Automated Natural Experiment Detection for Causal Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Businesses face challenges in efficiently identifying and analyzing natural experiments within their networks, as existing methods such as econometric modeling test for correlation rather than causation and manual detection is time-intensive and error-prone, making it difficult to assess the impact of business initiatives on key performance metrics.

Innovation Solution

A computer-implemented method that automatically detects natural experiments by analyzing historical data to identify changes in metrics across locations, determining the length and consistency of these changes, and selecting the best subset of test and control locations to generate a ranked list of experiments for analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If econometric modeling is used to assess historic changes, then correlation can be tested, but causation cannot be established

Engineering Contradiction:

Improvecorrelation testing accuracyVSAvoidcausation determination

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent segments the business network into test entities and control entities based on whether they experienced the initiative. This segmentation allows for comparing groups that differ only in exposure to the initiative, enabling causation determination rather than just correlation testing.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent creates a control group that copies the characteristics of the test group in all respects except for exposure to the initiative. By finding control entities that are similar to test entities but did not receive the initiative, the system can isolate the effect of the initiative itself, establishing causation.

Inventive Principle:

Principle #26Copying

2Reliability

If manual detection of natural experiments is performed, then experiments can be identified, but the process is time-intensive and error-prone

Engineering Contradiction:

Improveexperiment identification accuracyVSAvoiddetection time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces the manual mechanical process of detecting natural experiments with an automated computer-implemented system. The system automatically scans historical data, identifies entities that experienced initiatives, finds matching control entities, and generates experiment descriptions without human intervention, eliminating time consumption and human error.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system performs self-service by automatically detecting natural experiments using its own algorithms and data processing capabilities. It independently identifies test entities, selects control entities, determines experiment parameters, and generates reports without requiring manual analysis or external tools.

Inventive Principle:

Principle #25Self-service

3Reliability

If actively designed tests are executed, then causal insights can be obtained, but significant resources are required

Engineering Contradiction:

Improvecausal insight accuracyVSAvoidresources required

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The patent converts the harm of uncoordinated store-level variations into a benefit by automatically detecting these variations as natural experiments. Instead of viewing decentralized initiative implementations as problematic, the system leverages them as valuable data sources for causal analysis, eliminating the need for resource-intensive designed tests.

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

Solution Approach 2:

The system serves multiple functions: it identifies natural experiments, selects control entities, determines experiment parameters, and generates analysis reports. This multi-functional approach consolidates what would otherwise require separate manual processes and resources into a single automated platform.

Inventive Principle:

Principle #6Universality (Multi-functionality)

4Loss of information

If store-by-store variations are tracked manually, then natural experiments can be discovered, but the process is complex and error-prone

Engineering Contradiction:

Improvenatural experiment discoveryVSAvoiddetection process complexity

Core Design Contradiction:

Loss of informationVSDevice complexity

Solution Approach 1:

The patent replaces complex manual tracking and detection processes with automated computer algorithms. The system automatically scans historical data, identifies variations in entity characteristics, matches test and control entities, and discovers natural experiments without human intervention, simplifying the process while improving accuracy.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS10776738B1Natural experiment finder

Publication Date: 2020.09.15 APPLIED PREDICTIVE TECHNOLOGIES

AI summary

The methods and systems described herein attempt to address the above-mentioned need by providing an algorithm and software capability to automatically detect and assess “natural experiments” that exist in any underlying dataset. In one embodiment, a computer-implemented method for identifying an experiment to use in an analysis of a business initiative comprises storing, by a computer, a set of historical data regarding entities in a business network; receiving, by the computer, a selection of inputs for the historical data; detecting, by the computer, a natural experiment based upon changes in the historical data; and outputting, by the computer, a report of a detected experiment from the historical data.