Every performance team eventually produces an ad that works and should not exist. It is louder than the brand allows. The headline overpromises slightly. The colours are not in the palette, or they are in the palette but used at an intensity the guidelines describe as "sparingly." It might be a screen recording shot on a phone by someone in ops. And it is beating the carefully made, fully compliant creative by a margin nobody can argue with.
So a meeting happens. The brand side says it is off-brand and has to come down. The performance side says it is the only thing working and taking it down means missing the number. Both are correct, which is why the meeting resolves nothing and the ad usually keeps running until someone senior gets annoyed enough to kill it.
The reason this argument is unresolvable as stated is that both sides are answering a question nobody asked out loud: is this a violation, or is this information? Those have opposite responses. Almost every published treatment of this collapses into "brand and performance are not really opposed, do both", which is true, useless, and does not tell you whether to pause the ad on Monday.
Here is the way to actually decide it, in the order the decisions have to be made.
First: separate the three things "off-brand" is being used to mean
Before anything else, work out which of three completely different problems you have, because the word "off-brand" is covering all of them and they have different answers.
1. It breaks a design rule. Wrong font, wrong colour ratio, logo too small, the layout the guidelines call for is not the layout used. Nothing untrue is being said. The ad simply does not look like the system.
2. It breaks a voice or positioning rule. The ad promises speed when you position on thoroughness. It leads on cheapness when you have spent two years moving upmarket. It uses urgency language your brand has deliberately avoided. Nothing is inaccurate, but the ad is selling a different company than the one you are trying to build.
3. It says something that is not true, or not defensible. An implied guarantee. A comparison you cannot substantiate. A claim about outcomes that the product does not reliably produce. This is the FTC endorsement / substantiation category, not a brand debate.
These get bundled together in the meeting and they must not be. The third category is not a brand disagreement at all, it is a compliance and honesty problem, and it comes down immediately regardless of performance. That decision requires no deliberation and no committee. Get it out of the conversation first, because as long as it is in the room, the brand side is arguing from the strongest possible position about the weakest possible instance, and the performance side is defending something indefensible.
Once category three is removed, you usually find you are arguing about category one. And category one, honestly assessed, is the one where the guideline is most often the thing that is wrong.
The question that actually decides it: what is it a promise about?
The test is not "does this match the guidelines." It is "does this ad create an expectation the rest of the business will fail to meet?"
That reframing is the whole thing, and it works because it converts an aesthetic dispute into an operational one that has a checkable answer.
A brand is not a colour palette. It is the accumulated set of expectations a person has about what dealing with you is like (Ehrenberg-Bass work on distinctive brand assets makes the same point from the other direction), and the thing that damages it is not inconsistent typography, it is the gap between what you led someone to expect and what they got. An ad that looks different from your other ads costs you very little. An ad that makes someone expect same-day onboarding when onboarding takes nine days costs you a refund, a support ticket, a review, and the next customer that review talks out of buying.
So take the winning ad and ask, concretely:
| Ask | If the answer is yes |
|---|---|
| Does it promise a speed, price or outcome the business does not reliably deliver? | Kill it. This is category three wearing category one's clothes. |
| Does it attract a buyer you cannot serve well? | Kill it, and look at what it reveals about demand. |
| Does it contradict what the sales conversation says two steps later? | Fix the mismatch, it is not obvious the ad is the wrong end to fix. |
| Does it merely look unlike your other ads? | Let it run, and read on. |
That last row is where most of these land. And the honest answer, when an ad breaks nothing but the visual system and outperforms everything you made properly, is that your guidelines just failed a test with real money attached, and the test result is data.
Why the guideline is so often the thing that is wrong
Most brand guidelines were written to solve a coordination problem, not a persuasion problem, and are being applied to a job they were never designed for.
Think about how the document came to exist. Someone, often an agency, often two or three years ago, was asked to make the company look consistent across a website, a deck, a business card and a trade stand. That is a real problem and the document solved it. What that document was not doing was predicting which combination of image, headline and offer would persuade a cold audience on a nine-inch screen in a feed, against competing content, in under a second.
Those are different disciplines. The guideline is an answer to "how do we look like one company." The ad is an answer to "how do we get a stranger to stop." Requiring the second to be produced entirely inside constraints written for the first, by people who were not thinking about the second, guarantees a permanent tax on performance that nobody ever measures.
