One customer is half our revenue. Do we chase another like them, or sell to everyone else?
By Greg Rosner
Founder of PitchKitchen · Author of StoryCraft for Disruptors
· 7 min read
TL;DR
When one customer is a large share of your revenue, the risk everyone tracks is financial and the risk nobody tracks is the message. Over time the homepage, the case studies and the roadmap all describe that account's problem, so every other buyer reads material written for somebody else. Call it the one-account story: a company story written from one customer's problem, which no other buyer recognizes. AI makes it heavier, because engines now summarize your public material to buyers you never meet. The work is to write the problem down in the market's words rather than one customer's, as a Magnetic Messaging Framework.
A CEO told me one customer pays for 47 percent of his company.
He wasn't worried about losing them. The contract runs three more years and the relationship is good. He was worried about the other half of the business, which had gone flat two years ago while the big account kept expanding.
We pulled up his homepage together. Every sentence on it was true. Every sentence was also about a workflow that exists at exactly one company, the one paying him 47 percent.
His case studies were that customer. His demo data was that customer. The last eleven things his team shipped were that customer's requests, in the order they were asked for.
Nothing there was dishonest. It was one customer's story wearing a company's name.
The question he asked me is the one this post answers. Does he go find another customer like that one, or does he build something the rest of the market can see itself in?
What is a one-account story?
He has a one-account story: a company story written from one customer's problem, which no other buyer recognizes.
It doesn't start as a mistake. It starts as a reward.
The big account showed up early and paid real money. They told you exactly what they needed, in their own words, and building it worked. Their vocabulary became your vocabulary, because theirs was the vocabulary that had been paid for.
Land a large customer early and you get handed a free script. The script works on one person.
Here's the part that makes it hard to catch. A one-account story doesn't read as generic. It reads sharp, specific and confident, which is why nobody on your team flags it. The page passes every clarity test you can run, and the only buyer who fully recognizes themselves already has a contract with you.
You're probably thinking specificity is the whole point, and you're right. The question is who it's specific to. Specific to a problem pulls in everyone who has that problem. Specific to an account pulls in that account.
Why is this worse now than it used to be?
Two years ago this cost you quietly. A prospect landed on your site, didn't see themselves, and left without telling you why.
Now the machine reads it first.
When a buyer asks ChatGPT or Claude what your company does, the engine builds its answer out of your public material. Your public material describes one customer's problem. The engine files you under that problem and repeats it, confidently, to every buyer who asks.
AI didn't create the one-account story. It started repeating it at scale, for free, to buyers you will never meet.
That's the change in the cost. The wrong impression used to sit on your website and wait. Now it travels.
It also compounds in the wrong direction. The longer the engines describe you as one account's vendor, the more of the right buyers read an answer about somebody else and move on. Every quarter you leave it, the engines get another quarter of material saying the same thing.
How do you tell if your biggest customer wrote your messaging?
Three tests. Each one runs off material you already have, and none of them takes longer than an hour.
- 1The Second-Customer Read. Open your homepage and read it as your second-biggest customer, not your biggest. Mark every sentence describing work you only do for the largest account. If more than a third of the page is marked, the page belongs to one customer.
- 2The Proof Origin Count. List the last ten pieces of proof you sent a prospect: case studies, reference calls, screenshots, outcome numbers, logos on a slide. Write the account name beside each one. Then count the distinct names. Companies in this position usually come back with two or three.
- 3The Roadmap Ledger. Take the last twelve things your team shipped and write the name of the customer who asked for each one. If a single name owns more than half the list, your product has been following one buyer and your story is describing where it went.
None of this says fire the customer. Keep them. They're paying for the runway you'll use to fix it.
What I see across companies in this position
The pattern is consistent enough that I ask about it early now.
A company lands a large account in year two or three. Revenue jumps. Nobody calls it a problem, because the number is good. Two years later the founder tells me growth outside that account is flat and the marketing team is getting blamed for it.
Public companies have a rule for this. Under SEC disclosure rules, a company has to name any customer accounting for 10 percent or more of consolidated revenue when losing that customer would materially hurt the business. Ten percent is where a regulator decides the market deserves a warning.
A private company at 40 percent has no such requirement, so nobody says it out loud, including the people running the company.
The revenue concentrates first. The message follows, quietly, and then holds the company in place long after the revenue is ready to move.
On the ground it shows up in three places:
- Sales can only sell confidently into one profile, and any call outside it feels like starting the company over.
- Marketing ships material the big account loves and nobody else ever asks about.
