Constellations, patent pending

The Go Live Cliff · Process

Brand Guidelines Are a Hypothesis

A guideline document records what a room decided a brand should look like, once. It was almost never tested against the audience it was written for. Here is how to check it.

Quick answer

A brand guideline document records what a group of people decided a brand should look like at one moment. It's an argument, and it was almost never tested against the audience it was written for. Over the following years the work drifts from the guide, and every drift gets treated as a compliance failure. Some of it is. Some of it is the organization correcting toward something the guide got wrong. The way to tell the difference is to test representative work against the people closest to the brand and see which pieces read as the brand and which don't, then look at what the winners have in common.

What a guideline document actually is

Somebody made a set of decisions. Palette, type system, photographic register, logo behavior, tone. Those decisions came from a creative team applying judgment to what they understood about the business and its market, then a group of stakeholders approving it.

That's a hypothesis about how the brand should be read. It might be an extremely good one. It's still a hypothesis, and in most organizations it was never checked against anyone outside the approval room before it became law.

Then it becomes law, and everything gets measured against it. Which is fine until the measurement starts producing results nobody knows how to interpret.

The drift problem

Run a standard brand audit and you get an inventory. Website templates, deck templates, print material, campaign assets, everything produced over some period. Each item gets checked against the guide and marked on or off.

The inventory is useful. What it can't tell you is what to do about it.

Say a third of the work has drifted. There are two readings and they lead opposite directions:

  • The guide is right and execution has degraded. Pull the templates back in, retrain, tighten governance.
  • The guide is partly wrong and the organization has been correcting toward something that works better in practice. Update the guide.

Figure 1 · One inventory, two readings

An audit inventory grid with a third of the tiles marked off-guide, forking into two opposite readings On the left, a grid of twenty-four tiles standing for the audited work, eight of them outlined in red as off-guide. Two arrows lead right to two cards: reading one, the guide is right and execution degraded, so retrain and tighten; reading two, the guide is partly wrong and the work corrected it, so update the guide. The audit inventory every item checked against the guide, on or off on-guide off-guide, a third of the work READING ONE The guide is right. Execution degraded. Pull the templates back in, retrain, tighten governance. READING TWO The guide is partly wrong. The work corrected it. The organization found what works in practice. Update the guide. Same inventory, opposite instructions. The inventory can’t choose.
A compliance count supports both instructions equally. Most teams pick the first because it's the one the guide implies, not because anything in the inventory pointed there.

An inventory gives you no way to choose. Most teams default to the first reading because it's the one the guide implies, and they spend a year enforcing a rule the market stopped rewarding.

The check

The missing step is a read from people, and the people to start with are the ones closest to the brand. The method is audience perception mapping: an audience responds to the images instead of describing them, and the pattern shows where the group agrees, where it splits, and why.

Collect everything. The full inventory. This part is the same as any audit.

Sample it for range. Pull the pieces that represent the brand across as broad a spectrum as possible. On-guide and off-guide, old and recent, internal and customer-facing. The sample needs to contain the disagreement, or the test has nothing to find.

Put it to the organization. Send it to employees, and to the people who work with the brand daily. Ask which pieces feel most representative of the brand and its values, and which feel least. Both halves matter, because the rejected set is what makes the accepted set legible.

Look at the results side by side. What scored high, what scored low, and what stood out as clearly off. Contrast the two collections rather than admiring the winners in isolation.

Find the shared quality. What do the high scorers have in common that the low scorers don't. That common quality is the real guideline, whether or not it's written in the document.

Figure 2 · The check, in order

A five-step sequence from collecting the inventory to finding the shared quality Five numbered steps in a row: collect everything, sample for range, put it to people, hold side by side, find what's shared. The third step is marked as the read and the fifth as the real guideline. 1 Collect everything the full inventory, same as any audit 2 Sample for range on and off guide, old and new, inside and out 3 Put it to people employees and those who work with the brand daily 4 Hold side by side most representative against least 5 Find what’s shared what the high scorers have that the low ones don’t THE READ THE REAL GUIDELINE The sample has to contain the disagreement, or the test has nothing to find.
The first step is any audit. The third is the one audits skip, and the fifth is the only one that produces a guideline rather than a scorecard.

Keen Learning ran their existing material through this and found strong internal alignment on what looked appropriate and what looked off. The conclusion was to get back to the brand guidelines they'd drifted from. That's a legitimate outcome and it's the first reading above, arrived at with evidence rather than assumed.

Why employees are the right first audience

It seems like the wrong group. They're inside, they're biased, and nobody sells to them.

They have also seen more of the brand than anyone alive, including the customer, and they're the ones who will produce the next thousand assets. An employee sample tells you two things at once: whether the guide describes something the organization can consistently recognize, and whether the people executing it agree on what it is.

If 500 employees can't reliably sort their own material into on-brand and off-brand, the guide has a legibility problem, and no amount of governance fixes that.

It's also the cheapest test available. The audience is already assembled, already reachable, and already motivated.

Customer-facing questions come next, and they're a different test with a different sample.

What this changes about the document

A guide that's been checked reads differently. Instead of a list of rules with a rationale nobody remembers, you get a set of constraints with evidence attached, plus an explicit note about which parts were confirmed and which parts were the team's judgment call.

Figure 3 · Rules on authority, constraints on evidence

The same six guideline rules before and after the check, the second set tagged confirmed or judgment call Two panels. On the left, six rules, palette through layout, each with a grey bar and a chip reading rationale unknown. On the right, the same six rules with four tagged confirmed in green and two tagged judgment call in cyan. The guide before rules, enforced on authority Palette no rationale Type system no rationale Photography no rationale Logo behavior no rationale Tone no rationale Layout no rationale The guide after the check constraints with evidence attached, and a note on which is which Palette confirmed Type system confirmed Photography judgment call Logo behavior confirmed Tone judgment call Layout confirmed Confirmed constraints need strong evidence to move. Judgment calls stay open, and everyone knows which is which.
The rules don't change much. What changes is that each one now says how it was earned, and that decides how hard it is to move later.

That distinction earns its keep later. When someone proposes a change, the conversation is about which category the rule is in. Confirmed constraints need strong evidence to move. Judgment calls are open for reconsideration, and everyone knows which is which.

It also gives whoever holds the brand something to stand on. Enforcing a guide on authority is exhausting and it makes the brand person the obstacle in every project. Enforcing it on a documented read from 500 colleagues is a different conversation, and it's one the brand person wins without having to be the difficult one.

When to run it

Three moments where the check pays for itself:

Before a refresh, not after. Once a direction is in flight the audit becomes justification. Run it while the question is still open and it becomes input.

When the guide is three to five years old. Long enough for the market to move and for drift to accumulate, early enough that correcting is cheap.

When compliance keeps failing in the same place. If every team drifts on the same rule, that rule is probably wrong or unusable. Repeated failure in one spot is a signal about the guide, not about the teams.

Frequently asked questions

What should a brand audit measure beyond compliance?

Whether the guidelines still describe something the audience and the organization recognize as the brand. Compliance measures conformity to the document. It says nothing about whether the document is right.

How do you know if brand guidelines need updating or enforcing?

Test representative work with the people closest to the brand. If the pieces they identify as most representative are on-guide, enforce. If the off-guide work consistently reads as more representative, the guide has drifted from what the brand has become.

Who should evaluate brand materials in an audit?

Start with employees, because they have the widest exposure to the brand and they produce the next round of work. Customer-facing perception is a separate test with a different sample.

How much material should be in the sample?

Enough to cover the range, including the work you suspect is off. A sample containing only compliant material can only confirm what you already believe.