Not all customers contribute to a store's business in the same way. Some buy often, others make large baskets, some have been loyal for a long time and others are only starting to show interesting behavior.
Identifying your best customers therefore means combining several criteria rather than simply ranking your clientele by revenue. The goal is not just to produce a ranking. It is to better understand the different customer profiles so you can tailor loyalty and communication efforts.
A best customer is not necessarily the one who spends the most
Take two customers. The first made a €1,000 purchase two years ago and never came back. The second has spent €80 every month for a year.
The first generated more revenue in total: €1,000 versus €960. Yet the second's behavior reveals a much more regular and recent relationship with the store. This does not automatically mean they have greater “value” in every context. It mainly shows why cumulative revenue is not enough to characterize the customer relationship.
Which criteria should you look at?
Several metrics can be used depending on what the store is trying to identify.
Amount spent
How much has the customer spent over the period studied? It is important to define this period. Comparing the revenue generated over the last twelve months does not tell you the same thing as cumulative revenue over five years.
Purchase frequency
How many transactions does the customer make over a given period? A customer who comes back every month does not behave like one who buys once a year.
Recency
How long ago was the last purchase made? This metric helps distinguish a customer who is still active from a former good customer who may no longer visit the store.
Average basket
How much does the customer spend on average per transaction? Two customers generating the same annual revenue can get there in very different ways: many small purchases for one, a few large purchases for the other. See increasing the average basket.
Regularity
Frequency tells you how many times a customer buys. Regularity adds another piece of information: are these purchases spread fairly evenly over time? Two customers may each have made twelve purchases in the year. The first comes about once a month; the second made all twelve purchases over a very short period and then disappeared. Their annual frequency is identical, but their behavior is not.
The RFM method: recency, frequency, monetary amount
A classic method for segmenting a clientele uses three dimensions:
- Recency: how long ago did the customer buy?
- Frequency: how many times have they bought?
- Monetary amount: how much have they spent?
This is the RFM method. The principle is to give customers a score on each of these dimensions. A customer can thus be:
- very recent;
- very frequent;
- high in amount.
Or have a different profile: a high amount but low frequency, for example. The point of RFM is not to mathematically determine “who is the best customer” in absolute terms. It mainly brings out different behaviors within the clientele and helps build usable segments.
You need to define what “best customer” means
The notion actually depends on the objective. If the store is looking for its biggest revenue contributors, the amount spent will be decisive. If it is looking for its most loyal customers, frequency and regularity will matter more.
If it is looking for customers to win back, it may look for people who were historically important but whose recency is slipping. If it is looking for high-potential customers, it may focus on recent profiles whose frequency or amount is growing quickly.
There is therefore no universal ranking of best customers. The right ranking depends on the business question being asked.
A few useful profiles to tell apart
A segmentation can, for example, reveal several types of customers:
Very active, high-value customers
They buy recently and regularly, and represent a significant amount over the period studied. They are naturally a group to build loyalty with.
Regular customers
They come back often, even if their baskets are not necessarily the largest. Their regularity can be especially valuable for a store built on recurring purchases.
Big occasional buyers
They spend large amounts but buy infrequently. Depending on the business, this behavior can be perfectly normal: a store selling slow-replacement products will not interpret this frequency the same way as a bakery or a grocery store.
Former good customers
They used to have a high frequency or amount, but their last purchase is getting old. They are an interesting target to analyze before any follow-up.
New customers with potential
They still have little history, but their first purchases may show an interesting frequency or spending level. Avoid labeling them too quickly, though: a few transactions are not always enough to predict a lasting relationship.
Adapt the analysis to the type of store
The same thresholds cannot be applied everywhere. A customer who has not come back in 30 days may be considered unusually absent in a store where purchases are weekly. In a furniture store, 30 days without a purchase may on the contrary be perfectly normal.
Likewise, five purchases a year may be a high frequency in some sectors and very low in others. Segmentation should therefore be built on the store's actual purchasing rhythm, not on generic thresholds.
What to do once you have identified your best customers?
The analysis becomes useful when it lets you adapt an action.
Build loyalty with active customers
Customers who are already regular can benefit from perks consistent with their behavior or from special attention, through customer loyalty capabilities.
Win back former good customers
A historically important customer who no longer comes back may justify a specific win-back campaign. But before following up, it is worth checking that their absence is really unusual compared with their usual purchasing rhythm (see following up with inactive customers).
Personalize offers
The categories or products usually purchased can help you offer more relevant communication than a generic promotion sent to the whole base.
Avoid unnecessary discounts
A very good customer who already buys regularly does not necessarily need a big discount to come back. Loyalty should therefore not boil down to systematically handing out reductions to your best customers. It can also take the form of a service, priority access, personalized information or a perk truly suited to the profile.
Watch the margin
Another point deserves consideration when the data is available. A customer who generates a lot of revenue is not necessarily the one who contributes the most to margin. If they mainly buy during promotions or regularly get large discounts, their revenue may give an incomplete picture of their economic contribution.
Customer analysis can therefore be enriched with margin when that data is available and relevant. Customer revenue and customer profitability are not synonymous.
How can AI help?
When it can use customer data and the associated sales, AI makes it easier to explore different criteria. The merchant can ask:
“Which are my 20 customers who generated the most revenue over the last 12 months?”
This wording is more precise than:
“Who are my 20 best customers?”
because it explicitly defines the criterion used. They can then change angle:
“Which customers bought most often over the same period?”
Then:
“Among my most regular customers, which ones haven't bought for longer than usual?”
You then move from a generic ranking to an analysis truly suited to the business question. These answers rely on the store's sales analysis.
How can Gillia help?
Gillia can query the available customer data by several criteria: recency, frequency, amount or average basket. A first request can focus on revenue:
“Which customers generated the most revenue over the last 12 months?”
Then the analysis can be refined:
“Among them, which ones buy regularly?”
or:
“Which ones haven't bought in more than two months?”
The value is being able to gradually combine several dimensions without reducing the notion of “best customer” to a single metric. Once a relevant segment is identified, it can serve as the basis for a tailored loyalty or follow-up action, which is the very principle of an AI agent for retail.
The case of a multi-location business
For a chain with several stores, the analysis benefits from being consolidated across the network when customer identification allows it. The same customer can buy in several locations.
If each store is analyzed separately, a customer who is important at the chain level can look fairly ordinary in each point of sale. A consolidated view therefore lets you take into account their whole relationship with the chain, not just their activity in one store (see multi-store management). More customer use cases are available in our full catalog of Gillia use cases: identify your best customers or check shared loyalty across locations.
Identify to act better
The question “Who are my best customers?” seems simple, but it always needs to be made more precise. Best by what measure?
- revenue;
- frequency;
- recency;
- average basket;
- regularity;
- possibly margin.
The right metric depends on the objective. By combining these dimensions, the store can tell apart its regular customers, its big buyers, its former good customers and even some profiles with potential.
The point is ultimately not to build a podium. It is to understand customer behavior so you can choose a more relevant action.