Descriptive Analysis
Describe the state before explaining it. Descriptive analysis reconstructs what was observed, in which population, at what time and under which measurement rules—without smuggling cause, mechanism or prediction into the result.
The observed state,made explicit.
Descriptive analysis is not a weak version of explanation. It is the controlled operation that determines exactly what the evidence says before any explanatory operation begins.
Descriptive analysis organizes, reduces and represents observed data inside a declared boundary so that the state, composition, distribution and variation of the object can be inspected without claiming why that state exists.
The output must preserve the observation unit, eligible population, measurement time, transformation rules, exclusions and missing records. A number without these controls is not a stable description; it is an unanchored value.
Six questions.Six different summaries.
The analytical question determines the correct representation. Count, composition, distribution, location, recurrence and change are not interchangeable views of the same thing.
How many?
Returns a count of eligible observed units. The eligible population and duplicate rule must be declared.
OUTPUT / FREQUENCYHow often?
Relates an event count to exposure, time or another valid denominator. A raw count cannot substitute for a rate.
OUTPUT / NORMALIZED FREQUENCYHow distributed?
Shows center, spread, concentration, tails and outliers rather than collapsing the population into one average.
OUTPUT / DISTRIBUTIONHow composed?
Expresses categories as parts of an explicitly complete or qualified whole while retaining unknown categories.
OUTPUT / COMPOSITIONWhere located?
Places units inside a spatial, architectural, semantic or market territory using a declared coordinate system.
OUTPUT / LOCATIONWhat changed?
Describes a delta only when states share compatible units, definitions, collection rules and time boundaries.
OUTPUT / BOUNDED CHANGEEvery percentagehas a population.
The numerator describes the observed condition. The denominator defines the universe to which the statement applies. Change either and the meaning changes—even if the displayed percentage remains identical.
Count ≠ rate ≠ share.
A count reports magnitude. A rate normalizes events against exposure. A share reports composition within a whole. A ratio compares quantities that may not form a whole.
One dataset.Four legitimate views.
Switch the analytical operation. The source values remain illustrative and fixed; only the representation and the claim it supports change.
Observed query states
SOURCE NOTE: values are an interface example used to demonstrate analytical representation; they are not live market measurements.
Magnitude
Inside the declared 250-query set, 146 observations are rival-only, 82 are shared and 22 are target-only.
Nothing is notone single state.
Zero, absent, unavailable, not applicable and excluded encode different realities. Collapsing them into one empty cell corrupts counts, shares and every later interpretation.
Preserve why a value is missing.
A descriptive system must carry null-state provenance. If absence is transformed into zero, the interface creates evidence that was never observed.
The property was eligible and measured; the resulting value is zero.
VALID NUMERIC STATEThe expected property was not present within the inspected boundary.
OBSERVED ABSENCEThe value may exist, but collection failed or access was unavailable.
UNKNOWN STATEThe measure does not logically apply to this unit or category.
OUTSIDE MEASUREThe record existed but was excluded through a declared eligibility rule.
CONTROLLED REMOVALA snapshot is nota trend.
Move the control from one to five observations. Descriptive language must change as temporal depth increases, while causal language remains prohibited.
Snapshot
One bounded observation describes the state at collection time. It establishes neither direction nor persistence.
The average cannotcarry the whole state.
Center, spread, range and composition expose different properties. A defensible description uses the smallest set of summaries that preserves the structure relevant to the question.
Typical position
Mean, median and mode answer different questions and respond differently to skew and extreme values.
Variation
Range, interquartile range and standard deviation describe how far observations extend around the center.
Quantiles
Percentiles locate an observation inside the distribution without pretending that intervals are equal in meaning.
Parts of the whole
Category shares require mutually intelligible groups, an explicit whole and preservation of unknown states.
From records toa bounded statement.
Two domains, one grammar. The interface changes the object and values while preserving scope, unit, denominator, missingness and claim limits.
Describe the footprint.
The unit is one canonical HTML URL. The population contains every eligible URL discovered inside the declared domain snapshot.
Descriptions failbefore explanations begin.
Most descriptive errors are boundary errors disguised as clean numbers. These eight failure modes contaminate every downstream comparison, pattern and conclusion.
Percentage without denominator
The reported share cannot be mapped back to an eligible population.
Mean without distribution
A center value hides skew, spread, clusters and extreme observations.
Mixed categories
Categories overlap, change definition or combine unlike observation units.
Snapshot as trend
One state or one difference is narrated as persistent directional movement.
Rank as quality
An observed position is silently converted into an explanation or value judgment.
Missing as zero
Unavailable evidence is recoded as a measured absence and changes the state.
Aggregation blindness
A total conceals materially different segments, cohorts or territories.
Stale description
A valid historical state is presented as if it still describes the current object.
Eight controls.One stable state.
The descriptive chain remains auditable because every reduction can be traced back to a declared object, source record and transformation.
Define identity and observation unit before collecting or grouping records.
CONTROL / IDENTITYFix population, locale, surface, period, exclusions and eligibility rules.
CONTROL / SCOPERetain source values, timestamps, provenance and original null conditions.
CONTROL / SOURCEApply documented mappings without erasing materially different states.
CONTROL / TRANSFORMBind each count, rate and share to its valid eligible population.
CONTROL / UNITSelect measures that preserve center, spread, composition or location.
CONTROL / REDUCTIONSeparate zero, absence, unavailable, not applicable and excluded records.
CONTROL / LOSSState what was observed, where and when—and what was not established.
CONTROL / CLAIMComplete description.No causal leakage.
A descriptive output is ready for exploratory or comparative analysis only when another analyst can reconstruct the population, operations and claim boundary.
Within [boundary], at [time], [value or distribution] was observed across [eligible population], subject to [missingness and exclusions].
This grammar keeps the statement attached to its evidence. Explanation, mechanism, forecast and decision value remain separate analytical operations.
One root.Twelve analytical nodes.
Descriptive Analysis is the second node: it converts governed observations into a bounded account of state before discovery, comparison, pattern detection or causal reasoning.