There is also a structural reason the tax stays invisible: the cost of an over-restrictive guideline never appears in a report. The ad that was never made, the angle that was ruled out in a review, the format nobody proposed because it obviously would not get approved, none of that shows up anywhere. Whereas the cost of an off-brand ad is loud, visible, and gets a meeting. The feedback loop is asymmetric, so guidelines ratchet tighter over time whether or not tightening helps. This is the same failure that turns brand documents into shelfware generally, which is the argument in a brand kit nobody follows is just a PDF.
The decision rule
Run the winner for a fixed window, decide from the read, and change the guideline rather than repeat the argument.
Concretely:
1. Keep it running while you decide. Unless it is category three, pausing costs real money for a decision that will take a week. A brand does not get damaged in seven days by an ad with the wrong font. Not deciding is a decision; make it the one that costs less.
2. Give it a defined window and a defined read. One buying cycle, or enough conversions that the difference is not noise, whichever is longer. Write down before the window starts what result would make you keep it, because otherwise the read gets argued afterwards by whoever is more senior.
3. Read further down the funnel than the click. This is the step that most often changes the answer and the one most often skipped. An off-brand ad frequently wins on click-through and loses on everything after it, the traffic is worse, the refund rate is higher, the support load is heavier, the customer is worse-fitting. That is not a brand argument, it is a unit economics argument, and it is far more persuasive in the room than an appeal to consistency. Judge it on contribution after refunds and servicing cost, not on cost per acquisition. If you do not have that view assembled, the sheet that tells you whether your marketing makes money is where to start, because you cannot settle this argument without it.
4. If it wins on the full-funnel read, change the guideline. Not "make an exception." Exceptions are how guidelines rot, you end up with a document everyone knows is routinely ignored, which is worse for consistency than a document with a wider boundary that people actually follow. Amend the document: name the channel, name the permitted latitude, name what stays fixed.
5. If it loses on the full-funnel read, you have won the argument properly. And you now have a reason the performance team accepts, which the guidelines alone were never going to give you.
What should never move, and what should
The way to make this decision fast every time is to have decided in advance which parts of the brand are structural and which are stylistic.
Most brand systems do not make this distinction, which is why every disagreement escalates. A useful system has three tiers:
Fixed, never varies, in any channel. The name. The logo and its clear space. The legal and claim boundaries. The promise you make about what the product does. The things that, if inconsistent, mean people cannot tell it is you, or mean you have misled someone. This tier should be short. If it runs past a page, it is not a fixed tier, it is a wish list.
Fixed in owned space, flexible in bought space. Typography, layout system, colour ratios, photographic treatment. Your website, your product, your invoices, your onboarding emails, the places someone experiences you deliberately, hold the system. Paid social, where you are interrupting someone who did not ask for you, does not have to.
Deliberately variable. Headline structure, hook, format, tone within a stated range, offer framing. This is the working surface. Nobody should need approval to change it.
The reason to write this down before you need it is that in the meeting, everything feels tier one. With the tiers written, "the colours are off" is visibly a tier-two issue in a tier-two channel and takes ninety seconds instead of two weeks.
One useful sharpening question when you are drafting the tiers: if a customer saw this and the logo were covered, would they still know it was us, and would they mind that they could not tell? The first half is the recognition test. The second half is the one that matters, and the answer for a feed ad is very often no. Recognition is not free, but on an ad seen once by someone who has never heard of you, it is worth less than most brand documents assume.
Check that it is actually winning before you rebuild the brand around it
Confirm the win is real in the business, not just large in the report, because off-brand creative has a specific tendency to look better than it is.
This is the check that should happen between step three and step four, and it is worth its own section because the failure mode is subtle. A loud, high-contrast, urgency-heavy ad does not just persuade differently, it gets clicked differently, and some of those clicks are people who were going to buy anyway. Bold creative on a broad audience harvests existing demand efficiently. Your platform report will credit it for that demand. Nothing in the dashboard distinguishes "created a customer" from "was the last thing a customer saw."
So before you amend a brand system on the strength of a number, rule out the three ways this number flatters:
It is claiming conversions it did not cause. The clearest tell is that the winner's apparent gain does not show up in total revenue. If the ad reports a large lift and the business is flat, the lift is reallocation, not incrementality. The reliable read here is a holdout or a geo split, not a platform report, and the window the platform uses to make that claim is a setting somebody chose, which is the argument in attribution windows are a commercial term, not a technical setting.
