Identity work is judged in a presentation and tested in production. The gap between those two moments is where most of it dies, and it dies in three predictable places.

The printer.

An identity approved on a screen has been approved under conditions it will never encounter again. The colour was rendered in RGB on a calibrated display. The logo was viewed at 400% on a white field. The packaging existed as a flat visual, not as flexographic printing on kraft corrugated running at speed, or as reverse print on film, or as four colours because the category cannot support six.

What follows is familiar. The colour cannot be held on board, so it shifts, and it shifts differently at each of the three printers used across three markets. A detail in the mark disappears below a certain size and is quietly removed by someone at the plant. The substrate specified turns out to cost more per unit than the margin allows, so a substitute is used, and the substitute changes everything the design was relying on.

Design that cannot be manufactured at the price the category supports is not a design. It is a proposal. A system built properly starts from the press, the substrate and the unit cost, and treats them as the material rather than as an obstacle to be discovered later.

The calendar.

The season does not wait for anyone. A promotion is confirmed on a Tuesday and needs artwork by Thursday. A market asks for a different declaration on the back panel. A customer wants their reference added to a case. These requests do not arrive at convenient moments and they do not stop.

If satisfying them requires the original designer to be available, the system will be bypassed the first week nobody answers the phone. Whoever is free will make the label. They will do a reasonable job with the file they can find, which may not be the current one. Twelve months later there are four versions of the identity in circulation and nobody made the decision that produced them.

A system survives the calendar only when the people inside the company can operate it without asking permission. That is a question of rules, files and decisions already taken — what is fixed, what can move, what the hierarchy is when text has to be added, where the assets live. Not taste. Taste does not scale and it is not available on a Thursday.

The next SKU.

The system was built around a range that existed at the time. Then a new format appears. A new variety. A line for a market that requires a different alphabet and a different set of mandatory statements. A pack size the original grid never anticipated.

If each of these requires the identity to be reinvented, what was delivered was a set of layouts, not a system. The test is whether someone can add a product that did not exist when the rules were written and have it look like it belonged there from the beginning. That capacity has to be designed deliberately. It rarely appears by accident, because the pressure during a project is always toward perfecting the products in front of you.

This is why the review question that gets asked in most presentation rooms is the wrong one. Whether the people around the table like it is close to irrelevant, and it is a question everyone present feels qualified to answer, which is why it dominates.

The question that predicts the outcome is whether it will still be intact in three years, across four printers, six formats, two new markets and the departure of the person who commissioned it.

Most identity work does not fail because it was bad. It fails because it was never asked to survive anything.