How to Tell If You Have a People Problem or a System Problem
Do you have a people problem or a system problem?
Ask one question: could a capable, motivated person in that seat have got it right with the information and authority they actually had? If no, the design is at fault. If yes, and it went wrong twice, you are looking at the person.
Who is this for?
You run a business that already sells. The revenue is real, the clients are real, and at least one other person touches the work, whether employed, freelance, or a long standing collaborator.
And something keeps going wrong in the same place. You have had the conversation about it, maybe twice. The person nodded, meant it, and three weeks later it happened again. You are quietly wondering whether you hired the wrong person or built the wrong thing, with no way of telling which.
Key takeaways
- A missed deadline or a dropped client is produced by both causes, so the symptom tells you nothing about which you have.
- Start from information and authority. Someone who lacked either could not have succeeded, whatever their qualities.
- Three signals point at design: the problem survives a replacement, repeats on a schedule, or appears across several people at once.
- One signal points at a person: everything was written, available, inside their authority, and it still went wrong twice.
- Deciding between a people problem or a system problem is worth ten minutes, because the two fixes share nothing and one wastes a hire.
- Write the answer down before acting. Diagnosing in the same breath as reacting is how founders replace people twice.
Table of contents
- Why one symptom has two causes
- The four questions that settle it
- The same missed deadline, two versions
- What Deming actually claimed, and what he did not
- Three signals that look like people and behave like design
- What to do once you know which one it is
- Frequently asked questions
- Recommended tools and resources
- Where to explore next
- Final summary
Why one symptom has two causes
A dropped client looks identical from the outside whatever produced it. So does a late delivery, a pricing mistake, an unanswered supplier.
That is the difficulty. You are reading an outcome and inferring a cause, and the outcome is silent about the cause. Founders resolve that silence with whatever is most visible, and the most visible thing in any business is a person. So a people problem or a system problem gets decided by who stands nearest.
I sat in a doctor's office last June for something that kept coming back. He reached for anti inflammatories without asking what was producing it. I remember thinking, in the chair: "Ça va traiter le symptôme, pas la cause. Dans 2-3 semaines ça reviendra." It will treat the symptom, and in two or three weeks it will be back.
It came back. And I have watched myself do the identical thing inside a business. Something breaks, I reach for whatever stops the pain fastest, and I am surprised three weeks later.
A replaced person is an anti inflammatory. It relieves something real. If the design produced the failure, the relief has an expiry date, and you spent a hire to buy it.
The four questions that settle it
Here is a demonstration you can run right now, and it takes about eight minutes.
Pick the last thing that went wrong. The most recent one, not the worst, because you still remember the detail. Open the actual thread. Then answer these four in writing.
One. Was the information available to them? Not in principle, somewhere in a shared drive. In the place they were actually looking, at the moment they had to decide.
Two. Was the decision inside their authority? Could they have said yes or no without asking you? If it needed your approval and you were unreachable, they did not fail. They waited, correctly.
Three. Was there a rule, or only a precedent? A rule has a threshold in it: "Refund in full under 48 hours, credit after that." A precedent is "we usually do what feels right", which is a rule only for whoever has been here five years.
Four. Had this been made explicit, once, in words? Not demonstrated, not implied by how you do it yourself. Said or written, so there was something to follow.
Four yeses and a repeat failure points at the person. Any no, and you have found the thing to fix.
Most founders are surprised on question three. They find a precedent where they were certain a rule lived.
The same missed deadline, two versions
An illustration rather than a case study. Same week, same person, same missed Thursday.
The design version. Your collaborator has the client's file and a deadline. On Tuesday the client sends an extra round of changes. Nobody has written down whether extra rounds move the deadline, or who may agree to that. She does the changes, because refusing feels like bad service, and Thursday passes. She did the considerate thing with what she had. The deadline was never protected by anything.
The person version. Same brief, and this time one line exists: "Extra rounds after the brief is signed move the date by the same number of working days. You can confirm that with the client." She has read it and used it before, in June. This time she does the changes, says nothing, and lets Thursday pass.
Identical outcome for the client. Completely different cause. In the first, a sentence was missing. In the second, a sentence existed, was understood, sat inside her authority, and went unused.
Only the second is about the person. Notice how little separates them: one line containing a number and a name.
What Deming actually claimed, and what he did not
W. Edwards Deming wrote the most quoted line in this territory, on page 315 of Out of the Crisis: he estimated that troubles and possibilities for improvement divide roughly 94 percent to the system, the responsibility of management, and 6 percent special.
Now the limits, which matter more than the number. Deming's own wording was "I should estimate that in my experience". It is a judgement from decades of industrial quality work, never a measurement, and no dataset was published behind it. The Deming Institute reproduces the quote with that framing intact. Practitioners have noted it gets cited as a law of organisations when it was offered as an estimate, and that it came from manufacturing lines rather than a four person service business with a founder in every conversation.
So do not carry the 94 into your business as a ratio. Carry the direction of the prior. When you genuinely cannot tell whether you have a people problem or a system problem, the thing you have not examined is the design, because the design is invisible and the person is standing right there.
