Teachify

Can we better match tutors and students?

Team

Matthias Lee
Cephas Yeo
Carina Chu
Oishin Wong
Nada (me!)

Contributions

User Research
Data Analysis

Results

Awarded Best Value Proposition award for Design Odyssey DRIVE Bootcamp 2022

Context

Teachify aims to create a digital solution for unequal access to tuition and high barriers of entry in becoming a tutor, through creating a circular system that curates teaching experience and verifiable credentials through volunteering experiences. Supported by a platform that provides student-tutor matching, review system and upskilling resources, Teachify strives to be a one-stop solution for all things tuition related.

In this case study, we focus solely on the types of questions and inputs to curating the a good tutor-tutee match. As such, we won’t be testing this via UI designs (of Teachify!) as we want to abstract away the platform that provides the matching service, but instead focusing truly on how the matching is being done.

Exploratory Research

We conducted a series of user interviews with tutors, parents and students to try identify their painpoints. In this case study, we mainly focus on the students & parents’s painpoints.

Focus

We realised that the current matching services lack depth which creates high turnover rates for tutors found via tutor agencies.

While we highlighted several other painpoints which we explored in other aspects of our work behind Teachify, we wanted to also focus on how the matching is being done to try promote matching services as a way of finding our tutors instead of solely relying on word-of-mouth*.

*Out of 25 parents interviewed, 13 rely solely on word of mouth for finding tutors, with only 9 parents having tried using tutor matching services.

Solution Exploration

Competitor Research

In our competitor research, we sent the same student profile with academic information adapted from this school’s data. This profile was sent out to multiple tutoring agencies to try investigate the type of information they require before curating a match, and the type of tutor profiles given out. This allowed us to get a better understanding of the core or basic information that seems like a necessity in these profiles or pre-matching process.

Within this study, we also looked into submitting multiple student profiles to the same agency to see how the response will differ. While this study wasn’t extensively done with all the tuition agencies, we could extract that the deterrence from tutor matching services was that it lacked personalisation and curation— no matter how different the student profiles were, we got the same teacher recommendations as long as they’re within the same level of study, subjects and gender.


To give a more meaningful analysis of these responses, we sat in a small focus group of 4 parents and shared our findings with them. In general, they shared similar sentiments of:

[01]
Profiles are quite generic, not much personalisation
[02]
Usually just back and forth with personal preferences via chat before getting a suitable match
[03]
Goes through many trial lessons to see if it match
[04]
As a parent, hard to pinpoint what will make it compatible for child, relies on child for approval of tutor

Will making these changes to tutor matching
efforts make it more effective ?

Hypothesis

Introducing more well-rounded factors when considering matches will allow us to better adapt our matching algorithm based on the top features that a student or parent looks for in a tutor, and also expand on tutors not just being there to teach, but also be a good role model**.

** Why is this important?
Teachify as a project has 2 main drivers: one of which is to streamline and improve the current tutor landscape, and secondly, to improve the system to be more inclusive and create more access in education. For us, this also means provided better mentors to teach you not only academics, but also beyond that— to truly find a holistic match.

Proposed Changes

[01]
Initial sign up questionnaire to include more well rounded information
[02]
Instead of a standardised profile template used for replies, curate the layout based on top features.
[03]
Curated Recommendations

Developing our own Matching Algorithm

We interviewed 11 tutors with their 31 respective tutees (each tutor brought in at least 2 students), to try gather data on certain student preferences and seeing if that matches how their current tutor describes themselves. For this, we’ve used the students’ satisfaction score as a benchmark factor to measure the relevance of their inputs.

After some calculations and data analysis, we soon realised that the following factors have quite a strong correlation:

[01]
Hobbies + Interest
[02]
Lesson + Teaching Style
[03]
Preferred type of working relationship (ie: Professional or friend-like)

As such, other than accounting for the core information such as background and subject compatibility, we’ve also factored in these features into our matching algorithm. After several rounds of refinement, we’ve tested this matching system with the 14 parent-child pairs.

For each parent-child pair, we’ve asked them to each complete an onboarding questionnaire which is hosted on Typeform. Using their inputs, we responded with 6 profiles :

[01]
2 being with the standardised information given out as per Competitor A’s profiles
[02]
2 being with the additional factors that we’ve considered (aka with our matching algorithm)
[03]
2 being random profiles (acted more as a control, to allow participants to utilise the 1-10 scale and give lower rankings)

For the sake of the comparison test, all profiles were presented in a similar layout as shown above to remove any bias created based off visual presentation. We’ve asked the parent-child pair to separately rate each profile from 1-10 based off confidence levels of compatibility, and have asked them to rank their 6 choices.

Outcome

While the confidence level remained somewhat within a similar range for parents*, we noticed that there was a 21% boost in confidence levels for compatibility for students when it came to profiles including the newly introduced features of our algorithm. 71% of the profiles with our algorithm was also ranked in the top 2.

*there was only a 3% increase which we forgo as it might be random distribution!

Designing an Onboarding Experience

While we mainly covered the research behind Teachify, the project was quite an extensive one– ranging from concept research, product pitches, and even some UI prototyping!

This short snippet focuses on designing our onboarding & initial user experience— which inherently captures the way we yield and incentivise user adoption.

Fun fact: this (project) was my first time tinkering around with UI design– and I'm glad I did :~)