A Star With Nothing to Measure: Factless Fact Tables

Opening Scene

Not every entry in the observatory’s log records a measurement. Sometimes the only thing worth recording is that an event occurred at all: a particular telescope was scheduled for a particular slot, a particular astronomer was present for a particular session, whether or not any reading was actually taken. There’s no number to log, no brightness or distance. The event itself — the pairing of telescope, slot, and astronomer — is the entire point, and it’s still worth recording carefully.

A factless fact table captures this exact same kind of event, with nothing numeric attached.

In Plain English

A factless fact table is a fact table with no actual numeric measures — just foreign keys connecting the dimensions involved in an event. It sounds like a contradiction, but it’s a genuinely useful pattern for tracking events or associations where the occurrence itself is what matters, like which students attended which classes, or which promotions ran in which stores, without any measure needing to be attached to that occurrence.

The Old Way

Recognizing when a factless fact table is genuinely the right pattern has always required understanding what it’s actually good for:

Event tracking, where recording that something happened is the entire point — an appointment being kept, a class being attended — is the classic use case, since counting the rows themselves (a count of attendance events) becomes the useful measure, derived rather than stored.
Coverage or eligibility tracking, capturing which combinations of dimensions were genuinely possible or eligible during a period — which products were available for promotion in which stores, for instance — is a second common use case, useful for answering “what didn’t happen” as much as “what did.”
The row count itself becomes the implicit measure: querying a factless fact table is usually a matter of counting matching rows rather than summing a stored numeric column, since there simply isn’t one to sum.

Getting this right has always meant recognizing that “no measure” doesn’t mean “no value” — a factless fact table still needs a genuinely clear grain, and it still answers real, important business questions, just through counting occurrences rather than summing quantities.

What’s Changing (and Why AI Is the Reason)

AI-assisted pattern recognition can identify genuine factless fact table candidates in raw source data, distinguishing event or coverage tracking from data that’s actually missing a measure it should have. Rather than a modeler assuming a table without an obvious numeric column is simply incomplete, AI-assisted analysis of the data’s actual usage and structure can confirm whether it genuinely represents a coverage or event pattern, appropriately modeled as factless.
AI agents answering “what didn’t happen” questions rely on factless fact tables specifically designed for coverage tracking, since this kind of negative-space question can’t be answered from a table that only records events that did occur. A question like “which eligible promotions never actually ran in a given store” requires a coverage-style factless fact table capturing the full set of eligible combinations, not just the ones that happened.
AI-assisted query generation needs to correctly recognize when a factless fact table calls for counting rows rather than summing a column, avoiding a category of query error unique to this pattern. An agent unfamiliar with the factless pattern might search for a numeric column to sum and fail, or worse, sum an unrelated column incorrectly; clear schema metadata identifying a table as factless helps an agent generate the correct counting query instead.

The Metaphor, Fully Extended

Observatory Element	Factless Fact Table Concept
A log entry recording only that a telescope, slot, and astronomer were paired, with no reading attached	A factless fact table row, connecting dimensions with no numeric measure
Counting how many sessions a particular astronomer attended	Deriving a measure by counting rows, rather than summing a stored numeric column
A log of every telescope slot that was genuinely available during a period, whether used or not	A coverage-tracking factless fact table, capturing eligible combinations for “what didn’t happen” analysis
An observatory reviewing whether an event log is genuinely factless or simply missing a reading it should have	AI-assisted pattern recognition distinguishing genuine factless design from an incomplete measure
An assistant correctly counting log entries rather than searching for a number to add up	AI-assisted query generation correctly recognizing when to count rows instead of summing a column

For Beginners: What to Actually Do

Practice recognizing that a fact table with no numeric measures isn’t a mistake — it’s a real, deliberate pattern for tracking events or coverage.
Get comfortable with the idea that a factless fact table’s implicit measure is usually the row count itself, derived through counting rather than stored directly.
Before assuming a table is missing data because it lacks a measure column, consider whether it might genuinely be tracking events or eligibility instead.
Notice that factless fact tables are particularly good at answering “what didn’t happen” questions, which measure-based fact tables generally can’t answer at all.

For Practitioners and Leaders: The Deeper Layer

Use AI-assisted pattern recognition to correctly distinguish genuine factless fact table candidates from tables that are actually missing a measure they should have.
Build coverage-tracking factless fact tables deliberately when “what didn’t happen” questions have genuine business value, since this pattern can’t be retrofitted from event-only data after the fact.
Maintain clear schema metadata identifying factless fact tables explicitly, helping AI agents generate correct counting queries rather than searching for a measure column that doesn’t exist.
Treat factless fact tables as a genuinely first-class dimensional modeling pattern, not an edge case, given how often event and coverage tracking come up in real business analysis.

Quick Recap

A factless fact table has no numeric measures, only foreign keys connecting the dimensions involved in an event, and its implicit measure is usually a count of rows.
Event tracking and coverage or eligibility tracking are the two classic, genuinely useful applications of this pattern.
AI-assisted pattern recognition can distinguish genuine factless design from an incomplete measure, and clear metadata helps AI agents generate correct counting queries.
Coverage-tracking factless fact tables are particularly valuable for answering “what didn’t happen,” a question measure-based fact tables generally can’t address at all.

Where This Fits in the Series

Article 11 covered sweeping up the loose stars. This article covered a star with nothing to measure. Article 13 looks at what happens when two stars share a planet — bridge tables.