A supplier is not dropped for making a mistake. They are dropped for repeating it.

No small company is going to do everything in-house, and it does not need to. The choice was never between having suppliers and not having them — it is between having several and governing none, or having those same several with somebody conducting.

The symptom that nobody is conducting is always the same: each area has a supplier, each supplier has a version of the problem, and when something breaks in between, they all point at each other. The owner becomes the judge of a technical argument they have no way to judge.

Whoever conducts is not whoever plays

The fix is not to replace them all with one, nor to bring everything in-house:

IT does not carry out every area — it governs whoever does. Each area has its specialist (internal or external); the company has one conductor.

The conductor can be someone in-house, or hired. What they cannot be is one of the musicians. Whoever runs an area does not assess their own area — not out of dishonesty, but because nobody sees their own work from the outside.

And there is a practical detail that makes almost all the difference: the supplier receives work through the same queue as everyone else, with a defined priority — not through side requests from whoever thought of it. A supplier who quietly serves whoever asked first is not being helpful. They are building a relationship only they can see in full.

The yardstick, and the question nobody asks

Assessing a supplier usually turns into a conversation about price, because price is the only number that shows up on its own. There are four questions, and price is the third:

Delivery — do they deliver what was agreed, when it was agreed? Quality — do they fix it for good, or generate rework? Price — at renewal, does it still make sense against a market quote? Presence — do they answer when needed, or vanish during an incident?

The fourth is the one almost never asked, and the one that reveals most. A cheap, competent supplier who disappears on the day the operation stops is, in real terms, the most expensive of them all — except their cost never appears on an invoice.

The rule that prevents the two biggest mistakes

There are two opposite errors, and the second is far more common: switching too early, in the heat of a problem; and never switching, because switching is hard work.

Golden rule (from real practice): a supplier is not dropped for making a mistake; they are dropped for making it again after being warned. Repetition without correction is the last straw — always.

Mistakes are in the nature of the work, and whoever has never got anything wrong with you has probably not yet done anything difficult. What you watch is not the mistake: it is what happened after the warning. That turns an emotional decision into one with a date — there was a warning, there was a deadline, and the correction either came or it did not.

The behaviour that gets no second chance

There are four triggers for switching, and any one is enough: repetition without correction, a price out of line with the market and unexplained, a security risk ignored after warning, and this one:

hostage behaviour (black box, withholding access or information)

The first three are performance failures, and performance can be corrected. This one is of another nature: it is the supplier making themselves irreplaceable on purpose, holding on to the password, the access, the drawing of how the thing works. It is not that they are too good to be replaced. It is that they arranged things so that you cannot.

When that appears, the yardstick stops making sense — because you are no longer assessing. You are asking permission.

How you switch, and never on impulse

The part that usually goes wrong is this one, and it is procedure, not judgement: a formal conversation with a deadline to fix; quotes obtained and a replacement chosen before switching off the current one; and a planned handover — access revoked, knowledge transferred, the exit recorded.

The order matters more than the decision. Switching off before you have a replacement is what turns a justified change into two weeks of stopped operations — and what makes the owner swear, next time, that putting up with it is better.

The test

Pick your most important supplier and answer:

If they stopped answering tomorrow, how long until somebody else took over — and who holds the access today?

If the answer is "I don't know", the problem is no longer the supplier. It is that the decision to switch stopped being yours.