- The roadmap has no room for a bet without a named requester, so every release deepens the same relationship.
Read down your own artifacts and the same sentence keeps appearing, written five different ways:
| The artifact | What it says today | What a buyer outside that account concludes |
|---|---|---|
| Homepage hero | The workflow your biggest account runs | Built for companies like them, not for a company like us |
| Case studies | One implementation, told in that account's internal vocabulary | No evidence they've solved this in a situation resembling ours |
| The demo | Data, screens and edge cases from the anchor environment | We would be their first real attempt outside that one customer |
| The roadmap | Releases with a single named requester behind them | The product follows somebody else's priorities, and we'd be in line |
| The AI answer about you | A summary built from all four of the above | That's a vendor for that industry, and the search moves on |
The last row is the one that's new. The other four have always been true of concentrated companies. Only the fifth one runs without you, at volume, to buyers who never reach your site at all.
None of those look like a messaging problem from the inside. They look like a sales problem, a marketing problem and a product problem, which is why three different people get blamed and nothing changes. Ask what your company sounds like to a buyer who has never met your biggest customer, and the three problems turn into one.
A real example
Back to the CEO with the 47 percent customer.
We ran the three tests in an afternoon. The Second-Customer Read marked nine of the fourteen sentences on his homepage. The Proof Origin Count returned two account names across ten artifacts, and one of the two was a pilot that ended in 2024. The Roadmap Ledger had a single requester on eight of the last twelve releases.
He knew all three of those facts separately. He had never seen them in a row.
What changed wasn't the customer. It was who the material was written for. The problem that health system hired him to solve is a problem every multi-site employer with shift labor has. Same problem, different buildings.
Once that problem was written down in language that didn't depend on one org chart, the case study stopped being about a customer and started being about a situation a stranger could recognize. His biggest account went from being the story to being the proof.
[Greg: if you want a named engagement and hard after-numbers attached here, give me the two figures you're comfortable printing, pipeline outside the anchor account and the timeframe. June did not invent them.]
What this means for you
If one customer is a large share of your revenue, you don't have a concentration problem yet. You have a story problem that becomes a concentration problem later.
Going out to find a second account like the first one is the same bet twice.
The work is to separate the problem from the customer who taught it to you.
That's the job a Magnetic Messaging Framework does here. It writes the problem down in the market's words instead of one account's words, decides which buyers own that problem, names the villain they're already fighting, and locks the language everything downstream inherits: the homepage, the deck, the case studies, and the AI your team writes with. It matters because every one of those artifacts is currently teaching a machine who you're for.
Three things you can do this week:
- 1Run the Proof Origin Count. Ten artifacts, one account name each. Twenty minutes tells you whether you have one story or several.
- 2Write the problem your biggest customer hired you to solve without using their name, their industry, or their internal system names. If you can't get through it, that gap is the work.
- 3Take that problem statement to three customers who aren't the big one and ask whether it describes their situation. Change the words until it does.
Your biggest customer handed you a real problem. Other companies have it too. Go tell them.
Questions People Ask
FAQ
Is customer concentration a sales problem or a marketing problem?
It usually starts as neither. It starts as a story problem. When one account funds most of the business, the homepage, the case studies and the roadmap gradually get written from that account's problem, so new buyers read material built for somebody else and don't recognize themselves. Sales then struggles outside the one profile and marketing ships work nobody else asks about. Both are symptoms. The fix is to write the problem down in the market's language rather than one customer's.
How much of our revenue should come from one customer?
There's no single safe number, but there is a useful marker. SEC disclosure rules require a public company to name any customer accounting for 10 percent or more of consolidated revenue when losing that customer would materially hurt the business. That's the point at which regulators decide the market deserves a warning. Private companies carrying 30 or 40 percent have no such requirement, which is why the risk often goes unnamed internally for years.
How do we sell to new customers when all our proof comes from one account?
Change what the proof is about. A case study that reads as one company's implementation only convinces buyers who look like that company. The same story rewritten around the problem, the trigger that made it urgent, and the decision the buyer had to make will land with anyone who has that problem, regardless of industry or size. Count the distinct account names across your last ten proof artifacts first. If the answer is one or two, that's the constraint.
Should we turn down more work from our biggest customer?
Not usually. That revenue funds the time you need to fix the story. The change to make is upstream of the contract: stop letting one account's vocabulary set what the company publishes. Keep serving them, and write the underlying problem in words that a stranger with the same problem would use. Your biggest customer then becomes proof of a problem you solve rather than the definition of what you do.