docs/upskilling/prd/Assessment Service.docx.md

GrowQR

ASSESSMENT SERVICE

Business Requirements Document • v2.0 • March 2026

Upskilling Module • Microservice Specification Integrates with: Roleplay • Interview • Course • Pathways • Dashboard

Document Title	Assessment Service — Business Requirements Document

Service Type	Standalone Microservice (Part of Upskilling Services)

Version

Integration	Shares data with Roleplay, Interview, Course services + Pathways + Dashboard

Third-Party LMS	Moodle, Thinkific (no proprietary LMS built)

1. SERVICE OVERVIEW

The Assessment Service provides AI-generated, role-specific skill assessments structured around a three-box engine: Input, Processing, and Output. Assessments are micro-format (3–4 minutes), designed for engagement and measurability. The platform integrates with third-party LMS platforms (Moodle, Thinkific) and does not build a proprietary LMS.

1.1 Three-Box Engine Architecture

All assessments flow through a standardized three-stage engine:

1.2 Core Features

• Three-box engine: Input (corporate/institutes/LLM) → Processing (intent + tier) → Output (standardized report)

• Micro-assessments: 3–4 minutes per session for high engagement and completion

• AI-generated role-specific questions (21 per assessment); GenAI rephrases for binary clarity

• Difficulty levels: Easy, Medium, Hard (adaptive or fixed per question)

• Multiple choice format (4 options per question), timed with countdown timer

• STAR schema evaluation for situational/behavioral questions

• Certificates awarded for passing score (70%+) on eligible assessments

• Fallback pre-built question bank when AI generation fails

• Third-party LMS integration: Moodle, Thinkific (no proprietary LMS)

• Recruiter and university posting capability via corporate input channel

1.3 Value Proposition

Small, focused assessments help college students and professionals measure fit for roles, motivation, culture alignment, and company values. The question bank concept enables HR integration and recruiter-led workflows. Assessments directly feed into Q-Score improvements and job opportunity matching.

2. FUNCTIONS

FUNCTION 1: Assessment Generation

2.1.1 Generation Sources

Assessments are generated from two input channels:

1. Corporate Input: Recruiters or universities post role-specific assessments directly to the platform. Questions are pre-authored and stored in the question bank.

2. LLM Input: AI generates 21 questions tailored to the user’s role, covering technical skills, behavioral scenarios, and domain knowledge. GenAI rephrases questions for binary clarity where needed.

2.1.2 Assessment Design Principles

• Micro-format: Each assessment completes in 3–4 minutes to maximize engagement

• Questions must be clear and binary where possible; GenAI rephrases ambiguous questions

• Situational questions evaluated using STAR schema (Situation, Task, Action, Reaction)

• Background verification: Situational responses cross-referenced against user profile data

• Communication style evaluated alongside content accuracy

• Open-ended questions are not used — monitoring and scoring reliability is insufficient

2.1.3 Data Requirements

Field	Description
Assessment ID	Unique identifier
User ID	Who is taking the assessment
Role / Title	Job role (e.g., Python Backend Developer)
Input Source	Corporate (recruiter-posted) / LLM (AI-generated) / Hybrid
Question Bank	Array of 21 question objects
Question Object	Question text, 4 options, correct answer, difficulty, topic, STAR tag
Time Limit	Total time allowed (e.g., 20 minutes)
Tier	User subscription tier — determines question depth and volume
Started At	Timestamp when assessment started
Status	Generating / Ready / In Progress / Completed

2.1.4 States

State	Backend Response
Generating	Show: 'Generating Assessment… Our AI is crafting unique questions for this role.' (10–15 seconds)
Ready	Questions generated; return assessment ID + start link
AI Error	AI generation failed → Use fallback question bank; flag question source in UI
Invalid Role	Role not recognized → Ask user to specify or use generic assessment

FUNCTION 2: Taking the Assessment

2.2.1 Assessment Flow

• User starts assessment → Timer begins (countdown: 19:59, 19:58…)

• Questions displayed one at a time: 'Question 3 of 21'

• User selects one of 4 multiple choice options (radio buttons)

• User can navigate: Previous / Next Question

• Difficulty badge shown per question: 'Medium Difficulty'

• STAR-tagged questions display scenario context before answer options

• User can submit early OR timer expires → Auto-submit

2.2.2 Measurement Framework

For situational / behavioral questions, the platform applies the STAR schema:

Component	Stands For	How Evaluated
S	Situation	Cross-referenced against user’s profile and background data
T	Task	Verified for role-relevance and complexity level
A	Action	Evaluated for decision quality and communication style
R	Reaction	Assessed for outcome awareness and self-reflection

2.2.3 Data Captured

• User answers: Array of selected options per question

• Time per question: Seconds spent on each question

• Navigation pattern: Sequence of question views (for analysis)

• Submission type: Early submit / Timer expired

• Completion time: Total time taken

• Score: Number correct / 21 (percentage)

FUNCTION 3: Results & Analysis

2.3.1 Standardised Output Report

Every completed assessment produces a well-defined, standardised report. Outputs feed into Q-Score calculations and job opportunity matching pipelines.

