A signal matters only in relation to a decision
A viral launch may have no effect on your activity. An update quietly published in a regulatory source could have more impact than a hundred press stories. Relevance is assessed according to the question: potential impact, available evidence, novelty compared with the history, urgency of verification and the consequences of being wrong.
Noise does not mean false information. It can be true information that changes no decision in our organisation. Conversely, a highly uncertain signal may deserve monitoring precisely because the cost of ignoring it would be high. «Low reliability» should not become «never look».
Three common reading errors
Counting mentions as confirmations. If an aggregator reproduces an article, the apparent growth of the topic may be simple duplication. In Company A, S04 and S05 point to the same content: they are not two independent verifications.
Selecting what confirms our thesis. If the team wants to bet on a new technology, it may give more weight to S11 (trial funding) than to S13 (performance variability), although both matter. Funding does not prove technical success.
Confusing an intention with an execution. Press releases, job offers, applications and consultations show different movements. A job offer may point to a need for capabilities, not to a completed hire.
Interactive · Cards
The three reading errors, on cards
Flip each card to see the Company A example and how to avoid it.
Three reading errors: counting mentions, selecting what confirms, and confusing intention with execution.
Challenge · survivorship bias
The planes that came back
The red dots are hits on the planes that returned from their missions. Armor is heavy: you can only reinforce one area. Which do you choose?
The planes that came back: this activity needs JavaScript.
Simplified account of Abraham Wald's work for the Statistical Research Group (US, World War II). See Mangel, M. and Samaniego, F. J. (1984), “Abraham Wald's Work on Aircraft Survivability”, Journal of the American Statistical Association, 79(386).
Reading matrix without false precision
You do not need to assign arbitrary probabilities. For a first classification, use four operational decisions:
| Category | When to use it | Next action |
|---|---|---|
| Escalate | High potential impact and a time-sensitive need | Verify quickly and notify the owner |
| Cross-check | Plausible but incomplete signal | Look for independent corroboration |
| Follow | Does not yet change the choice | Add to follow-up and define an indicator |
| Discard from focus | Repetition with no novelty or outside the scope | Record why it is not prioritised |
A signal can change category. That is why the log must keep the date, the criterion and the person who decided.
Do it yourself
Do it yourself · Six signals, four decisions (15–20 minutes)
Review S03, S04, S05, S08, S11 and S13 from the corpus. For each one write: 1) what it actually claims; 2) what it could mean; 3) what it does not allow you to conclude; 4) what you would do now. Then compare S04/S05 and explain why they are not independent.
Rule against biases: before closing a conclusion, deliberately look for evidence that weakens it. For example, set the signals of economic momentum against the trial limits in S13 and the uncertainties in S09.
Log of signals and evidence
Distinguish observation, source and interpretation.
See the template →Check it · closed answer
What would you do now with each signal?
There is no universal ranking: it depends on the timeframe and the constraints of Company A. Several reasonable answers are accepted.
S03 · Published patent application (simulation).
See the answer
Accepted answers: Follow or Cross-check. Follow or cross-check: it does not prove a patent has been granted.
S04 · Secondary news item about a demonstrator.
See the answer
Answer: Cross-check. Cross-check against the original test and its results.
S05 · Reproduction of the S04 news item on an aggregator.
See the answer
Answer: Discard as independent confirmation. S04 and S05 are not independent sources. Keep it for traceability if appropriate.
S08 · Unsourced rumor about a plant in Europe.
See the answer
Answer: Do not escalate as fact. Do not escalate as fact; verify only if the possible impact justifies the effort.
S11 · Funding granted (simulation).
See the answer
Answer: Follow. Follow: funding is not technical effectiveness.
S13 · Variable performance in a trial.
See the answer
Accepted answers: Escalate to the technical expert or Cross-check. Cross-check or escalate to the technical expert, because of its potential impact on the decision. S11 and S13 can both be true at the same time.
And now, your own answer
This part needs professional judgement: there is no automatic marking. Complete your work in the practice, rate it with the lesson's rubric and compare it with the indicative solution.
What your answer should include
- For each signal you write what it claims, what it could mean, what it does not allow you to conclude and what you would do now.
- You explain why S04 and S05 are not independent.
- You deliberately look for evidence that weakens your conclusion (for example, S13 and S09).
Indicative solution
S04 and S05 are not independent sources; S11 and S13 can be true simultaneously within the fictional assumption. No universal ranking is prescribed: the classification depends on the timeframe and the constraints of Company A.
Illustrative solution for a fictional case: reasoned alternatives are valid.
AI and prioritisation: useful, but supervised
An AI can propose labels, group news items and suggest signals to review. It does not automatically know the value of each decision for your organisation, nor does it guarantee that the sources are independent. Require it to cite the content actually available, to set out its reasons for the classification and to flag missing information.
Optional extensionGoing deeper · Two dimensions that must not be confused
The relevance of a signal depends on its relationship to a decision; the quality of the evidence depends on provenance, independence, precision and the conditions under which it was obtained. A verified news item can be irrelevant, and an unverified rumor can justify an urgent check if its potential consequences are large. The rumor must not be disseminated as fact however high its hypothetical impact.
A particularly frequent error is apparent independence: S05 repeats S04. Counting both as corroborations artificially increases confidence. Label them «same chain of origin» and look for a primary document, a different trial or an external confirmation before escalating the conclusion.
Interactive · Classify
Relevance and quality of evidence are not the same
Place each Company A signal in the quadrant where it belongs.
Activity: place signals in a matrix of relevance to the decision and quality of the evidence.
Guided exercise · optionalGuided exercise · Contradiction between signals
The team celebrates S11, a simulated funding grant, as proof that the technology works. However, S13 presents variable performance results. Write a mini note with three parts: «What drives the research», «What limits a claim of maturity» and «Which trial or independent source would resolve part of the tension».
Model reasoning: funding a trial is evidence of interest and resources to investigate it, not of the trial’s success; a technical limitation requires checking the test conditions and the possibility of mitigation. It is not compulsory to pick one of the two signals as the «winner».
Acceptance criterion for the exercise
The classification must state an operational action and a justification; separate out at least one duplicated source; recognise a signal that contradicts the preferred hypothesis; and establish which new observation would change the priority.
Key takeaway: a signal does not become important by appearing many times; it becomes important when it can change a decision and we know what evidence supports it.
References and limits
This lesson cites no external sources. It develops the Academy method from Innguma's training material.
- Company A is a fictional case: its data, companies and figures do not describe real markets or organisations.
- The cycle, the critical factors and the intelligence questions are presented as a general Academy method; they do not reproduce the requirements of any standard.
Intelligence Lab · L04
Signal log
Ready to apply it? Open the Signal log in the Lab and keep observation, source and interpretation separate in your log. Output: signal log. Works in the browser, no server, no account and no connected AI. The result may need reviewing and adapting to your case.