Expanded KPI Computation Graph for Non-Monotonic Change Attribution

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

General computing systems are incapable of identifying the source of changes in metrics defined by non-monotonic functions executed on derived key performance indicators (KPIs) due to the nested nature of the functions, which obscures the impact of individual dimension values on the overall metric.

Innovation Solution

The implementation of an expanded data structure and computation graph that breaks down the metric into subsets associated with different dimension values, allowing for the determination of change indicators through derivatives, thereby identifying the source of changes in the metric.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a non-monotonic function is executed on derived KPIs to calculate a metric, then the metric can capture complex business trends and anomalies, but the nested nature of the function obscures the impact of individual dimension values on the overall metric

Engineering Contradiction:

Improvemetric accuracyVSAvoiddimension value impact

Core Design Contradiction:

Measurement precisionVSLoss of information

Solution Approach 1:

The patent segments the nested non-monotonic function into multiple computational layers, representing each layer as separate nodes in a computation graph. This segmentation allows the system to track how changes in individual dimension values propagate through each layer to affect the final metric, thereby preserving information about dimension value impact while maintaining metric accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces a new dimensional representation by expanding the data structure to include not only the metric value but also attribution information linking metric changes to specific dimension values. This dimensional expansion allows the system to simultaneously store both the precise metric calculation result and the source attribution information.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Loss of information

If the metric is calculated using nested functions on derived KPIs, then comprehensive business insights are obtained, but identifying the source of metric changes becomes difficult or impossible

Engineering Contradiction:

Improvebusiness insightVSAvoidchange source identification

Core Design Contradiction:

Loss of informationVSDifficulty of detecting and measuring

Solution Approach 1:

The patent implements a feedback mechanism through the expanded data structure that captures attribution information at each computational layer. When the metric is recalculated after data changes, the system can trace back through the computation graph using stored attribution data to identify which dimension values caused the metric change, providing feedback about the source of changes.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The computation graph serves as an intermediary structure between the nested non-monotonic function and the final metric output. Each node in the computation graph represents an intermediate calculation step and stores information about its inputs and outputs, allowing the system to trace the origin of metric changes through this intermediary representation.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Device complexity

If traditional data structures are used to store metric data, then storage is simple, but the system cannot identify which dimension values are driving changes in the overall metric

Engineering Contradiction:

Improvedata structure simplicityVSAvoiddimension value attribution

Core Design Contradiction:

Device complexityVSLoss of information

Solution Approach 1:

The patent applies the nested doll principle by creating an expanded data structure where the metric value contains nested attribution information, which in turn contains nested dimension value references. This nested structure allows the system to maintain a compact representation while embedding multiple levels of detailed information about data sources and their contributions to the metric.

Inventive Principle:

Principle #7Nested doll (Nesting)

Data Source

PatentUS12547526B1Programming improvement using expanded data structure to identify a source of a change in a metric inferred from a non-monotonic function

Publication Date: 2026.02.10 INTUIT INC

AI summary

A method including receiving a processor command to identify a source of a change in a metric. The metric includes a non-monotonic function executable on a derived key performance indicator (KPI) derived from a base KPI, which includes a combination of measured data for different dimension values. The is represented metric as a computation graph in terms of the base KPI for each of the dimension values. Representing includes generating an expanded data structure that stores the base KPI as a number of subsets of the measured data. A number of change indicators in the metric is generated for the number of different dimension values by determining a corresponding derivative of the computation graph with respect to each of the number of subsets of the measured data stored in the expanded data structure. A change indicator is identified as being the source of the change in the metric.