the season, the weather, promotions or the calendar.
A sales anomaly is a gap unusual enough compared with the normal pace of activity to deserve a check. The question is therefore not just: “have my sales changed?”, but rather: “is this change consistent with what usually happens?”
Detecting these gaps early enough lets you look for their cause before a problem sets in — or before an opportunity goes unnoticed.
A few examples of anomalies
Unusual drop in revenue
A drop becomes worth analyzing when it is clearly larger than what the season, the calendar or usual activity can explain.
Sudden fall in sales of an item
An item that sells regularly and suddenly hardly sells at all can reveal a stockout, an availability problem, a change in demand or another situation to check.
Unusual increase
A sharp rise is also worth understanding. It may come from a promotion, an event, a change in behavior or exceptional demand.
Change in the average basket
A sudden rise or fall in the average amount per ticket can reflect a change in the composition of purchases, in prices or in the sales profile.
Change in payment methods
An unusual shift in the split between card, cash or other payment methods can sometimes signal a change in behavior or an operational problem to check.
Compare against a relevant reference
An isolated figure doesn't tell you whether it is abnormal. To interpret a variation, you need to choose a relevant point of comparison. Depending on the situation, it can be:
- the previous period;
- the history over several weeks or several months;
- a comparable day;
- another location in the network;
- a comparable family or item.
Comparing a Saturday's sales with the previous Monday's can, for example, produce a large gap without there being any anomaly at all. Likewise, a drop compared with the previous week can be perfectly normal if that week was exceptionally strong. The quality of the reference therefore matters as much as the observed gap.
Possible causes of a sales anomaly
Once the gap is confirmed, you need to look for what might explain it.
A stockout
A drop in sales may simply come from a product that was no longer available. In that case, the drop in sales doesn't necessarily indicate a drop in demand.
Seasonality
Some variations naturally repeat at certain times of the year. A peak or a dip can look unusual when compared with the previous week while being perfectly consistent with the same period last year.
A promotion
A commercial campaign can cause a one-off increase in sales or change how they are spread across several items.
A price change
A price increase or decrease can influence quantities sold, the average basket or the trade-offs between products.
An inventory problem
A gap between theoretical stock and the stock actually available can give a misleading picture of the situation.
A change in footfall
Revenue can also vary simply because more or fewer customers walked into the store. You then need to separate the change in traffic from the change in average basket or products purchased.
A four-step method
To analyze an anomaly without jumping to conclusions, you can follow four steps.
- DetectA metric departs from its usual behavior: revenue, quantity sold, average basket, category or item.
- CompareThe gap is checked against a relevant reference to make sure it isn't just a normal variation.
- Find the causeYou examine the factors that may explain the gap: product availability, promotion, price, seasonality, footfall or another known event.
- ActOnce the cause is understood, the action becomes far more relevant: check stock, adjust restocking, keep an eye on an item or take advantage of an unexpected increase.
This last step is important: an anomaly is a signal to examine, not automatically a problem to fix.
Why are anomalies hard to spot by hand?
A store can track hundreds or thousands of items, on top of revenue, number of tickets, average basket, categories, inventory or even different locations. Most of this data changes normally.
The problem is therefore less about having the figures than about spotting, amid all these variations, the ones that truly deserve attention. A major anomaly on a single item can easily stay invisible in an overall revenue figure that looks stable.
How can AI help detect anomalies?
AI can compare a recent period with a reference period and bring out notable gaps in the available data. The merchant can, for example, ask:
“Has anything unusual happened in my sales this week?”
Then dig deeper:
“Why did sales in this family drop?”
Or:
“Which items changed the most compared with the previous 7 days?”
The analysis can then cover revenue, tickets, average basket, or the families or items concerned. The benefit is being able to start from a signal and gradually dig into its origin, rather than going through several reports by hand — relying on the business's sales analysis.
From detection to action
Spotting an anomaly quickly mostly tells you where to look. If the drop in an item seems linked to a stockout, you need to check its available stock. If sales slow down while stock remains high, it may be worth checking whether the item is becoming dead stock. An unusual increase, on the other hand, may justify checking the remaining stock to avoid a future stockout.
The action therefore depends on the cause identified, not simply on whether a curve goes up or down.
The multi-location dimension
In a network of stores, comparing locations brings extra information. A drop seen across all points of sale may reflect a general trend. If it affects only one location, you should instead look for a local cause: a stockout specific to that store, an operational problem, a different assortment or a change in its footfall.
Comparing locations thus lets you tell a global trend from a local anomaly.
How does Gillia help with this analysis?
Gillia can use the available sales data to analyze a period and compare it with another. A request can start simply:
“Analyze my sales for the last 7 days.”
Gillia can then summarize the main available metrics — revenue, number of tickets, average basket — and bring out the families or items that are rising or falling. The analysis can then be taken further:
“Is there anything unusual?”
then:
“Compare with the previous 7 days.”
The goal is to let the merchant move quickly from the observation to checking the cause, then use that information to decide on the right action — the principle of an AI agent for retail. This same reading of gaps also feeds revenue forecasting, and shows up every day in the store dashboard. To go further, explore the Gillia use cases related to sales tracking: spot an anomaly in sales.
To go further, discover the Gillia POS: every sale updates inventory, customers and the online store.