Every operator I have watched buy a platform negotiates the same three things. Price per terminal, contract length, support tier. Two weeks on those three numbers, taken to the mat.
Then the operator says the operation does a few things differently and asks whether the platform can handle it. The vendor says yes. Yes to the exception, yes to the custom step, yes to reproducing the way the operation already does the thing. The operator hears a partner who finally understands the business.
That yes is the most expensive line in the agreement, and it is the only line nobody priced.
It gets monetized twice. First as the implementation and configuration work. Then as the remediation required when the custom build sits on a surface the vendor had already scheduled for deprecation. The two charges arrive years apart, under different names, usually with different people on both sides of the table, and nothing in your own books connects them. So you pay both halves of one trade and never see it as one transaction.
The uncomfortable part is that nobody has to intend any of it. The group inside the vendor that says yes is not the group that inherits the yes. The motion that closes is insulated from the cost of closing. Which is why the read cannot be built on intent, and has to be built on direction of travel instead: does the vendor’s revenue go up or down if you take the standard build.
Underneath all of it is one piece of arithmetic the industry stays quiet about. The same differentiation promise is made to every operator in the category. A capability sold to the whole market cannot differentiate anyone inside it.
The inputs converged, and not because operators failed. A handful of platforms carry the category. Broadline distribution consolidated, and your product list came off the same truck as the operation two blocks over. The labor pool is one pool, and those people move between you freely. Every input-side edge that does exist is thin, copyable, and sitting on a clock somebody else controls.
So the operator’s real question is not how to be different at every layer. It is which layers to stop competing on entirely, on purpose, and where the capacity that frees up goes instead.
That posture has a name now. Ordinary By Design: the deliberate decision to be unremarkable on a named surface, meeting the standard in full, because that surface does not differentiate the operation and the capacity it would consume is needed somewhere that does. Three conditions, and two out of three is just drift with better vocabulary. The surface named and the reason on the record. The standard met in full rather than quietly undercut. And the freed capacity reallocated somewhere you can point to.
The Guest decides which surfaces qualify, not you. A surface is non-differentiating if no Guest has ever chosen you or returned because of it. That is answerable, and running it will move more items off your differentiation list than you expect.
What is left when the purchasable layers are conceded is the part nobody can order: the arrangement you chose, the cast you built, and the relationships they produce on the stage.
Read the prosecution: https://hacksterism.jeffreysummers.com/the-yes-is-the-product
Read the architecture: https://physics.jeffreysummers.com/differentiation-was-never-for-sale
One move this week. Take every technology exception, custom step, and special configuration your operation runs, and put one column next to it: name a Guest who chose you or came back because of this. The items with a name stay. Everything else is a preference you have been paying to hard-code.
— Jeffrey