Noise is the default. Signal is the exception. A four-part test for knowing which one you are acting on — before you act.
The assumption most executives make
Information is reliable until proven otherwise. When data arrives — a report, a rate, a valuation, a headline — the default is to treat it as an accurate picture of reality and act accordingly.
What this framework argues instead
Noise is the default. Signal is the exception. The burden of proof runs in the other direction. A piece of information earns the right to drive a decision — it does not inherit it.
Why your information environment produces more noise than signal
1
Reporting lag
By the time a number is published, aggregated, and distributed, the underlying reality has already moved. The information is technically accurate and structurally misleading at the same time.
2
Incentive distortion
The people producing information have interests in how it is read. The information is not fabricated — it is framed. Framing is the most powerful form of noise because it looks exactly like signal.
3
Aggregation smoothing
Averages hide distributions. Any time an aggregate is used where a distribution would be more accurate, noise is being introduced structurally — and the picture it paints may be the opposite of the truth.
The four-part test
A piece of information is signal when it passes all four. Click each criterion to understand what it is testing and why it matters.
Test 01 — Falsifiable
What this test is looking for
If you cannot name a specific condition under which this piece of information would be wrong, it is not signal — it is narrative. Narratives can be compelling and useful, but they cannot drive a decision the same way a falsifiable claim can. The test: say out loud "this would be wrong if ___." If the blank stays empty, you are holding noise.
Example: "Consumer spending is resilient" is not falsifiable as stated. "Consumer spending by the top two income quintiles is resilient, while the bottom three are contracting" is falsifiable — you can name the condition that would make it wrong, and you can test it against the data.
Run it on real information
US unemployment
Consumer spending
SpaceX at $1.75T
US unemployment rate: 4.3%
Falsifiable
The 4.3% figure is technically accurate within its own definition. But the definition excludes 6.2 million people who want work and stopped looking, plus millions in involuntary part-time roles. The number is not wrong — the definition is doing the work.
Partial
Upstream
Unemployment is a lagging indicator — it reflects decisions already made by employers, not signals about what is coming. The leading signal is labor force participation rate and the transition rate into discouraged-worker status, both of which are deteriorating.
Fails
Incentive-clean
The BLS methodology is consistent and not politically manipulated in construction — but administrations have strong incentives to lead with the headline figure (U-3) rather than the broader measures (U-6: 8.1%). The framing is the distortion.
Partial
Distribution-stable
The aggregate rate hides a sharp bifurcation: unemployment among college-educated professionals is near historic lows while underemployment among workers without degrees is running well above pre-pandemic levels.
Fails
Verdict: noise dressed as signal. The 4.3% figure is useful for trend comparison over time but should not drive decisions about labor market health. The upstream signal is the U-5 and U-6 measures, the participation rate, and the transition rate into discouraged-worker status — none of which make the headline.
US consumer spending: "resilient"
Falsifiable
"Resilient" is a narrative frame, not a falsifiable claim. Retail sales hitting a record $752B in March 2026 is falsifiable. The word "resilient" applied to the whole consumer economy is not — it is a conclusion embedded in the framing.
Fails
Upstream
Retail sales are downstream of spending decisions already made. The upstream signal is the personal savings rate (collapsed to 4.0% from 6.2%) and the credit card delinquency transition rate — both indicating the fuel source for "resilient" spending is debt, not income growth.
Fails
Incentive-clean
Retailers, banks, and policy communications all benefit from a resilient-consumer narrative. The framing is produced by parties with interests in how it lands.
Fails
Distribution-stable
This is where the aggregate fully breaks down. The top 10% of households by wealth account for 49% of all consumer spending. The aggregate looks resilient because one segment is carrying the other four. The bottom 60% are actively contracting.
Fails
Verdict: pure noise. "Resilient consumer spending" fails all four tests. The signal is in the distribution: two economies running simultaneously, one of which is being used to describe both. Any strategy built on aggregate consumer resilience is built on a misleading foundation.
SpaceX IPO valuation: $1.75 trillion
Falsifiable
The valuation is falsifiable — Morningstar's independent DCF puts fair value at $780B, a $970B gap. You can name specific conditions under which the $1.75T figure would be wrong: if Starlink subscriber growth slows, if xAI does not achieve profitability, if Starship misses commercial deployment timelines.
Passes
Upstream
The IPO price is downstream — it reflects the demand that was assembled in the bookbuilding process, not the underlying cash-generation capacity of the business. The upstream signal is the xAI capex burn rate ($7.7B in Q1 2026 alone) and the $20B bridge loan maturity in 2027-2028.
Fails
Incentive-clean
The banks running the bookbuild earn fees on the raise. The insiders selling have 85% voting control and a 366-day lockup — they benefit from a high opening price. Every party producing the $1.75T figure has a financial interest in you believing it.
Fails
Distribution-stable
The $1.75T figure aggregates three businesses with radically different profiles: Starlink (profitable, $4.4B operating profit), Space (loss-making, Starship at 40% recovery rate), xAI ($6.35B operating loss, indeterminate moat). The aggregate masks a distribution where only one of three divisions supports the valuation.
Fails
Verdict: one test passed, three failed. The valuation is falsifiable — that is the most important test, and it passes. But the other three fail clearly. The signal for an investor is not the IPO price but the xAI burn trajectory, the bridge loan maturity, and whether Starlink ARPU stabilizes. The price is noise. The cash mechanics are signal.
The one rule
Run the four tests before acting on information — not after. Most executives reach for this framework retrospectively, when they are looking for reasons a decision went wrong. Used that way it is an autopsy tool. Used beforehand it is a decision filter. The whole value is in the timing.
The trap
Mistaking familiarity for signal. The number you have seen a hundred times feels more reliable than a new one. The official rate feels more trustworthy than the alternative measure. The published valuation feels more real than the independent estimate. Familiarity is not evidence. The most dangerous noise is the noise you have stopped questioning.
The Signal/Noise Split · A CULT+MATH Framework · Ipalibo Da-Wariboko · 2026