What a brand book cannot say

A brand is a set of decisions that a small group of people made together, usually over months, usually with more argument than anyone admits afterwards. What this company is. What it refuses. What it should feel like to encounter it.

The output of that process is then handed to people who were not present, and who will apply it for years under conditions the original group never anticipated. The document exists to make that possible.

Be precise about what is being attempted here, because the difficulty is built into the task rather than a matter of effort.

We know more than we can tell

Michael Polanyi's formulation is the starting point: we can know things we cannot fully articulate.1 A person who can reliably recognise a face, judge a proportion, or tell that a sentence is subtly wrong is not applying rules they could write down. Polanyi's point was not that such knowledge is mystical. It is that its structure resists complete expression.

Nonaka and Takeuchi built a theory of the firm on the consequences.2 Their argument is that organisational knowledge is created by conversion between tacit and explicit forms, and that the conversion from tacit to explicit, which they call externalisation, is the difficult and valuable step. Crucially, they found that documents are among the weakest instruments for it. What actually transfers tacit knowledge is shared practice: working alongside someone, watching a judgment being made, being corrected.

A brand guidelines document is an attempt at externalisation under unusually bad conditions. It must convert the tacit judgment of a small group into explicit form, then be re-internalised as judgment by a stranger, alone, at speed, without access to anyone who was there.

Written that way, the surprising thing is not that guidelines underperform. It is that anyone expected otherwise.

What the files say

The evidence is easier to read than the theory.

Going through a decade of one studio's brand systems, the pattern is consistent: the documents had not been opened since the week they were delivered, and the working files beside them were in constant use.

In one identity the numbers state the case starkly. Four colour styles, and roughly five thousand components.

That imbalance is a map of where effort went, and it is not a filing error. Nobody wrote the colours down carefully, because colour feels like the easy part and a hex code looks self-explanatory. The components were built properly, because components are obviously hard, and building them properly meant building them to be taken and used by people who would never meet anyone involved in making them.

That identity carried its company from seed to a Series A with no redesign.

At a different company, a count of the colours in the working files reached 3,080. Eight of them were the brand. Nobody decided that. It accumulated one defensible afternoon at a time, and every single one of those decisions was reasonable on the day somebody made it.

Both findings describe the same thing. Where a decision had been embedded in the object people actually used, it held. Where it had been written down and left to be applied, it did not.

Two ways to move a decision

This suggests a distinction that guidelines practice mostly ignores.

Externalising a decision means writing it down so that someone else can read it, understand it and apply it correctly. This requires the reader to acquire the judgment, at the moment of use, from text. It is the mechanism a document relies on and it is the expensive one.

Embedding a decision means building it into the thing the person is already using, so that the correct outcome requires no judgment at all. A named style, a component, a template, a token, a default. The decision is not communicated. It is simply present.

Most of what a brand document contains could be embedded instead, and historically was not, because embedding was expensive and writing was cheap. A document could be produced by the people who made the decisions. Embedding required engineering.

That constraint has largely dissolved. Design tokens, component libraries, shared template systems and the generators built on top of them mean that a very large share of what used to be written down can now be built in. This is the practical change worth noticing, and it is more consequential than any development in how guidelines are designed.

So the question to ask when specifying brand work has changed. It is no longer how thoroughly the decisions are documented. It is what proportion of them a person can get right without reading anything.

The instruction and the execution

Sol LeWitt described this arrangement before any of the tooling existed. His formulation was that the idea becomes a machine that makes the art.3 His wall drawings are instruction sets, executed by people he never met, in rooms he never entered, often decades later. The execution is delegated entirely and the work survives, because what was specified was sufficient and the specification travelled.

Two things in that practice are directly instructive.

The instructions are precise about what matters and silent about what does not. They do not attempt to transfer LeWitt's taste in general. They specify the conditions under which a correct instance can be produced by somebody else.

And the work is designed on the assumption that it will be made by strangers. This is not a compromise or a concession to scale. It is the premise. A brand handed over to an organisation is in exactly that position from the day it ships, and is usually designed as though it were not.

What cannot be embedded

There is a limit, and it is the reason this is not simply an argument for better tooling.

An embedded system holds the decisions that have already been made. It cannot make the next one.

When a company moves somewhere its system did not anticipate, a new market, a harder audience, a category that has shifted underneath it, the templates produce something technically correct and slightly dead. Recognising that, and saying why, remains a human act, and nothing currently available changes this. Polanyi's problem is untouched: the judgment needed is the kind that resists being written down.

So the design question is which decisions belong in which place, and that allocation is most of the work.

What a leader should specify

Ask what proportion of the brand can be applied correctly without reading anything. This is a better test of a brand system than any review of the document, and it can be run in an afternoon. Give three people who make things a real task and watch what they reach for.

Commission the working system, not only the document. The deliverable that protects the investment is the file, the components, the templates and the tokens. A document as well, by all means, but as a companion rather than as the asset.

Separate the codifiable from the judgment explicitly. Colour values, spacing, type scales, logo behaviour and naming can be embedded and should be. Tone in an unfamiliar context, what to do when a campaign needs to break the system, whether something is any good: these need a person, and the organisation should know which person.

Name who holds the judgment, internally. Every functioning brand has one or two people whose reading is trusted when the system runs out. If that is not acknowledged, it happens anyway and informally, which means it stops when they leave.

Test on strangers, not on the people who made it. The people who were in the room cannot evaluate a system for transferability, because they carry the tacit knowledge it is supposed to replace. Their judgment about whether the document is clear is systematically unreliable, and that is a property of the situation rather than a criticism of them.

Notes

1 Michael Polanyi, The Tacit Dimension (Garden City, NY: Doubleday, 1966).

2 Ikujiro Nonaka and Hirotaka Takeuchi, The Knowledge-Creating Company: How Japanese Companies Create the Dynamics of Innovation (New York: Oxford University Press, 1995).

3 Sol LeWitt, "Paragraphs on Conceptual Art," Artforum 5, no. 10 (June 1967): 79–83.