• Overall score: Percentage (e.g., 85%)

• Score by topic/category (e.g., 'Python Basics: 90%, Backend: 75%')

• STAR schema breakdown for behavioral questions

• Communication style rating (derived from scenario responses)

• Correct vs. incorrect breakdown with explanations

• Time spent per question

• Comparison to average user in same role

• Recommended focus areas (topics with low scores)

• Q-Score contribution: Which Q-Scores are updated based on results

• Job opportunity flag: Whether results unlock specific job matches

2.3.2 Certificate Generation

• Certificate issued when: Score >= 70% AND assessment is certificate-eligible

• Practice / diagnostic assessments are not certificate-eligible

• Certificate includes: User name, Assessment title, Score, Date, Certificate ID

FUNCTION 4: Third-Party LMS Integration

GrowQR will not build a proprietary LMS. All courseware and assessment delivery integrates with established third-party platforms.

Platform	Integration Type	Use Case
Moodle	API / Webhook	Open-source LMS for institutional and corporate assessment delivery
Thinkific	API / Embed	Course and assessment hosting for third-party content creators
Internal Fallback	Native	GrowQR question bank used when no external source is available

3. DATA SHARING WITH OTHER SERVICES

Target Service	Data Shared	Purpose
Course Service	Low-scoring topics, skill gaps identified	Recommend courses to address weak areas
Interview Service	Technical knowledge gaps, topics scored low	Tailor interview questions to validate weak areas
Roleplay Service	Behavioural/scenario performance, weak competencies	Suggest roleplay scenarios for practice
Pathways Service	Assessment completion, Q-Score deltas, role fit scores	Update weekly cycle priorities and pathway recommendations
Dashboard Service	Assessment count, scores, certificates earned, focus areas	Update Q-Score, show progress, display achievements
Job Matchmaking	Role fit score, Q-Score contributions, skill signals	Surface relevant job opportunities post-assessment

4. TECHNICAL REQUIREMENTS

4.1 Performance Requirements

Operation	Target SLA
Question generation (21 questions)	< 15 seconds
Assessment start (after generation)	< 1 second
Question navigation	< 200ms
Score calculation	< 3 seconds after submission
Report generation	< 5 seconds
Certificate generation (if eligible)	< 3 seconds
LMS sync (Moodle / Thinkific)	< 10 seconds

5. ACCEPTANCE CRITERIA

#	Must Pass	Pass / Fail
1	Three-box engine operates correctly: Input (corporate/LLM), Processing (intent + tier), Output (standardised report)	[ ] PASS [ ] FAIL
2	AI generates 21 questions within 15s. Fallback question bank activates on AI failure.	[ ] PASS [ ] FAIL
3	Micro-assessment format completes within 3–4 minutes for standard question sets.	[ ] PASS [ ] FAIL
4	GenAI rephrasing produces binary, clear questions. No open-ended questions in scored flow.	[ ] PASS [ ] FAIL
5	STAR schema correctly applied to situational questions. Background cross-reference works.	[ ] PASS [ ] FAIL
6	Timer counts down correctly. Navigation (Prev/Next) works. Auto-submit on timeout.	[ ] PASS [ ] FAIL
7	Score calculated correctly. Topic breakdown accurate. Comparison to average shown.	[ ] PASS [ ] FAIL
8	Standardised report generated with Q-Score contributions and job opportunity flags.	[ ] PASS [ ] FAIL
9	Certificates generated for eligible assessments with score >= 70%. Download works.	[ ] PASS [ ] FAIL
10	Moodle and Thinkific integrations functional for corporate/institutional assessment delivery.	[ ] PASS [ ] FAIL
11	Skill gaps and Q-Score signals shared with all connected services correctly.	[ ] PASS [ ] FAIL
12	Recruiter and university posting workflow functional via corporate input channel.	[ ] PASS [ ] FAIL