---
name: anomaly-variance-explanation
description: Explains material changes in a business metric by decomposing period, volume, price, mix, and data-quality effects, then ranks evidence-backed causes in a cited analysis sheet and management memo. Use for KPI variance reviews, monthly business reviews, forecast misses, anomaly investigation, and driver analysis.
license: Apache-2.0
metadata:
  adlass.categories: "data-tables/analysis-commentary"
  adlass.industries: ""
  adlass.tags: "anomaly,variance,driver-analysis,kpi,forecast-miss,root-cause"
  adlass.adaptation: "mapping"
  adlass.source: "original"
  adlass.version: "1"
---

# Anomaly variance explanation

## Purpose

Explain why a reported metric differs materially from its comparison value. The output separates arithmetic drivers from contextual evidence, ranks causes by quantified contribution, and gives a reviewer a traceable basis for follow-up.

## Scope

Analyze one metric, comparison period or plan, and the dimensions that explain its movement. Cover volume, rate or price, mix, timing, data completeness, and documented operational events. **Excluded:** changing source data, inventing a target, forecasting beyond the supplied evidence, or assigning corrective work to people.

## Data basis

- KPI table with `metric`, `period`, `actual`, `plan_or_prior`, `dimension`, `unit`, and `source_row_id`.
- Transaction or event table with `date`, `quantity`, `price_or_rate`, `category`, `customer_or_account`, and `status`.
- Forecast, budget, calendar, launch, outage, or pricing documents in scope.
- Metric definition and materiality threshold supplied by the company.

## Result

Produce a variance bridge sheet with one row per metric segment and a memo that states the absolute and percentage variance, quantified drivers, evidence citations, confidence, and unresolved data gaps.

## Quality criteria

- The bridge reconciles actual minus comparison to the sum of classified drivers, subject to a stated rounding tolerance.
- Every driver names its formula, source row or document page, and confidence level.
- The same period, unit, currency, and population are used on both sides of the comparison.
- Timing effects, excluded rows, and unproven hypotheses remain visibly separate.
- No explanation is called causal without supporting evidence.

## Instructions

Use the metric definition and company materiality rule before interpreting movement. Calculate absolute change as `actual - comparison` and relative change only when the comparison denominator is non-zero. Decompose revenue-like measures into volume, rate, and mix only when the source fields support that decomposition; otherwise label the residual. Rank explanations by quantified contribution, then evidence strength, and label a contribution as “hypothesis” when it is supported only by correlation or a dated event note. Cite metric-table rows for figures and document section or page for contextual claims. Preserve negative, zero, and missing values rather than converting them to blanks.

## Adapt before use

- Map KPI, transaction, dimension, date, unit, and source-ID columns to the company data model.
- Add the metric definition, comparison hierarchy, materiality threshold, and rounding tolerance.
- Define currency conversion, timezone, fiscal calendar, and treatment of cancellations or late postings.