Intelligence Analysis & OSINT

Intelligence analysis is what data work becomes when a real decision hangs on it and the picture is never complete. It's the discipline of turning fragmentary, sometimes contradictory information into an assessment a decision-maker can act on — and being honest about how confident that assessment deserves to be. The maths and models from the rest of this section are tools it uses; the discipline itself is about judgement under uncertainty.

It's the heart of my current work in government, and it has its own tradecraft — a body of method built precisely because the hardest adversary an analyst faces isn't the subject of the analysis, but the predictable ways their own mind gets things wrong. This page is that tradecraft, made plain.

Data with a decision attached

The defining feature of intelligence is its purpose: it exists to inform a specific decision, for a specific person, who will act on it. That distinguishes it from analysis done out of curiosity. An intelligence product isn't judged on how clever it is, but on whether it helped someone make a better call with imperfect information — under time pressure, with consequences.

So intelligence is fundamentally about assessment under uncertainty. You will almost never have all the facts; the job is to make the best-supported judgement you can from what you have, state how much weight it can bear, and hand it over in time to be useful. Certainty is not on offer; calibrated judgement is.

The intelligence cycle

Intelligence work runs on a recognised loop, the intelligence cycle, which keeps the effort tied to the decision it serves:

Direction — what does the decision-maker actually need to know? The requirement that drives everything.
Collection — gather the relevant information from available sources.
Processing — turn raw material into usable, organised form.
Analysis — the core: assess what it means, weigh the hypotheses, form a judgement.
Dissemination — deliver the assessment to the decision-maker, clearly and in time.

Like the data-science lifecycle, it's a loop, not a line — dissemination raises new questions that feed back into direction. And the same lesson applies: the analysis is only as good as the question at the top, and only matters if it reaches the decision-maker in a form they can use.

The intelligence cycle. A decision-maker's need drives collection, processing, and analysis into a finished assessment — which, once delivered, raises the next question. A loop, always tied to a decision.

Intelligence vs data analysis

Intelligence and data analysis overlap, but the emphasis differs in a way worth naming. Data analysis often asks what does the data show? Intelligence insists on the next step: what does it mean for the decision, and so what should we do? The "so what" isn't optional polish — it's the product.

Intelligence also routinely reasons from incomplete and unreliable information, where a clean dataset is a luxury you don't get. So it leans less on a single number and more on weighing competing explanations, grading how much each source can be trusted, and being explicit about the gaps. The quantitative toolkit from the rest of this section absolutely helps — but the core skill is structured reasoning under doubt.

The enemy is your own mind

The central insight of modern intelligence tradecraft is humbling: the biggest threat to a sound assessment is not bad data — it's the analyst's own cognitive bias. Human minds take shortcuts that served us on the savannah and betray us on hard problems:

Confirmation bias — seeing the evidence that fits the theory you already hold and discounting the rest.
Anchoring — over-weighting the first piece of information you got.
Premature closure — settling on an answer too early and stopping the search.

You can't switch these off by trying harder — willpower doesn't fix a wiring problem. What works is method: structured processes that force you to consider what you'd otherwise skip. That's the entire reason structured analytic techniques exist.

Structured analytic techniques

Structured Analytic Techniques (SATs) are formal methods that externalise reasoning — get it out of your head and onto paper where its flaws show. They make analysis more rigorous, more transparent, and more defensible. The most important is the workhorse of the craft:

Two more that earn their keep daily: a Key Assumptions Check — write down every assumption your judgement rests on and ask what happens if each is wrong — and rigorously separating the reporting from your interpretation: keeping "here's what the source said" distinct from "here's what I think it means", so a reader can see exactly where the facts end and your judgement begins.

OSINT and source grading

Open-Source Intelligence (OSINT) is intelligence drawn from publicly available information — news, public records, social media, company filings, imagery. It's vast and powerful, and it's exactly where the discipline matters most, because open sources are often contradictory, incomplete, and sometimes deliberately deceptive.

So you never take a source at face value — you grade it on two separate axes: how reliable is the source (its track record and access), and how credible is this particular piece of information (does it fit what else is known, is it corroborated)? A reliable source can still pass on a dubious claim, and an unreliable one can occasionally be right — keeping the two judgements apart is the discipline. Corroborate across independent sources, trace claims to their origin, and stay alert to the verification problem that the same false story echoing across ten sites is still one claim, not ten.

The language of confidence

Because intelligence trades in uncertainty, how you express confidence is part of the product. Vague words betray the reader: "likely" might mean 55% to one person and 90% to another. Good practice uses a consistent set of probability yardsticks — a defined ladder from "remote" through "even chance" to "almost certain" — and separates that estimative likelihood from your confidence in the underlying evidence (a high-likelihood judgement built on thin sourcing is a different thing from one built on strong sourcing).

This is the statistics lesson of being honest about uncertainty, turned into disciplined language. Calibrated wording — neither falsely precise nor uselessly hedged — is what lets a decision-maker weigh the assessment correctly.

Probity and the law

Intelligence work, especially in government and policing, runs inside hard ethical and legal limits. Collection must be lawful and proportionate; handling must respect privacy and governance; and the analyst carries a duty of probity — being honest, impartial, and rigorous, precisely because the assessments can affect people's lives and liberty. The discipline isn't only about being right; it's about being right in a way that's defensible, traceable, and fair.

Where it shows up in my work

The core of the current role

This is the centre of what I do now. As a senior analyst in a government professional-standards setting, the work is exactly this: turn incomplete information into a defensible assessment for a decision-maker, under real constraints. The discipline on this page is the daily practice — guarding against my own bias with structured techniques like ACH, grading sources rather than trusting them, keeping reporting separate from interpretation, and being calibrated and lawful about what I can actually conclude.

It's where the rest of the section comes together in service of a decision: the statistics for honest uncertainty, the spatial analysis for the "where", the communication for the hand-off — all pointed at the same target: a sound, defensible judgement that helps someone decide well.

Refresh in 60 seconds

Tradecraft on this page reflects established intelligence-analysis references (the CIA/IC "Tradecraft Primer" on structured analytic techniques, OSINT practice) alongside hands-on government work.