WebMEM™

The Protocol for Structuring, Delivering, and Conditioning Trust-Scored AI Memory on the Open Web

DerivedStatsFragment

Part of the WebMEM® Protocol
Fragment Class: DerivedStatsFragment
Location: /protocol/fragments/derivedstatsfragment/
Status: Current Draft
Last Updated: 2026-08-24

Overview

A DerivedStatsFragment is a WebMEM fragment class for publishing calculated, aggregated, summarized, or otherwise derived knowledge produced from identified source data.

DerivedStatsFragments represent values created by applying a defined calculation or transformation to underlying facts. Typical examples include counts, percentages, averages, distributions, ranges, rankings, ratios, and conditional aggregates.

A DerivedStatsFragment may contain one or more Machine Fact Objects. Each Machine Fact Object represents an individual derived factual assertion within the fragment.

Unlike a DataFragment, which represents source-observed facts, a DerivedStatsFragment represents publisher-derived facts whose evidentiary basis remains connected to the underlying source data.

Semantic Purpose

The semantic purpose of a DerivedStatsFragment is to state:

These factual values were calculated or derived from identified source data and apply to this defined entity or context.

A DerivedStatsFragment provides a coherent boundary for derived knowledge that shares common context such as:

  • the entity or context to which the derived values apply;
  • the source dataset or datasets used in the derivation;
  • the applicable time period;
  • the scope of the calculation;
  • the provenance of the underlying data;
  • and other information necessary to interpret the resulting values correctly.

The fragment distinguishes the resulting knowledge from the source-observed facts from which it was produced.

DerivedStatsFragment Within an SDT

A DerivedStatsFragment exists as one modular knowledge object within a Semantic Data Template (SDT).

The relationship is:

Web Resource → SDT → DerivedStatsFragment → Machine Fact Objects

An SDT may contain one or more DerivedStatsFragments depending on the derived knowledge represented by the resource.

For example, a resource describing a geographic market might contain one DerivedStatsFragment for market-level statistics while separate DataFragments expose source-observed facts and an IndexFragment identifies the entities available within that market.

Each fragment retains its own semantic purpose while participating in the same page-level machine knowledge representation.

DerivedStatsFragment Requirements

A conforming DerivedStatsFragment should establish enough information for its derived values and their evidentiary basis to be interpreted within the SDT.

Element Purpose
Fragment Class Identifies the knowledge object as a DerivedStatsFragment.
Fragment ID Provides a stable identifier for the fragment within the published knowledge structure.
Entity or Context Identifies the entity, location, collection, record set, or other scope to which the derived values apply.
Machine Fact Object(s) Contains one or more calculated, aggregated, summarized, or otherwise derived factual assertions.
Source Identifies the source dataset or datasets from which the values were derived.
Derived Status Explicitly identifies the resulting factual values as derived rather than source-observed.
Provenance Connects the derived knowledge to the source and lineage information supporting the calculation.

Additional information may be included when required by the derivation, domain, source, entity model, or applicable conformance profile.

What Counts as Derived Knowledge?

A factual value is derived when the publisher creates a new assertion by applying a calculation, aggregation, comparison, summarization, filtering operation, or other transformation to underlying source data.

Examples include:

  • an average calculated across multiple source records;
  • a count of records meeting defined criteria;
  • a percentage calculated from a numerator and denominator;
  • a minimum or maximum identified across a collection;
  • a distribution across defined categories;
  • a ranking based on an observed metric;
  • a range calculated from source values;
  • a ratio between two source-derived quantities;
  • or a conditional aggregate calculated from a defined subset of source data.

The distinction between observed and derived knowledge concerns the origin of the assertion.

A value copied or faithfully normalized from a source remains source-observed. A new factual assertion created through calculation or analysis is derived.

Machine Fact Objects

Individual derived values within a DerivedStatsFragment are represented as Machine Fact Objects.

A derived Machine Fact Object may identify:

  • a field identifier;
  • a defined term;
  • a description;
  • the calculated value;
  • a datatype;
  • a unit of measurement;
  • a derived status;
  • a source reference;
  • a provenance reference;
  • a confidence designation where applicable;
  • and other information necessary to interpret the derived assertion.

Where multiple derived facts share the same source, entity, and calculation context, that context may be established once by the fragment rather than repeated unnecessarily for every Machine Fact Object.

Read the Machine Fact Objects specification →

Provenance and Derivation

Provenance is particularly important for a DerivedStatsFragment because the published values do not appear directly in the underlying source.

The fragment must preserve the relationship between the derived knowledge and the source information from which it was produced.

Provenance may identify:

  • the source dataset or datasets;
  • the source creator or publisher;
  • the canonical source location;
  • an archive or source-file location;
  • publication or effective dates;
  • retrieval or import dates;
  • source version information;
  • source format;
  • checksum or integrity information;
  • and other lineage information required to establish the evidentiary basis of the derived values.

Where the method of derivation is necessary to understand or reproduce a value, the publisher should also expose sufficient information to identify how the value was calculated.

The detailed requirements for source lineage and derivation are defined by the WebMEM Provenance specification.

Read the Provenance specification →

Entity and Context Scope

A DerivedStatsFragment must establish the scope within which its calculations are meaningful.

That scope may represent:

  • a specific entity;
  • a geographic area;
  • a collection of entities;
  • a dataset or subset of records;
  • a product category;
  • a time period;
  • or another explicitly defined context.

