The graveyard of performance-only brands

Somewhere there is a conference talk called “We cut our brand budget 40% and grew.” It has a good chart and a confident speaker, and the room nods along. There is no talk called “We cut our brand budget 40% and slowly died,” because the brand that did that is not at the conference. It is in the graveyard, and the graveyard does not send speakers.

This is one of the oldest traps in reading evidence, and Taleb spends a good part of two books on it1 . Survivorship bias Reasoning from the winners because they are the ones you can see, while the losers who ran the same play have dropped out of the sample. Study the survivors and any method looks like it works. : you study the cases that survived to be studied, and infer that their method works, without ever seeing the larger pile that ran the same method and vanished. Look at the funds that beat the market and momentum trading looks brilliant. Look at the ones that blew up using the same strategy, and you cannot, because they are gone. The brands that cut equity and grew are on stage. The ones that cut equity and faded are not, and they are not a rounding error.

Why a real survivor can honestly tell you the wrong thing#

Cutting equity works for a while, because you spend down the demand already built. The good years come first and the decline comes later, so a survivor gets the stage slot during the phase that looks like proof.

The strongest version of this trap is an honest person on stage with real numbers. Cutting equity does work for a while, because for a while there is a backlog of demand your brand already built, and you can harvest it cheaply while spending nothing to replace it. I laid that mechanism out in the first essay. The account grows and the graphs go up. On paper it looks like proof. The decline comes later, when the backlog runs dry, and it comes slowly enough that it rarely gets pinned on the decision that caused it. So the speaker on stage is not wrong about their own numbers. They are showing you a real good year that arrived before a bill that had not yet landed.

The speaker is not wrong about their own numbers. They are showing you a real good year that arrived before a bill that had not yet landed.

The sample is the whole problem#

Real brand, real data, real person on stage, and still no use as evidence, because the sample was selected on the outcome. The failures ran the same play and are not in the room.

Everything about the case study reads as rigour. It names a real brand, the numbers were audited, and the person presenting was in the room when it happened. What it does not have is a control group. To know whether cutting brand spend is a good idea, you would need the brands that made the same cut and did not survive, and those are the ones a conference cannot book and a case-study library will not publish. No vendor is going to volunteer them either. The sample was selected on the outcome, which means it can prove almost anything you point it at. It is a graph of the winners with the losers cropped out of frame.

Where this sits on the triangle#

The public case against funding Equity is built almost entirely from survivors. Ask for the graveyard and the case gets much thinner.

On the model, this is how the case against the corner that compounds gets made in public: from a sample of brands that cut it and lived, with the ones that cut it and died left out. The survivors are not lying. A room full of survivors just is not evidence about a decision, no matter how real each story in it is.

So the next time a case study proves that brand spend does not matter, ask one thing. Where is the graveyard? Name the brands that ran this exact play and are not here to present it. If the speaker cannot, they have not gone looking, and the target is being painted around the bullet holes after the fact.