[Product Roadmap] How upGrad followed a three-phased approach to close to a million individuals globally
For several years now, many reports, surveys and white papers have stated that engineers in the country are unemployed. Two years back, the HRD Ministry had pointed out that 48 percent of engineers in the country are unemployed. Pearson Research adds that 95 percent of Indian engineers are not equipped for development and coding jobs.
The pandemic has simply amplified the need for specialised skills. With many people facing job loss, furloughs and salary cuts, there is a growing need for upskilling. Though edtech seems to be the flavour of the season, the idea of helping people upskill themselves came to Mayank Kumar, Ronnie Screwvala, and Phalgun Kompalli in 2015, later in 2020, Rohit Dhar, and Puneet Tanwar joined the core team.
Recently, in June 2020, it earmarked an outlay of Rs 150 crore to expand courses offered through its platform, acquisitions and partnerships, and announced its collaboration with Jamia Hamdard University and O P Jindal Global University.
The early upGrad team
Explaining their journey, Mayank says when they started in 2015, they knew that technology had to be the backbone. They broke their journey down into three phases.
Mayank explains that in the first phase, the problem technology was solving was making content accessible. “If you want content from Ivy league - online education can offer that. But just making content accessible is not necessarily 'education', that is equivalent to 'content libraries'. While our technology platform made possible seamless delivery and access to best quality content, we believe the role of technology is a lot more and has evolved over time,” explains Mayank.
Phase II was about engagement and completion. Mayank says that they realised users’ engagement was very less across online education platforms, and as a result, course completion rates were also low. They focused on adding multiple elements of engagement and allowed the team to take “human touch” to scale.
“Further, technology helped us take out anonymity from online education. We brought a strong peer to peer learning element into education that made online learning at upGrad less anonymous and more participative. All these helped us drive high engagement and over 80 percent completion compared to sub 10 percent in the online education space in the early days,” says Mayank.
Now the team is working on phase III, focused on outcomes and personalisations. Since the pandemic has accelerated the growth of edtech, the team will now build personalised learning journeys and experiences that will help achieve strong outcomes for the learners.
The performance system
Initially, the team launched two programs for market validation using Canvas — a third-party Learning Management System (LMS). upGrad used Canvas as a base platform and customised the experience by adding discussion forums and visual elements on top of it, since most of the content was in video format.
“It took us about three months to customise experience on Canvas and launch the program. This was followed by the first version of a learning platform that we built in-house, which took another four to five months. Overall, it was about 10-11 months before we launched our first learning platform which was built completely in-house. We went with micro services architecture, initially. We used python and JAVA. For UI we used backbone.js,” says Mayank.
Evolution across phases
“We addressed the problem of accessibility in phase I (content access) by building a strong interactive platform that ensured we can deliver high quality content globally even if the bandwidths are low. While we solved for making content accessible, that was not the end goal for us,” says Mayank.
Initially, working professionals comprised the user segment with programmes in data science, technology, software development, digital marketing and others.
He explains, content accessibility for upGrad was easy stakes, which led them to phase II of high engagement and completion rates. To achieve this, the team focused on bringing in a lot of peer to peer engagement and leveraged technology to scale human intervention.
This allowed the team to take grading, mentoring, coaching and academic support to scale. And thus, achieve completion rates of around 80-90 percent (90 percent+ for corporate learners), the best in the industry.
“Once we achieve course completion and learner engagement in the platform, phase III was about delivering outcomes and personalisation to the learners in terms of career transition, growth acceleration and professional confidence,” says Mayank.
Basis the learner survey they had conducted recently, close to 79 percent of upGrad’s learners have attained meaningful career benefits out of all who were looking for professional growth, transitions, promotion, upskilling, and other career outcomes, bearing testimony to how they fare in terms of achievement of their goal.
“To elaborate, we are capturing about 100,000 plus data points per learner. These data points are allowing us to know, which skill are they good/bad at, what are their strengths/weaknesses etc. Depending upon what they are good at and what they are bad at, we design the right product interventions, so that they find their desired outcomes,” says Mayank.
For instance, any learner during his onboarding is asked on the career platform as to why they joined upGrad, what are they looking for in their career, where do they see themselves in five years etc. Accordingly, the team is designing their entire learning path. And these learning paths are unique for every learner.
Technology for all
“As we are now launching Bachelor’s and Master’s degrees in Arts, Science, Commerce etc., we are talking about students who are school or college pass-outs, who are yet to join the workforce and can be from Tier II and III markets. So, there’s the massy element of it coming into place that is meant for Bharat and not necessarily India,” says Mayank.
He explains that the infrastructure that working professionals have access to in terms of connectivity, horsepower and devices is starkly different from the access that Bachelor’s & Master’s degree audiences have.
“We cannot expect the same quality, and hence the same technology cannot provide them the optimum experience. So now, we are building in new features and making improvements that lighten our stack and make it more resilient to difficult infrastructure environments. So, if somebody has poor connectivity or has poor quality device, we will still be able to deliver results owing to our robust technical capabilities,” says Mayank.
The next focus is on the learning experience itself. This, Mayank explains, can be further broken down into three sub-segments, which are: recorded, live, and career centre.
The recorded content includes latest videos as well as existing content repurposed with the help of technology. Mayank says the team has effectively applied gamification to make the recorded experience interesting for learners, which allows them to understand what they have completed and what they have not, along with text content in a video format.
