Conker-ing quizzes with AI

TL;DR

TL;DR

TL;DR

I led the design of Conker, an AI-powered platform that enhances the quiz creation workflow for educators. Focusing on educational efficiency and curriculum alignment, Conker attracted over 500,000 users, resulting in more than 700,000 quizzes created.

The Problem

The Problem

The Problem

Educators struggle to efficiently create engaging, curriculum-aligned quizzes for diverse learning needs. Conker was developed to provide a faster, more effective solution.

Discovery

Discovery

Discovery

We started by engaging in surveys and interviews with educators. Given AI's novelty in education, we were keen to explore teachers' potential hesitations alongside identifying their primary challenges, particularly with regards to time constraints in quiz creation.

"I mean that sounds awesome. Anything that can save me time and still meet the curriculum is gonna be super helpful."

"My initial reaction is skeptical. I generally am not trustworthy of anybody trying to do my job for me."

INSIGHT

Our initial survey, involving 57 educators, revealed that they spend on average 2.5 hours preparing for and 3.5 hours marking formative assessments.

Building a Proof of Concept

Building a Proof of Concept

Building a Proof of Concept

Our insights led to a week-long hackathon in Brighton, culminating in Conker's proof of concept. From research came our focus: streamline the workflow using a Google Forms integration. This would give us an interface with basic functionality that fulfilled fundamental requirements.

A (very) cold Brighton Pier

Prioritisation of educator needs derived from initial discovery

Prioritisation of educator needs derived from initial discovery

Our competitor analysis from discovery showed that no other tool at the time was doing what we were aiming to achieve. We hypothesised that if we could create an easy-to-use generative AI tool that integrates with Google Forms, we could gain a first-mover advantage to then build off.

Initial user flow diagram

Low-fi wireframes used to build the first prototype

Low-fi wireframes used to build the first prototype

Iterating in the wild

Iterating in the wild

Iterating in the wild

Post-Brighton, we launched the product. We knew it wasn't perfect, but launching gave us a platform to test and gather real data. This feedback was instrumental in refining the UI/UX. Our mantra became 'learn quickly and iterate'

TRACKING SUCCESS

  • Key success metrics were tracked (Created -> Shared)

  • Analytics tools were used to inform decisions.

  • A product feedback form was placed in the product.

  • We invited active users for interviews based on their product interactions.

Qualitative and quantitative feedback provided the insight we needed to make quick decisions. An early pivot included alterations to how we presented the hierarchy of information on our home page. We put more emphasis on the creation of topic-based quizzes.

"I'd want to use my own topic; I don't have source material most of the time."

Usability test on question edit functionality

Usability test on question edit functionality

We iterated multiple times during the beta phase

We iterated multiple times during the beta phase

INSIGHT

Requests for tailored grade-level difficulty came up multiple times during feedback. This resonated with earlier assumptions we had made.

Heat map analysis showing choice paralysis on an iteration

Heat map analysis showing choice paralysis on an iteration

Through iteration, we soon ran into an unintended conundrum. Although topic-based quizzes increased our landing page quiz generated conversion, the share rate later down the funnel was underperforming compared to source-material quizzes.

Many teachers didn't have up-front source material when first visiting Conker. So, we tweaked our UI again to make it easier for them to add their own, whilst also providing ready-to-use content as an alternative.

INSIGHT

Requests for tailored grade level difficulty came up multiple times during feedback. This resonated with earlier assumptions we had made.

Switching the visual hierarchy from source-material first to topic-first drove up landing page → quiz generated conversion by 24% for first-time users.

Customisation of user added prompt

The call for complexity

The call for complexity

The call for complexity

One prominent insight that we kept hearing was that teachers sought more challenging questions, they required additional depth to match curriculum-based requirements.

We introduced more question types to ensure the questions posed could challenge students and foster critical thinking.

"The questions are kinda accurate but kinda stupid, like the first question was ‘What was the name of the professor, and the answer was professor"

INSIGHT

On a scale of 1 to 10, how important is it for you to factor in meeting curriculum standards when creating quizzes? 


Yeah we must meet the curriculum standards when creating quizzes, that’s kinda the main point. (10)

Alongside new question types, we developed the Conker question bank. A curated collection of complex, curriculum-aligned prompts that when triggered generated standards-aligned quiz content specific to the user's learning objective.

This became invaluable, especially for educators keen on tailoring their quizzes to specific classroom curriculum criteria. It proved another step in ensuring Conker was truly functional in real-world classroom settings.

Impact of question bank in Texas (TEKS)

Impact of question bank in Texas (TEKS)

Integrated question bank search

Create a quiz from question bank search

Create a quiz from question bank search

Explore question bank view from navigation

Explore question bank view from navigation

Integration and expanding reach

We recognised the potential of integrating with platforms already familiar to educators. A strategic choice was made early on to leverage Google Forms. It was relatively simple to implement, and meant a lot of the recipient experience was taken care of.

However, for truly wide scale adoption we required our own platform. This meant supporting share codes, quiz result scorecards and our own recipient experience.

Quiz Results Tab

Mobile adapted layouts

Final thoughts

Starting with a hackathon in Brighton and leading up to the current state of Conker, my journey has been thrilling, yet filled with complex challenges.

The success of Conker is evident not just in its growing user base but also in the 700,000 quizzes created using the platform. Additionally, the continual positive buzz on social media throughout the development highlights Conker’s broad appeal and impact.

Next

Next

Testing to drive mobile acquisition

the wall street journal

© 2024 Nathan Van Hesteren