From a record to a decision takes several leaps
The words data, information, signal and intelligence tend to get mixed up. It is worth separating them because each level involves a different kind of work and a different risk.
- Data or record: a date, a figure or a captured text.
- Contextualised information: the record interpreted in terms of its origin, date, scope and conditions.
- Evidence: information whose provenance can be verified and which supports a specific statement; its strength depends on the question.
- Signal: a fact or combination of facts that could indicate a relevant change.
- Hypothesis: a provisional explanation that can be tested.
- Intelligence: a reasoned interpretation of evidence and alternatives in the service of a decision.
- Decision: a choice made by a responsible person, taking into account their own costs, risks, constraints and objectives.
There is no automatic lift between levels. The more confidence a text conveys, the more important it is to know what real support it has.
Interactive · Explore step by step
From record to decision, step by step
Walk through the seven levels with the same Company A record. At each step, notice what work is needed.
Walk through the seven levels, from data to decision, with the S03 example.
1. Data or record
A date, a figure or a captured text. It does not say anything by itself yet.
2. Contextualised information
The record interpreted in light of its origin, date, scope and conditions.
3. Evidence
Information of verifiable provenance that supports a specific claim. Its strength depends on the question.
4. Signal
A fact or combination of facts that could indicate a relevant change.
5. Hypothesis
A provisional explanation that can be tested.
6. Intelligence
An argued interpretation of evidence and alternatives in the service of a decision.
7. Decision
A choice made by the person responsible, who weighs their own costs, risks, constraints and objectives.
Company A case · simulation
Fictional caseExample · Company A · A patent is not a launch
In S03, a simulated register publishes a patent application from Supplier B. Within the simulation, we can state that an application described in the corpus was published. We cannot conclude, from that alone, that Supplier B holds a granted patent, masters the technology, has industrialised the result or has a commercial product.
The sentence «Supplier B has published an application» is a fact of the exercise. «Supplier B is strengthening its technology strategy» is an interpretation; it needs more signals. «It will launch a product in six months» would be a prediction without sufficient support. «We must invest now» is a recommendation that also requires analysing Company A’s capabilities and priorities.
Evidence is not a permanent label on a document. The same record can be strong evidence of a published application and weak evidence of commercial viability.
Educational simulation: fictional information and organisations.
The statement log
Before disseminating a briefing, break the relevant sentences down into a traceability matrix:
| Statement | Class | Documentary basis | What we know / what is missing |
|---|---|---|---|
| An application was published | Simulated fact | S03 | It does not indicate a grant |
| The firm is betting on this technology | Hypothesis | S01 + S03 + S07 | Executed investment and roadmap are missing |
| Commercialisation is near | Unsupported inference | None sufficient | Do not publish as fact |
| It is advisable to test before investing at scale | Decision option | S09 + S13, among others | Review cost and own capabilities |
Note: S01, S03 and S07 could share a corporate origin. Counting items is not enough: you must assess the independence and nature of each piece of evidence.
Do it yourself
Do it yourself · Audit of four sentences (10 minutes)
Classify and rewrite the following statements for a committee:
A. «A patent application proves that the technology already works». B. «A public consultation on energy efficiency is open». C. «A competitor’s entry could speed up market interest». D. «Company A should bet its entire innovation budget on this technology».
For each one, state: fact / hypothesis / inference / recommendation; source needed; communicable version with uncertainty.
Log of signals and evidence
Distinguish observation, source and interpretation.
See the template →Check it · closed answer
Classify the four sentences
Each sentence belongs to a different level: not everything that sounds certain is a fact.
A. «A patent application proves that the technology already works».
See the answer
Answer: Unjustified inference. An application does not prove that something works. Communicable version: «An application describing a solution has been published; its feasibility is not demonstrated here».
B. «A public consultation on energy efficiency is open».
See the answer
Answer: Fact. It is a fact of the corpus, conditional on S06 and its date. Check its later status in S15 before communicating it.
C. «A competitor's entry could speed up market interest».
See the answer
Answer: Hypothesis. It is a hypothesis: define which entry, adoption or investment would corroborate it.
D. «Company A should bet its entire innovation budget on this technology».
See the answer
Answer: Recommendation without grounds. It is an extreme recommendation without grounds: it requires alternatives, budget, risk, technical evidence and decision criteria.
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 sentence you state which source would be needed.
- You rewrite a communicable version that shows its uncertainty.
- You separate the source from the scope of the statement.
Indicative solution
A published patent application does not establish grant, performance or sales: it requires the source and the scope of the statement to be treated separately. Communicable version of A: «An application describing a solution has been published; its feasibility is not demonstrated here».
Illustrative solution for a fictional case: reasoned alternatives are valid.
A test for generative AI
When you use an assistant, demand four columns: statement, verifiable evidence, degree and reason for uncertainty, implication for the decision. If the assistant has not consulted sources, it must say so. A plausible reference that does not exist is worse than an explicit gap.
Optional extensionExtension · Traceability at the level of the statement, not the report
A bibliography at the end of a note does not prove every sentence it contains. Divide the conclusions into verifiable statements and assign each one the specific passage that supports it. Always add a scope field: «the source allows us to state X, but not Y». This matters especially when working with AI, because a fluent synthesis can join facts of different origin without warning.
Given S03, use four distinct tests: (1) existence of a published application, (2) legal grant, (3) proven technical feasibility and (4) commercial deployment. The first is represented in the fictional corpus; the other three cannot be deduced. Neither the same source nor five copies of it close that gap.
Micro-workshop · Rewriting an executive headline
Flawed headline: «Supplier B will lead the market thanks to its patent». Version that can be supported in the case: «The simulated register S03 indicates the publication of an application; we do not yet have enough information on grant, performance or commercial adoption». Useful implication: «Before recommending an alliance, cross-check technical demonstrations and customer commitments».
Compare both versions. The second looks less spectacular, but it lets you decide what to do about the uncertainty. Complete a third version that explicitly includes the cut-off date of the analysis.
Interactive · Rewrite
Rewrite the executive headline
Which of these versions could you send to the committee without claiming more than the corpus says?
Activity: choose the defensible rewrite of the headline “Supplier B will lead the market thanks to its patent”.
“Supplier B will lead the market thanks to its patent.”
Acceptance criterion for the exercise
Each relevant conclusion must include its category (fact of the case, inference, hypothesis, recommendation), reference, limit and next check. It is not enough to label the whole note as «based on sources».
What usually goes wrong
- Labelling as «evidence» a press repetition with no original document.
- Asking the AI for «firm conclusions» when the data only allow hypotheses.
- Mixing what is observed outside with what the company can do internally.
- Interpreting an isolated indicator as demonstrated causality.
Key takeaway: useful intelligence does not eliminate uncertainty; it makes explicit what we know, what we infer and what we need to check before deciding.
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 requirements of any standard.
Intelligence Lab · L05
Evidence auditor
Ready to apply it? Open the Evidence auditor in the Lab and review each claim: fact, inference, error or lack of evidence. Output: claim audit. Works in the browser, no server, no account and no connected AI. The result may need reviewing and adapting to your case.