Jira progress: loading…

AGGR

Aggregation Engines

1. Overview

Aggregation Engines (AGGR) are Micro Engines responsible for deterministically combining already-computed and validated results into higher-order, consolidated outputs.

AGGR engines do not compute new primary values.
They derive secondary results strictly by aggregating existing outputs along explicitly defined dimensions.

AGGR is an engine type in the MICE taxonomy.
Concrete aggregators are implemented as MEID-registered micro engines.

---

2. Design Principles

1. Deterministic

   • Given the same inputs and aggregation rules, AGGR engines always produce the same output.

2. No Assumptions

   • AGGR engines never estimate, extrapolate, or fill gaps.
   • Missing data must be handled upstream (e.g. SEM).

3. Post-Compute Only

   • AGGR engines operate exclusively on validated computation outputs — never on raw inputs.

4. Full Lineage Preservation

   • Every aggregated result retains a complete reference to all contributing executions.

---

3. Scope of Responsibility

3.1 What AGGR engines do

AGGR engines perform mathematical and structural aggregation across explicit dimensions:

Structural

• asset → site
• site → company
• company → group

Temporal

• day → month
• month → quarter
• quarter → year

Logical

• scope-level consolidation (e.g. Scope 1 + Scope 2)
• category-level roll-ups (e.g. Scope 3 categories)
• portfolio, region, or segment summaries

3.2 Supported aggregation modes

Depending on the metric type and rule definition, AGGR engines support:

• Sum
• Weighted sum
• Average
• Weighted average
• Minimum / maximum (only where explicitly allowed)
• Count

All aggregation rules are explicitly declared, versioned, and auditable.

---

4. What AGGR engines do not do

AGGR engines explicitly do not perform:

• ❌ Primary computation (CALC)
• ❌ Validation of raw inputs (VALI)
• ❌ Unit or format conversion (TRFM)
• ❌ Estimation or extrapolation (SEM)
• ❌ Policy routing or enforcement (ZSSR)
• ❌ Interpretation or narrative generation (ZARA)

This strict separation is essential for explainability and assurance.

---

5. Inputs

AGGR engines consume:

• Collections of validated computation outputs
• Aggregation context, including:
  • hierarchy or scope definition
  • time window
  • aggregation rule reference
• Optional weighting parameters (e.g. revenue, production volume, area)

All inputs must conform to declared output schemas.

---

6. Outputs

AGGR engines produce:

• A consolidated output payload (schema-validated)
• Aggregation metadata:
  • aggregation rule identifier
  • aggregation dimensions
  • engine version
• Provenance references:
  • list of contributing execution IDs
  • source method identifiers and versions

Aggregated outputs may be further consumed by:

• higher-level AGGR passes,
• reporting and disclosure pipelines,
• assurance and verification workflows.

---

7. Audit & Provenance

Every AGGR execution produces an auditable aggregation record linking:

• all source execution IDs,
• the aggregation rule definition,
• the AGGR engine version,
• timestamps and execution context.

This enables:

• full replay of aggregated results,
• verifier inspection,
• regulatory audit readiness.

AGGR engines are fully integrated with ALTD / DAL logging.

---

8. Position in the computation chain

AGGR engines operate after computation and normalization:

INPUT
  → VALI
  → CALC
  → TRFM
  → (optional SEM)
  → AGGR
  → Reporting / Disclosure / Assurance

This ordering ensures aggregation is applied only to trusted, normalized data.

---

9. Canonical identification

• Engine Type: AGGR
• USO role: produces aggregated signals from validated source signals
• Category: Micro Engine (MICE)

AGGR identifiers are used consistently across:

• execution logs,
• provenance records,
• USO naming,
• explainability surfaces.

---

10. Registry view

All engines tagged as aggr:

Loading micro engines…

---

11. Design rationale

By elevating aggregation to a dedicated engine type, ZAYAZ ensures:

• a clear semantic distinction between computed, estimated, and aggregated values,
• deterministic behavior suitable for assurance and verification,
• long-term scalability as reporting and disclosure complexity increases.

Aggregation is not an implementation detail — it is a semantic operation, and AGGR makes it explicit.

---

12. Related engine types

• CALC — Calculation Engines
• VALI — Validator Engines
• TRFM — Transformation Engines
• SEM — Stochastic Extrapolation Module
• ZARA — Explainability & orchestration

---

Stable