OVERVIEW

PROCESS

DESIGN

IMPACT

YT Thumbnail Gen 1

Automated creation of thumbnails for content teams

For

Testbook

Role

Lead Product Designer

Timeframe

2 Weeks

Tool Used

FIGMA

PROJECT

Overview

Overview

Testbook, an ed-tech brand under Classplus, runs 24 YouTube channels managed by 12 YouTube Managers and 6 Thumbnail Designers.The design team was stretched thin, handling all channels—though only 8 were high-performing and needed premium attention.The remaining 16 underperforming channels required a consistent yet time-efficient thumbnail solution.

Objective

Empower YouTube Managers to independently create thumbnails for the 16 underperforming channels without needing design intervention, so the design team can:

Focus on the 8 top-performing channels
Improve quality, drive experiments, and optimize performance

Problem Statement

The 6 thumbnail designers were overloaded, creating visuals for all 24 channels, while 12 YouTube Managers (non-designers with basic Canva skills) lacked the tools to contribute. Thumbnails for underperforming channels were inconsistent and lacked quality. Designers couldn’t focus on innovation for high-performing channels.

When tasked with optimizing Testbook’s YouTube thumbnail process, I realized the challenge: While AI tools like ChatGPT and MidJourney were advancing, they couldn't generate high-quality, branded graphics aligned with Testbook’s guidelines. The complexity was further compounded by the need to empower non-designers to create consistent, professional thumbnails without relying on advanced AI or skilled designers.

Challenges

AI tools like ChatGPT and MidJourney couldn’t generate the precise, branded thumbnails Testbook required. While they provided abstract visuals or text suggestions, they fell short in delivering brand-aligned designs.

Key limitations:

Faculty Image Consistency: AI couldn’t maintain branded faculty images.
Custom Font & Text: AI struggled with precise fonts and overlay text.
Layout Precision: AI lacked flexibility to maintain consistent layouts essential for branding.

DESIGN

PROCESS

To build a scalable system that would empower YouTube Managers and reduce the load on designers, I started by conducting focused research across both user groups. The goal was to deeply understand their workflows, frustrations, and opportunities for automation or simplification.

Research

Empathize with Users

Methods Used

1. Thumbnail Audit (Across 24 Channels)

Collected ~200 thumbnails across all Testbook channels
Evaluated based on performance, visual consistency, design layout, color usage, and typography
Benchmarked the top-performing thumbnails to identify recurring patterns

2. Card Sorting & Categorization

Sorted thumbnails into logical categories:
- Subject-based (e.g., Maths, English, Reasoning)
- Faculty-specific
- Exam-specific (e.g., SSC CGL, CTET, Banking PO)
- Campaign types (Live Class, Marathon, Strategy, Tips)
Mapped recurring layout patterns, such as:
- Faculty on left, text on right
- Full image background with overlay text
- Split background for dual topics

Layout Sorting

3. Stakeholder Interviews

Interviewed 4 YouTube Managers and 3 Thumbnail Designers
Key Questions:
- “Where do delays happen in your process?”
- “What’s the most common kind of thumbnail you need to create?”
- “What would you do if you had a thumbnail template but no designer available?”
Tools Used: Miro, Google Meet, Notion for synthesis

4. Usability Testing on Canva & Figma

Observed YouTube Managers edit existing Canva templates
Measured time taken to:
- Change faculty image
- Modify title text
- Match brand color
Repeated the same test using Figma components with simplified override properties

Process

followed

Define

Users' needs and problems

Key Insights

Insight	Impact
Most managers can edit templates but not create layouts	Focused on giving them fully-built, swappable components
80% of thumbnails follow just 4–5 layout patterns	Prioritized building only those into the system
Designers were repeating the same layouts manually	Systemized these layouts into components
Thumbnails for 16 channels were low-effort but high-volume	Targeted these for manager self-service
Managers lacked clarity on which colors/fonts/images to use	Embedded branding directly into the Figma system

User Persona

YouTube Manager Persona at Testbook

Attribute	Details
Role	YouTube Manager
Background	Content/Marketing background, basic knowledge of Canva (can edit templates but not create designs)
Responsibilities	- Manage one or more YouTube channels - Coordinate with faculty, video editors, and scheduling teams - Upload and optimize video content
Pain Points	- Can’t create thumbnails from scratch, only edit templates - High dependence on designers for basic tasks - Struggles with branding consistency across thumbnails - Time-consuming manual processes, especially for 16 underperforming channels
Goals	- Empowered to create thumbnails for underperforming channels - Speed up the process and improve self-sufficiency - Improve performance by aligning thumbnails with brand and target audience
Quotes	“I can tweak Canva templates but I can’t make new ones from scratch.” “The design team is always busy, I wish I could do basic thumbnails on my own.”
Solution Fit	- Figma component system allows them to easily swap images and edit text - Empowered to work independently on 16 underperforming channels while designers focus on the top-performing ones

