Work Item Sizing Prediction Using Team-Specific Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inaccurate work item sizing in software development leads to missed deadlines, revenue loss, and impaired client trust due to reliance on senior engineers' familiarity with project details, and the diversity in language and writing styles complicates consistent estimation across teams.
Innovation Solution
A system utilizing natural language processing (NLP) to extract keywords from feature requests and correlate them with team-specific keywords, generating vectors for similarity analysis and weighted averaging to estimate work item sizing accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If senior engineers perform work item sizing based on their familiarity with project details, then estimation accuracy is improved, but reliability deteriorates when they are unfamiliar with specific subjects
Solution Approach 1:
The system enables self-service estimation by automatically analyzing feature request descriptions, extracting keywords, comparing them with historical work items, and generating size predictions without requiring senior engineer intervention. This automated approach ensures consistent application of estimation criteria across all work items regardless of engineer familiarity with the subject matter.
Solution Approach 2:
The system introduces an intermediary computational layer that mediates between feature request descriptions and estimation outcomes. This intermediary automatically processes language variations, extracts meaningful keywords, and compares them against historical data, eliminating the variability introduced by different engineers' familiarity levels with specific subjects.
2Adaptability or versatility
If manual estimation by senior engineers is used, then adaptation to project requirements is improved, but productivity deteriorates due to time-consuming processes
Solution Approach 1:
The system replaces the manual mechanical process of senior engineer estimation with an automated computational system. The computational system processes feature descriptions, extracts keywords, compares with historical data, and generates predictions rapidly, maintaining adaptability to project requirements while dramatically increasing estimation speed and throughput.
Solution Approach 2:
The system changes the parameters of the estimation process from human cognitive factors (engineer familiarity, subjective judgment) to computational parameters (keyword extraction, vector similarity, historical data matching). This transformation maintains adaptability while enabling rapid processing of numerous work items.
3Adaptability or versatility
If diverse language and writing styles are accommodated in feature requests, then adaptability is improved, but measurement precision deteriorates due to inconsistency
Solution Approach 1:
The system transforms the parameter of language expression from diverse human writing styles to standardized computational representations through keyword extraction and vectorization. This allows the system to accommodate language flexibility in input while achieving consistency in the processed representation used for comparison and prediction.
Solution Approach 2:
The system introduces language processing intermediaries that mediate between diverse feature request descriptions and the estimation engine. These intermediaries extract meaningful keywords regardless of writing style variations and transform them into consistent vector representations, preserving adaptability to different expression styles while ensuring measurement consistency.
Data Source
AI summary
Method and apparatus for work item sizing prediction are provided. A feature request is received. A plurality of feature keywords are extracted by processing descriptions of the feature request. A plurality of team-specific keywords are identified for a work item associated with the feature request. A work time vector representing the work item is generated using the team-specific keywords. A plurality of prior work items that are related to the team-specific keywords are identified. A plurality of prior work item vectors are generated, where each respective prior work item vector corresponds to a respective prior work item, among the plurality of identified prior work items. A similarity score between the work item vector and each of the prior work item vectors is calculated. A time to complete the work item is estimated based on the similarity score.