In almost every store, a minority of customers carries most of the revenue. That is neither a problem nor an achievement in itself. The problem starts when next year’s plan quietly assumes that this year’s best customers will spend just as much again. Measure concentration over a longer period, check whether the strongest group is stable and why it buys, and only then decide how much attention – and how much of the plan – it deserves.
How to measure concentration
Rank customers by what they spent over the last 12 months and calculate the share of revenue excluding VAT that comes from the top 1%, 10% and 20%. Three cut-offs say more than one: the top 1% shows dependence on a few specific people, the top fifth shows the breadth of your core.
To keep the numbers comparable, follow three rules. Exclude cancelled orders and calculate revenue the same way in both periods. Identify customers by the same key – in a store with many guest checkouts that is usually the order e-mail, otherwise one person splits into several “customers”. And count several orders on the same day as one purchase.
An illustrative example with invented numbers: a store had 1,200 customers and CZK 3,000,000 in revenue excluding VAT over 12 months.
| Group | Customers | Revenue excl. VAT | Share of revenue |
|---|---|---|---|
| Top 1% | 12 | CZK 540,000 | 18% |
| Top 10% | 120 | CZK 1,350,000 | 45% |
| Top 20% | 240 | CZK 1,800,000 | 60% |
Twelve people carry almost a fifth of revenue. Whether that is good, we do not yet know – the next steps will show.
How much is “a lot”
The 80/20 rule is a shorthand rather than a law. A large study of US packaged goods found that the top fifth of a brand’s customers carried 73% of sales on average, ranging from 65% to 90% across categories, with niche brands higher. Across 339 publicly traded companies outside packaged goods, the average was 67%, and 68% for non-subscription businesses. A department store chain in East Asia drew 71% of revenue from its top fifth of customers.
These results differ in method, period and product range, and they are not a benchmark for a Czech online store. Two things are useful. Values of two thirds to three quarters are common, so a high number alone signals nothing. And niche ranges – collectibles, for example – tend to be more concentrated. More important than any outside average is the trend in your own figure: compare the last 12 months with the same period a year earlier.
Short windows inflate concentration
Over a short period only some customers buy, and chance in who happened to buy looks like a difference between people. A classic consumer-panel study shows this directly: for ketchup, the top fifth of households accounted for 86% of purchases in one month but 52% over two years; for yogurt, the share fell from 97% to 73%.
For an online store the rule is simple. Do not use a monthly or quarterly share of top customers to judge how dependent the business is. Twelve months capture the season and customers who buy once a quarter.
Who carries your revenue
See how much revenue your strongest customers carry and which formerly regular customers have stopped buying.
This year’s best customers will spend less next year
The same research names a second trap: a group that showed highly concentrated buying in one period will not deliver the same concentration in the next. This is regression to the mean. Some customers are at the top permanently; others got there through one exceptional year, a larger order or chance. The department store study followed the top fifth of customers over three years, and its stability over time ranged from 11% to 74% across departments.
So measure overlap as well as share. Take last year’s top tenth and check how many are in the top tenth again this year and how much they spent compared with last year. In the illustrative store, last year’s 120 strongest customers spent CZK 1,140,000 last year and CZK 700,000 this year, and 55 of them are back in this year’s top tenth. This year’s top group is therefore largely made up of different people. A plan built on last year’s ranking would have overstated revenue.
When concentration is a strength and when it is a risk
The same percentage can mean a solid core of loyal customers or revenue hanging on a few chance orders. Three questions tell them apart.
| Question | More likely a strength | More likely a risk |
|---|---|---|
| How do they buy? | Repeatedly, on many different days, with a recent last purchase | One or two large orders, with no purchase for a long time |
| At what margin? | At regular prices and on goods with a reasonable margin | Mainly on discount, at individual prices or on low-margin goods |
| What would replacement cost? | A few ordinary new customers cover the loss of one | Replacement would need dozens of new customers and costly acquisition |
Buying rhythm. Ranking by spend is not enough. Research on customer value shows that a customer’s future contribution depends on the combination of how recently they last bought, how often they buy and how much they spend; customers with different histories can have the same future value. A customer with 15 purchases a year and a customer with one order of the same value are not the same risk.
Margin. Revenue concentration does not tell you where the profit is. For publicly traded companies, preliminary results found profit even more concentrated than sales. In an online store it can be the other way round if the strongest customers buy with volume discounts. Calculate the gross margin of the top group as well.
Cost of replacement. Compare the annual spend of a typical top-group customer with a new customer’s first-year spend. Illustratively: if a top customer spends CZK 25,000 a year and a new customer CZK 1,800 in their first year, replacing one takes roughly 14 new customers. Only with the cost of acquiring them do you know what caring for existing customers is worth – and what losing them costs.
Handle data about specific people sparingly
Measuring concentration, overlap or rhythm needs no names. A stable customer key, such as a pseudonymous hash of the e-mail address, is enough. The GDPR requires personal data to be limited to what is necessary and, by default, processed only as far as needed for a specific purpose. The European Data Protection Board lists minimisation, pseudonymisation and access limitation among such default measures.
A named list of top customers makes sense where someone actually acts on it – for example, when the owner personally contacts a long-standing customer who has stopped buying. Shared reports, exports for an agency or inputs for AI should contain aggregate figures only. This section is orientation, not legal advice.
What to do now
- Measure the share of your top 1%, 10% and 20% of customers in revenue excluding VAT for the last 12 months and the 12 months before.
- Check the overlap. How many of last year’s top tenth are in it again this year, and how much they spent compared with last year.
- Split the top group by rhythm. Separate repeat customers from one-off large orders, and for those inactive for a long time, find out why they stopped.
- Add margin and replacement cost. The gross margin of the top group and the number of new customers needed to replace one strong customer.
- Adjust the plan. Plan revenue from the top group on its actual overlap, not last year’s spend, and keep the named list only with people who act on it.
In its Customers section, Korzaro shows what share of all revenue comes from up to ten of the largest customers across the full order history and lists them in a ranking visible only to people with access to the project. Customer groups show how much spend comes from loyal, returning, new and one-off customers, and a separate list of formerly regular customers with no purchase in the last 90 days helps you find whom to contact. Customers are matched by order e-mail, so guest purchases count too. Names and e-mail addresses are not sent to AI inputs. Korzaro does not yet calculate the 12-month share of the top tenth or the year-on-year overlap.