My 2026 So Far — What We Fixed on the NIAT Website and the Learning Portal, and What the Numbers Say

In January, the marketing team brought a clear problem. Visitors were barely scrolling: page depth was only about 30%, half of them dropped off by the 20% scroll mark, and only 37% of clicks went to Apply Now. The site had recorded more than 1.1 million unique visitors and 2 million pageviews the year before, so small gaps mattered. Instead of replacing the live page, we ran an A/B test in Framer. The new version cut the page from 9 sections to 5 and added an Apply Now button halfway down, and UTM parameters followed every click all the way to the application portal so Google Analytics could tell the two versions apart. Over about 20 days, the old page turned 4,647 views into 806 clicks (17.3%), while the new one turned 4,856 views into 1,021 clicks (21%). That is a 21.2% uplift, and it became the homepage before the March peak.

The NIAT website never stands still. New universities join, admission statuses change, and legal and compliance text gets updated, and each change comes in as a small request that cannot wait. In 2026 we shipped internal and external pages for many universities, updated the university count to 30+ across the site, and refreshed the startup, outcomes, curriculum and parents’ testimonial sections. We also built the GRIT pages for tracks, skills, levels and miles, built the NIAT Tech Decode site (the former Masterclass), and kept the terms, refund and compliance pages current. The biggest behind-the-scenes fix was updating UTM tags across 354 links, so every click can now be traced back to where it came from. There is no single headline number for this work, but a website that parents read before they trust it has to be right, and it is.

In January, about 26,600 learners started 1.12 million practice questions on the portal. 3.6% of attempts ended with the student peeking at the answer, and 79.5% of first attempts were correct. This year we added a Match the Following question type, upgraded the Question Bank with difficulty filters, search, filters that stay after a refresh and automatic access for every NIAT plan learner, and redesigned the SQL practice screen to show one question at a time with the tables always visible. By September, 32,400 learners had started 1.30 million questions, answer peeking had fallen to 1.1%, and first-attempt accuracy had risen to 82%. Active learning time, the time students spend doing and not just watching, also rose from 11.8% of learning time in April to 14.0% in June, which is when we started measuring it. Content, the AI Tutor and the interface all changed at once, so I cannot give design the whole credit, but this is the direction we wanted.

At the start of the year, the AI Tutor was used by about 2,300 learners a month, in about 10,000 conversations. We kept working on it as an adaptive coach that sits right inside the question and unit screens, where students get stuck, and not as a chatbot in a corner. By September, 12,842 learners were using it in 54,427 conversations in a single month, about five and a half times more of both. The tutor inside questions grew from about 2,000 to 10,800 learners, and the tutor inside units grew from 442 to 4,671. Some of that growth comes simply from more students reaching it, and I cannot split exactly how much, but the habit of asking for help in the moment has clearly taken hold.

In January, live classroom quizzes reached 6,832 students across 23 institutes, with 313 quizzes that month. This year we kept improving the experience for both sides of the classroom. Students get performance segments, rank ranges and a 75-second timer per question, and instructors get section selection, a live “% joined” view, a list of students who have not joined, and copy-paste lockdown. By September, the live quiz reached 17,257 students across more than 40 institutes, with 692 quizzes in the month. That is about two and a half times the students and more than double the quizzes.

The learning portal had grown screen by screen, and parts of it were still built on older components, which made some units feel out of step with the rest. This year I audited the learning screens and listed every component, explored the new WaveKit design system, and helped define which components belong to the central system and which belong to a single pod. We also designed new components such as vertical tabs and a horizontal stepper. I built the events page in Cursor using WaveKit, and I am now working with the developers on the learning portal revamp using Cursor and WaveKit. There is no number for this yet, because it is still in progress. What I can say is that designing and building in the same flow is the direction we are betting on.