The idea in one line. A forecast without a named basis is an opinion with a logo. A trend bureau report, a platform's yearly predictions, an industry ranking — every one of them is an object of analysis, never a ground for a conclusion.
Why it matters. Bureaus sell certainty. Their reports are written to be quoted, not to be checked, and a student who writes "the bureau says" has not done analysis — they have outsourced it. Nobody in a buying office pays for repeating a bureau; they pay for knowing whether the bureau is right this time.
The audit — five questions, in this order.
1. Basis. What is the forecast built on: named data (searches, sales, runway counts, a panel with size and dates) or narrative (mood, zeitgeist, "we are seeing")? If the data is not named, the claim is narrative. Narrative can be right; it cannot be checked.
2. Horizon. When exactly is this supposed to happen — which season, which year? A forecast with no date cannot fail, and what cannot fail cannot be trusted.
3. Falsifiability. Write the sentence that would prove this forecast wrong. If you cannot write it, neither could the author.
4. Track record. Open the same bureau's report from two years ago. Which calls landed, which quietly vanished? Bureaus do not publish this scorecard; you keep it.
5. Interest. Who pays for the report and what does that buyer want to hear? A colour of the year sold to paint and textile mills is a commercial product, not a discovery.
How to use the verdict. A forecast that names its basis, dates its horizon and can fail is a signal like any other — score it with Signal Strength and put it on the evidence board with its provenance. A forecast that fails questions 1–3 is taste at industrial scale: note it as context, never as evidence. In both cases you write down what you checked, not what the bureau said.
The trap. Reports are seductive because they are well designed and confident. Confidence is a design choice. Your job is to find the number under the adjective.
Practice. Take one page of any current trend report. Underline every claim. Mark each one: basis named? dated? falsifiable? Count how many survive all three. That number is the page's real content — and your first line when you present it.