Sign in

Blog · Board and management reporting

Concentration on ten customers: the whole arithmetic on one page

The complete customer concentration calculation on ten customers small enough to check by hand: the revenue per customer, the sorted shares, top-one, top-three and top-five share, the count to half, the Herfindahl index as the sum of squared shares, the same measures last year for the trend, the per-rep split that shows a rep whose count to half is one, and the identity, so a reader can reproduce every figure and then run it on their own export.

The short answerTen customers, revenue sorted largest first. Top-one share is the largest over the total; top-three and top-five likewise. Count to half is the number of customers, largest first, whose cumulative share first reaches 50 percent. The Herfindahl index is the sum of each customer's share squared. Computed for this year and last, the trend is the finding: top-three rose from 44 to 52 percent with the same ten customers. Split by rep, one rep's book is two customers and their count to half is one. The identity is that the ten revenues sum to the ledger. Every number can be reproduced by hand.

Concentration is three measures on a sorted list, and on ten customers all three can be computed by hand in a minute. This page works them, this year against last, and splits by rep to show what the company table hides.

The customers, sorted

Rank Customer Revenue Share Cumulative Rep
1 F $450,000 23.0% 23.0% R-1
2 A $310,000 15.9% 38.9% R-2
3 E $270,000 13.8% 52.7% R-1
4 I $210,000 10.7% 63.4% R-2
5 G $200,000 10.2% 73.7% R-3
6 C $160,000 8.2% 81.8% R-3
7 B $140,000 7.2% 89.0% R-2
8 J $95,000 4.9% 93.9% R-3
9 D $90,000 4.6% 98.5% R-2
10 H $30,000 1.5% 100% R-3
Total $1,955,000 100%

The three measures

Top-one share = 450,000 ÷ 1,955,000 = 23.0% Top-three share = (450 + 310 + 270) ÷ 1,955 = 52.7% Top-five share = 73.7% Count to half = 3 (cumulative first reaches 50% at E) Herfindahl = 0.230² + 0.159² + 0.138² + 0.107² + 0.102² + 0.082² + 0.072² + 0.049² + 0.046² + 0.015² = 0.0529 + 0.0253 + 0.0190 + 0.0114 + 0.0104 + 0.0067 + 0.0052 + 0.0024 + 0.0021 + 0.0002 = 0.136

Last year, same customers

Measure Last year This year
Top-one 19.0% 23.0%
Top-three 44.1% 52.7%
Count to half 4 3
Herfindahl 0.118 0.136
Customers 10 10

Every measure rose with the same ten customers. F and E grew; the rest did not. That is the finding, before any level is judged.

The per-rep split

Rep Customers Revenue Top-one share of rep's book Count to half
R-1 F, E $720,000 62.5% 1
R-2 A, I, B, D $750,000 41.3% 2
R-3 G, C, J, H $485,000 41.2% 2

Rep R-1's book is two customers and the larger is nearly two thirds of it. The company table said 23 percent.

The identity

Σ customers = $1,955,000 = ledger Σ reps = 720 + 750 + 485 = $1,955,000

Where it goes wrong, even at ten

Level read first. "Is 52.7 high?" The rise from 44.1 is the finding.

Duplicate customer. If A were also recorded as A2 with $100,000 split off, top-one would read 21 percent and concentration would look lower than it is; the identity would catch a total above the ledger.

Company view only. R-1's one relationship invisible.

Index read alone. 0.136 means little; the rise from 0.118 does.

From ten to ten thousand

The same sorted list, per company, per segment and per rep, trended, with the roll-up checked. Covirage runs it on the ledger every quarter. The customer concentration guide covers the measures, and the concentration reading guide covers reading them at scale.

Questions people ask

Why compute three measures?

Because each is read by someone different. Top-n share is what a board reads; the index is what trends cleanly; count to half is what a sales leader acts on, because it is a number of relationships. On ten customers all three take a minute.

What does the index mean?

The sum of squared shares. Ten equal customers give 0.10; one customer with everything gives 1.00. It is sensitive to the largest, which is why it trends well. The level is hard to read alone; the direction is not.

Why the per-rep split?

Because the company's largest customer at 22 percent is one rep's at 71 percent. That rep's number is one relationship, and the company-level table cannot show it.