We started with a tempting idea: that the music a market plays might predict where its spending is heading. There is peer-reviewed evidence that music valence tracks sentiment, so it was worth testing properly.
So we tested it properly, and pre-registered each test before running it. A country-level forecast. A brand and category backtest. A powered nowcast across France and Germany. And, after we found a cleaner domestic-only signal, a re-test on that. Four tests. All four failed. On the broader outcome, adding music actually made the forecast worse.
We published every result, with the pre-registration and the code, including the failures. Then we changed the product rather than the data. Cadence is a descriptive read of where a market’s attention, mood and watching are moving right now. It is not a forecast, and we do not imply one.
Why show the dead ends. Because in a category built on confident guesses, the working is the trust. An admitted gap earns more confidence than a confident guess, and every figure in a Cadence read names its public source and date. The discipline is the product.