This gives a ‘context of content’ which in turn emulates classroom learning as close as possible, thereby ensuring class participation, better ranking and meeting individual targets.
The second sub-segment is the live learning that enables real-time doubt resolution, attendance tracking, interactive live classes, in-class quizzes, polls, and also measures learner efficiency using detailed classroom analytics. Introduction of hybrid chat and group coaching strengthens peer-to-peer interaction online.
“And then we have the third sub-segment which is careers and outcomes, which was not there during the phase I of our growth. The career platform is a one-stop solution to drive career engagement and increase career readiness of learners by keeping a constant track of their performance and other tasks including the ones allotted by the career coach,” says Mayank.
The team has also added a layer of support function, where technology helps make internal process far more efficient to be able to scale out operationally. Personalised features like just-in-time interview preparation, resume building and counselling sessions, etc. which requires human intervention are added as features in the platform to help learners achieve the desired career goal.
Building on flexibility
As the platform started scaling, they realised that the business problem was that different universities have different requirements. Citing an example, Puneet Tanwar says the team first launched an entrepreneurship programme, where learners would pitch their business ideas, receive feedback on their business plans, and work on case studies.
Whereas, the second programme was a technical one on data science, where learners used to end up spending a lot of time writing codes, building data science models, writing SQL queries, and coding in the platforms.
“Next, we launched courses in digital marketing, product management, etc. again catering to widely different requirements. Hence, we needed to build in flexibility in the MVP. We built loosely coupled micro-services. For launching any new product, we would just add a feature or a service, on top of our existing system, without really touching our existing code base. Coding languages we used were JAVA, python and on the front end, we used backbone.js,” says Puneet.
The next was the ability to add different features. After they launched their learning platform, Puneet explains working on an add-on feature that would allow their internal graders to grade and provide quick feedback to learners on their submissions.
“This was us planting the initial seeds for personalisation, which today is our biggest focus,” says Puneet.
Working around challenges
The feedback on the first version was a mix of positive and healthy-negative. Learners appreciated the clean UI, UX of the platform look. There's no distraction and hence, learners were really able to focus on the content.
“The negative was that many felt it was too rigorous. We had purposefully built a rigorous programme to impart learning that will create a meaningful impact on the learners’ life. Given that our users were working professionals, we realised it was a real challenge to balance rigour,” says Rohit Dhar.
Thus, in the second leg, they created a Learner 360 platform. It is a data rich platform where personal student mentors assigned to individual learners can get a 360-degree view of the learners’ profiles and track their progress. If a learner is lagging behind or finding it difficult to manage time, mentors provide deeply personalised advice, based on the learners’ background and progress so far, making it a technology powered platform solving the consumers’ pain point.
While Learner 360 was facilitating learner - mentor interactions, the next level of feedback was to solve for the lack of peer to peer learning, one of the biggest ‘teachers’ in an offline setup. This feedback charted the path for building communities.
"We developed discussion forums. For every cohort, we started with a blank discussion forum, that in a way, encouraged learners to post more and more queries and not hesitate. Through 2016 to 2018, we were delivering on strong post course competition NPS (Net Promoter Score),” explains Rohit.
The team then decided to do an anonymous NPS for those who had completed the course six months prior. And that was an eye-opener. What they realised was that the immediate NPS was high due to the satisfaction with the education delivered. However, after some time, if the learner had not received tangible benefits from the course or any ROI, then the overall perception about the experience fell through.
Thus, the team focused on delivering outcomes.
“We launched an AI-backed job-board which provided: a) Unified job stream, a one-stop-shop to see job listings from upGrad's hiring partners + third party recruiter jobs + curated jobs from across the internet. b) Central tracker to provide seamless tracking of learners’ job applications, both on upGrad platform + any other via a chrome extension and self-logger,” says Rohit.
Currently, the team is focused on driving personalised outcomes. This also included personalised mentoring and one-on-one training and coaching sessions that have been built using technology. Plus, the tool is providing personalised job recommendations for a learner, thus facilitating career transitions.
“When we started, our Content Management System (CMS) was half-baked. Since the developers had to manually code, add content, videos, photos, etc. and program it live on the platform, it took about three weeks’ time to launch a programme. In addition to the time, the manual to and fro between the tech team, programme team and university stakeholders took significant coordination effort,” says Rohit.
In order to solve this problem, upGrad built a superior CMS that facilitated dynamic launches. Now without a single line of code written by a developer, without any involvement of engineers, new programs can be launched. This brought in agility.
“From launching five programmes per quarter, we’re now launching 50 programmes per quarter. The turnaround time taken to launch new programmes have been brought down from three weeks to two days and now we strive to do it within a few hours,” says Rohit.
The world of upskilling
Today, upGrad has several competitors focused on upskilling including, , , , and others. upGrad has got the growth they need, now they are looking at building a deeper experience.
“When you look at how technology can create deep personalisation and engagement - the canvas for applying cutting edge innovation from various fields becomes limitless. Big data and artificial intelligence (for personalised insights and interventions), voice assistants (for personal productivity), facial recognition (for measuring and improving engagement), Virtual Reality (for immersive learning experiences), simulation platforms (for simulation based learning), blockchain (blockcerts etc.), and virtual instructors (for real-time Q&A) are some examples,” says Puneet.