For example, an average premium has no complete meaning without knowing the collection of plans, geographic area, plan year, or other boundaries used to calculate it.

WebMEM therefore treats the scope of a derived statistic as part of the knowledge required to interpret that statistic correctly.

Example DerivedStatsFragment

The following example represents several derived statistics calculated from Medicare Advantage plan data within a defined county context.

<template
  data-webmem-fragment
  data-fragment-class="DerivedStatsFragment"
  data-fragment-id="county-ma-derived-stats">

  <section
    data-entity-type="County"
    data-entity-id="04015"
    data-plan-year="2026"
    data-provenance-ref="#provenance-cms-landscape">

    <div data-field="num_total_plans">

      <span data-role="defined-term">
        Total Medicare Advantage Plans
      </span>

      <p data-role="description">
        Total number of Medicare Advantage plans available in this county.
      </p>

      <data
        data-role="value"
        value="42"
        data-type="integer">42</data>

      <meta
        data-role="derived"
        content="true">

    </div>

    <div data-field="avg_partc_premium">

      <span data-role="defined-term">
        Average Part C Premium
      </span>

      <p data-role="description">
        Average monthly premium across Medicare Advantage plans in this county.
      </p>

      <data
        data-role="value"
        value="27.42"
        data-type="number"
        data-unit="USD">$27.42</data>

      <meta
        data-role="derived"
        content="true">

    </div>

    <div data-field="percent_0_premium">

      <span data-role="defined-term">
        Percent of $0 Premium Plans
      </span>

      <p data-role="description">
        Percentage of Medicare Advantage plans in this county with a $0 monthly premium.
      </p>

      <data
        data-role="value"
        value="61.90"
        data-type="number"
        data-unit="percent">61.90%</data>

      <meta
        data-role="derived"
        content="true">

    </div>

  </section>

</template>

The fragment establishes that:

  • the knowledge object is a DerivedStatsFragment;
  • the statistics apply to a defined county and plan year;
  • the fragment contains multiple derived Machine Fact Objects;
  • the resulting values are explicitly identified as derived;
  • and the underlying source information can be traced through the fragment’s provenance reference.

Multiple Source Datasets

A DerivedStatsFragment may be based on one source dataset or multiple identified sources.

When all derived facts within the fragment share the same evidentiary basis, common provenance may be established at the fragment level.

When different derived facts depend on different source datasets, individual Machine Fact Objects may reference the provenance appropriate to their derivation.

The publisher should preserve enough source separation that a consuming system can determine which underlying information supports each derived assertion.

DerivedStatsFragment vs. DataFragment

DataFragment DerivedStatsFragment
Represents source-observed factual knowledge. Represents calculated, aggregated, summarized, or transformed factual knowledge.
The factual assertion exists in the identified source. The factual assertion is created from underlying source data.
Example: a plan’s published monthly premium. Example: average monthly premium across plans in a county.
Example: enrollment reported for a specific plan. Example: total enrollment across a defined collection of plans.
Example: a plan’s CMS star rating. Example: percentage of available plans rated 4 stars or higher.

Both fragment classes preserve provenance. The difference is whether the published assertion was observed from the source or derived from source information.

DerivedStatsFragment vs. IndexFragment

A DerivedStatsFragment and an IndexFragment may both summarize information about a collection, but they serve different semantic purposes.

A DerivedStatsFragment publishes facts calculated about the collection.

An IndexFragment publishes the identifiable members or objects within the collection.

For example:

  • DerivedStatsFragment: There are 42 plans available in this county.
  • IndexFragment: These are the 42 plan identifiers available in this county.

The two fragments can therefore complement one another within the same SDT.

Derived Statistics and Resolution

Derived statistics primarily describe a defined entity or context rather than provide resolution instructions.

However, the objects and collections referenced by a DerivedStatsFragment may participate in the broader WebMEM Resolution Model.

For example, a DerivedStatsFragment may state that 42 plans exist within a county while an associated IndexFragment enumerates those plans and provides the identifiers needed to resolve individual plan resources.

This preserves the semantic role of the DerivedStatsFragment while allowing its knowledge to participate in a larger resolvable information structure.

Conformance

A conforming DerivedStatsFragment must:

  • identify itself as a DerivedStatsFragment;
  • have an identifiable fragment scope;
  • associate its derived facts with a defined entity or context;
  • contain one or more derived factual values represented according to the WebMEM fact model;
  • identify those values as derived;
  • identify or reference the source data supporting the derivation;
  • preserve sufficient provenance to establish the evidentiary basis of the resulting values;
  • and conform to the applicable WebMEM serialization requirements.

Where interpretation or reproducibility depends on the derivation method, sufficient information about that method should also be provided.

Relationship to WebMEM 1.x

The DerivedStatsFragment class originated in the WebMEM 1.x Semantic Data Template specification and remains a defined fragment class in the current WebMEM Protocol.

Earlier specifications serialized DerivedStatsFragments using YAML-in-HTML and associated them with Semantic Digest scope, export routing, and memory-layer behavior.

The current protocol preserves the semantic purpose of the class while separating that purpose from serialization and retired WebMEM 1.x mechanisms.

DerivedStatsFragment continues to mean what its name says: a bounded WebMEM knowledge object containing statistics or factual metrics derived from identified source data.

Related Protocol Documents