YouTube Thumbnail Designer Persona at Testbook

Attribute	Details
Role	YouTube Thumbnail Designer (Part of a 6-person team)
Background	Skilled in tools like Figma, Photoshop, and Illustrator; strong grasp of Testbook’s brand, style, and design needs
Responsibilities	- Design 10-20 thumbnails daily across multiple channels - Maintain consistency in design while experimenting with high-impact thumbnails - Source and edit faculty images, retouch thumbnails
Pain Points	- Overloaded with repetitive tasks for lower-performing channels - Time spent on repetitive design work - Lack of creative freedom due to workload - Constant requests from managers for basic thumbnails
Goals	- Focus on high-impact channels and experiments - Streamline the design process for repetitive thumbnails - Collaborate on more strategic campaigns, improving conversion rates
Quotes	“We’re stuck making the same 3 types of thumbnails all day.” “If managers could handle the basic ones, we could focus on more creative work.”
Solution Fit	- With the Figma-based system, managers now handle the repetitive thumbnail tasks - Designers can focus on creative and high-performance thumbnails, experimenting more

Ideate

Solutions

Research Method	Key Insight	Design Decision / Solution
✅ Card Sorting by content category (e.g., Strategy, Notification, Batch Launch)	Each category had a unique visual structure and urgency	Created category-specific layout components optimized for clarity and performance
✅ Thumbnail Audit	80% of thumbnails follow 4–5 recurring layout types	Focused only on high-ROI layouts to build as components
✅ Interviews with YouTube Managers	Managers can’t design but can edit Canva files	Built easy-to-edit Figma components using simple overrides
✅ Interviews with Designers	Time wasted on repetitive requests	Empowered managers to handle standard thumbnails for underperforming channels
✅ Usability Testing (Figma vs Canva)	Figma works if well-structured	Used component props: Text, Swap, Visibility, etc.
✅ Brand Audit	Inconsistent branding in lower-tier channels	Embedded brand tokens directly into styles and layouts
✅ Faculty Image Review	Low consistency in cutouts/names	Built and shared a ready-to-use, organized image repository

Rough Sketches for layouts from card sorting to get a idea of combination of layouts that are required to cover 80% of the use case.

FINAL

Designs

Hi- Fidelity

Design Mocks

Layouts

Layouts

AI assisted Image Generation

Component Library – System Design Overview

To create a scalable thumbnail system, I built a component library using the Atomic Design Principle. Each thumbnail was broken down into reusable units like backgrounds, faculty images, icons, and titles, making it modular and adaptable.

I leveraged Figma’s Instance Swap and Component Properties to allow non-designers to easily swap images, change layouts, and edit text without disrupting the structure. Layouts dynamically adjust based on content type, ensuring consistency without manual intervention.

This system was specifically built for Testbook tutors, focusing on practical components over decorative ones, unlike generic platforms like Canva. The library will evolve with changing content styles and needs, ensuring easy and impactful thumbnail creation for managers and tutors.

IMPACT &

Outcome

Impact

Created

YouTube Managers can now create thumbnails themselves for 16 channels
Reduced repetitive workload for the design team by 65%
Improved brand consistency across underperforming channels
Enabled designers to focus on creativity and performance for the top 8 channels

What's Next

Possible Upgrade

AI-Powered Thumbnail Generation

With the base design system in place and non-designers now equipped to create thumbnails manually, the next logical evolution is automation.
I’m currently exploring an AI-powered thumbnail generator, integrated with the existing Figma component system.

How It Would Work

Prompt-based input: YouTube Managers describe the video topic in plain English.
ChatGPT generates:
- Suggested thumbnail copy (based on trending hooks, emotional triggers)
- Relevant faculty image (if mentioned)
- Layout type recommendation (based on video type/category)
A Figma API script or plugin then:
- Selects the right component layout
- Inserts the generated text
- Swaps in the matching faculty image
- Outputs a thumbnail preview (export-ready)

Goals

Make thumbnail creation as easy as writing a prompt
Increase output speed while maintaining brand alignment
Personalize thumbnails to content without extra design input

Tools & Stack

ChatGPT for content generation and logic
Figma Plugin API to control component variants dynamically
Optional: Connect with Google Sheets or a CMS for batch generation

Next Project

Click to Claim

Back to work