Three signals that look like people and behave like design
Three patterns are nearly conclusive on their own.
It survives the replacement. You changed the person and the failure returned within a quarter. The seat produces it. This is the cleanest evidence available, and the most expensive, because you ran the experiment to get it.
It repeats on a schedule. Month end, the week after a launch, every August. Failures that keep an appointment are structural. A bad week does not recur on the first Monday of every month.
It shows up across several people at once. Two or three people, different temperaments, same error. The one thing they share is the instruction they were given.
And the counterpart, so this stays honest. A person problem has a signature too. It is narrow, specific to one individual in a seat where others did fine, and it persists after the rule has been written, read and confirmed. It is rarer than it feels at 6pm on a Friday.
What to do once you know which one it is
The two fixes share nothing, which is exactly why the diagnosis is worth eight minutes.
If it is the design, write one line. Not a process document, not a manual. One sentence with a threshold and a name, in the place the work actually happens. Then leave it for three weeks and watch whether the same question reaches you again. A question that stops arriving is a rule that works.
If it is the person, have one specific conversation about the specific rule, with the written rule in front of you both. Then decide inside a defined window rather than carrying it for another eight months.
And one caution worth more than the rest. Do not diagnose in the hour you discover the problem. That hour is when a design fault most reliably looks like a character fault, because you are annoyed and a person is in front of you. Write the four answers down the next morning. They change more often than founders expect.
Frequently asked questions
What is the difference between a people problem or a system problem?
A system problem means the work could not reliably be done right with the information, rules, and authority available in that seat. A people problem means it could have been, repeatedly, and was not.
Can it be both at the same time?
Often, and the order matters. Fix the design first, because a written rule makes the remaining behaviour visible. Address the person first and you cannot tell afterwards whether the improvement came from the conversation or from the clarity that arrived with it.
How many times does something need to repeat before it counts as a pattern?
Twice, under the same conditions, after the rule was made explicit. Once is an event. Founders tend to act on the first occurrence and ignore the fifth.
I replaced the person and the problem came back. What does that mean?
That the seat produces the failure rather than the occupant. It is the strongest single piece of evidence available, and the next step is a written rule for that seat rather than a third candidate.
Does this apply to freelancers and contractors too?
Yes, and more sharply. A freelancer has less context by construction, so they reach missing information faster than someone who absorbed your habits over two years. Contractor failures are design failures more often than owners assume.
What if I am the person who keeps dropping things?
Run the same four questions on yourself, answered as if about someone else. Founders usually find the information was scattered and no decision had a threshold. You occupy the same undesigned seat as everyone else, plus the burden of being the fallback for all the others.
Is a handover document enough to make it a rule?
Rarely. A handover lists what to do. A rule states what to decide, at what threshold, and who may decide it. Most handovers are long on tasks and silent on decisions, which is why the questions keep arriving.
How do I raise this without the person hearing an accusation?
Ask about the decision, never the intention. "When the extra round arrived, what did you have to go on?" gets you the actual answer. "Why didn't you flag it?" invites a defence and teaches you nothing about your design.
Could this be a training problem instead?
Training is part of the design. If someone was never shown, the instruction was missing, which sits on the same side as a missing rule.
Where does software fit into any of this?
After the diagnosis, never before. Software enforces a rule you have already written. Applied to an undecided situation it produces the same failure faster and with an audit trail, which is worse, because it now looks deliberate.
Recommended tools and resources
- A written decision list. A plain shared document holding every recurring decision in one line, with a threshold and an owner's name. This does more than any software here.
- W. Edwards Deming, Out of the Crisis. Source of the 94 percent estimate and the wider argument that management owns the system. Read it as experience, not data.
- Your own last four weeks of messages. The cheapest diagnostic material available and the only one specific to your business. Search your sent folder for questions you answered in under a minute.
- A three week review slot in the calendar. Written rules need one checkpoint to prove they work. Put it in when you write the rule.
- A single shared place where rules live. Which one matters far less than everybody knowing where to look before asking you.
Where to explore next
- What Is a Business Systems Audit?
- How to Know What to Fix First in Your Business
- How to Delegate Without Losing Control
- The 5 Decisions You Should Never Delegate
- 7 Signs Your Business Has Outgrown Its Systems
- Hire vs Automate: Which One Your Business Needs
- How to Document a Process Without a 40-Page Manual
Final summary
The symptom never tells you the cause. A dropped client, a missed Thursday, a supplier left waiting: each is produced as readily by a missing sentence as by the wrong person in the seat. Which is why a people problem or a system problem has to be answered deliberately, in writing, and preferably not on the day you discover it.
Four questions settle most cases. Was the information available where they were looking, was the decision inside their authority, was there a rule with a threshold, and had it been said once in words. Four yeses and a repeat tells you about the person. A single no tells you where to write one line. The second answer is more common than it feels, and cheaper to act on.
Ready to find out which one it is in your business?
If you already know which incident you would run the four questions on, that is the conversation worth having out loud. Bring it, and we will work out whether the seat or the occupant produced it.
