THE INTEL ANALYST ACADEMY
Professional Intelligence Training
Printable Lesson
Estimative Language
The Intel Analyst Academy · Lesson Notes
Intelligence analysts almost never deal in certainty. The future is murky, sources are incomplete, and the adversary is not cooperating with your timeline. Estimative language is the disciplined system analysts use to communicate how likely something is and how confident they are in that judgment. Get it right, and you empower decision-makers. Get it wrong, and you could launch a crisis -- or miss one entirely.
The modern framework for estimative language traces back to Sherman Kent, the father of modern intelligence analysis. In the 1960s, Kent noticed that when CIA analysts wrote "probable," some readers interpreted that as 80% likely while others read it as 55% likely. That gap could mean the difference between mobilizing troops and writing a memo.
Kent's solution was a standardized probability scale that assigns numerical ranges to specific words and phrases. While exact ranges vary by agency and country, the hierarchy is broadly consistent across the Western intelligence community.
99-95% Almost Certain / Nearly Certain
93-80% Very Likely / Highly Probable
75-60% Likely / Probable
55-45% Roughly Even Chance
40-20% Unlikely / Improbable
15-5% Very Unlikely / Highly Improbable
5-1% Remote / Almost No Chance
Sherman Kent once said that if intelligence analysts cannot agree on what "probable" means, we have bigger problems than the Soviet Union. He was right. And we still argue about it.
Not all agencies use the same scale. The U.S. Intelligence Community (IC) and the UK's Joint Intelligence Committee (JIC) use slightly different terminology and ranges. The Australian intelligence community has its own variant. When working in coalition environments or reading foreign intelligence, always check which scale is in use. "Likely" in a CIA product and "likely" in a JIC assessment may carry different implied probabilities.
When your report will be read across agencies or by coalition partners, include a brief footnote defining your probability terms. Two sentences can prevent a misunderstanding that leads to very expensive decisions.
One of the most misunderstood distinctions in intelligence writing is the difference between probability and confidence. They answer different questions, and conflating them is a fast track to misleading your reader.
Probability addresses the likelihood of an event occurring. When you write "Iran will likely continue uranium enrichment," you are making a statement about the expected future behavior of the subject.
Confidence addresses the quality and quantity of your evidence. A "high confidence" judgment means you have strong sourcing, corroboration, and analytical support. A "low confidence" judgment means you are working with thin, uncorroborated, or potentially unreliable information.
You can have high probability with low confidence, and vice versa:
High Probability, Low Confidence
"We assess that the regime will likely [70%] conduct a missile test within 60 days (low confidence), based on a single uncorroborated human intelligence report."
The event seems likely, but the evidence is thin. Proceed with caution.
Low Probability, High Confidence
"We assess that the regime is unlikely [25%] to abandon its nuclear program (high confidence), based on consistent signals intelligence, imagery, and diplomatic reporting over 18 months."
The event is unlikely, and we are quite sure about that judgment.
Confusing confidence and probability is like confusing "I'm pretty sure I locked the door" with "the door is probably locked." One is about your memory; the other is about the door. Both matter, but they are not the same thing.
Even analysts who know the scale make mistakes in application. Here are the most frequent offenders:
Hedge Stack
"It is possible that the group may potentially be likely to consider conducting an attack."
Four hedges in one sentence. What is the actual probability?
Clean
"The group will likely conduct an attack within six months."
One estimative term. Clear probability. Defined timeframe.
Using "likely," "probable," and "expected" interchangeably within the same report creates confusion. Pick one term per probability level and stick with it throughout. If you say "likely" on page 2 and "probable" on page 5, your reader may wonder if you are describing different levels of likelihood.
The words "could" and "may" are capability statements, not probability statements. "Iran could develop a nuclear weapon" says nothing about whether they will -- only that they have the potential. Analysts sometimes retreat to "could" and "may" to avoid committing to a probability judgment. This is a failure of analytical courage.
"The regime will likely test a missile." When? This week? This decade? Without a timeframe, an estimative judgment is incomplete. Always pair your probability with a defined period.
Before finalizing your report, do an "estimative language audit." Search for every instance of likely, unlikely, possible, probable, and may. For each one, verify: (1) it maps to a specific probability range, (2) it is consistent with other uses in the same report, and (3) it includes a timeframe.
Let us put it all together with full before-and-after examples drawn from common intelligence writing scenarios.
Before
"There is a possibility that the group might try to attack the embassy at some point in the future."
After
"We assess with moderate confidence that the group will likely (60-75%) attempt to attack the embassy within the next 90 days, based on intercepted communications and observed pre-operational surveillance activity."
Before
"The president might not survive the next election."
After
"We assess with high confidence that the incumbent is unlikely (20-35%) to win re-election in the March 2026 cycle, given persistent economic decline, falling approval ratings, and credible opposition consolidation."
Notice the pattern: the "after" versions include **a probability term, a confidence level, a timeframe, and the supporting evidence**. This is the gold standard for estimative writing.
The before version of Example 1 is the analytical equivalent of a fortune cookie. Technically true, but nobody is going to scramble fighter jets over it.
Continue your training
This lesson is part of The Intel Analyst Academy — professional intelligence analysis training built for analysts. Explore the full course library, structured learning paths, and practical tools at theintelanalystacademy.com.