It is winning on a broader audience, not better creative. Check whether the off-brand ad and the compliant ad actually ran against comparable audiences with comparable budgets. Frequently the winner was given the easier segment, or picked it up automatically because the platform's optimiser found cheap inventory. That is a targeting result being read as a creative result.
It is early in its life. New creative gets a honeymoon, from the algorithm, and from an audience that has not seen it yet. An ad judged in week one against creative in week six is not being judged fairly. Compare at equal frequency, or wait.
If the win survives all three, it is real and you should act on it decisively. If it does not, you have avoided rewriting a brand system on the basis of a reporting artefact, and you have given the brand side a genuine answer rather than a veto, which is what makes the next one of these conversations shorter.
When you do not control the creative at all
Set the boundaries upstream when the ad is made by a creator or a partner, because after it is live your only options are removal or acceptance.
Creator and affiliate-made ads are where this problem gets genuinely harder, and it is worth treating separately. The economics that make creator content work are the same economics that make it off-brand: it performs because it sounds like the creator and not like you, and the moment you brief the brand system onto it, it stops working. Meanwhile the compliance exposure is larger, not smaller, because someone outside your business is now making claims on your behalf to an audience that trusts them.
The workable arrangement inverts the usual one. Instead of approving executions, you set:
- A claims boundary, the specific things that may never be said, with the substantiation you do have for the things that may. This is category three, delegated.
- A short must-include list, the name pronounced correctly, the offer stated accurately, the required disclosure. Genuinely short.
- An explicit grant of latitude on everything else, in writing, so nobody wastes a cycle guessing.
- A pre-agreed takedown route with a stated response time, so a real problem comes down in hours rather than becoming a negotiation.
The part people underestimate is the last one. You will eventually need something removed, and whether that is a phone call or a two-week standoff was decided when the agreement was signed, not when the problem appeared. Get the takedown clause in early, along with the rest of the commercial terms worth settling before the volume arrives, that is the same lesson as what an affiliate agreement has to say before you need it.
The thing the off-brand winner is actually telling you
Treat a winning off-brand ad as a market signal about your positioning, not just a creative outcome, that is where the real value is.
Ask why it won. Not "why did this creative beat that creative," but what proposition it put in front of people that your compliant ads do not. A few patterns recur:
- It said the plain thing. Brand language drifts toward abstraction, "partner," "solutions," "empowering." The scrappy ad said what the product does in the words customers use. The lesson is not about design at all; it is that your positioning language has drifted from your buyers' language.
- It led with the objection. Compliant ads lead with the benefit. The winner led with the thing people are actually worried about, the price, the switching cost, the time it takes.
- It looked like a person, not a company. A screen recording, a face, a handwritten note. This one wins in feeds so reliably that it is worth treating as a channel property rather than a creative accident.
- It was specific where the brand is careful. Careful language is vague language. Specific claims convert, which is exactly why they need the category-three check first.
Any one of those is worth more than the ad. If your winning ad works because it says the plain thing, that is a finding about your website, your sales deck, your positioning and your onboarding, not just about your ad account. Feed it back. The audit worth running afterwards is whether every other surface is still speaking the abstract version, which is the argument in your brand is whatever your worst touchpoint is.
Who decides, and how to stop having this meeting
Name a single decision owner for this class of conflict before it arises, and give them the full-funnel number rather than the click number.
The reason this argument recurs is almost never that people disagree about brand. It is that the decision has no owner, so it defaults to whoever escalates hardest, which is a terrible allocation mechanism and produces inconsistent outcomes that then get cited as precedent in the next argument.
What works:
- One owner for creative that runs in bought media. Usually whoever is accountable for the revenue number, with a stated obligation to apply the tier-one and category-three checks.
- A standing review with a fixed agenda, not an approval queue. Approval queues are where the tax accumulates, because the cost of everything not proposed is invisible. A weekly look at what ran, what won, and what the winners suggest about positioning is a better instrument than a gate.
- A written record of the amendments. Every time an off-brand winner leads to a guideline change, note the date, the evidence and the change. Six months later that record is the most valuable brand document you have, because it is the only part of the system that was tested against real behaviour rather than asserted in a workshop.
The underlying point is that a brand is not protected by refusing to change it. It is protected by knowing which parts cannot change, being genuinely relaxed about the rest, and treating every ad that wins by breaking a rule as a question about whether that rule was ever earning its keep. Guidelines that never change are not evidence of a strong brand. They are usually evidence that nobody has been testing them, and a rule that has never been tested is not a standard, it is just a habit with a PDF around it.
