EdTech First to Fail
In July of 2021, I quit my job at a large digital agency and needed a reset. I was burned out and wanted to move to the client side. After taking a small SEO job at a local auto carrier, quickly I got bored and was ready for something new and bigger. One day a recruiter reached out asking if I was interested in the lead SEO role at Numerade, an education technology (edtech) startup. It just raised a funding round resulting in a $100M valuation.
The pandemic was still dominating everyday life and it wasn’t clear if my two children were going to get into the classroom. Universities all across the US were staying remote for the fall and students needed tools to learn. A quick look at the edtech landscape suggested it was the perfect time to get into this booming vertical. Stock and revenues were at all time highs for the market leader Chegg. And not for the best reasons.
Numerade was different from the rest of the competition and that’s what sold me – short form video explanations to millions of STEM questions created by real, (mostly) expert educators. A sort of asynchronous tutor in short form video – the “TikTok of Edtech” we liked to say.
This was their competitive advantage and unlike Chegg, it wasn’t designed to assist cheating. Numerade pitched itself as a “learning platform.” Finally I found myself firmly in an environment where I believed in the mission and could try to make a difference. Plus, they offered a generous equity package – I was going to be a part owner. Incentives aligned. Time to get to work.
EdTech SEO
Before I started, the CEO said we needed to “10X SEO” – that was the assignment. Diving into Search Console and analytics data, I couldn’t believe my eyes. All of their search queries were extreme long-tail, often longer than 30 word queries – sometimes paragraphs. Students copied and pasted text from their homework into Google. The traffic was so diffuse that the top landing pages on any given day was a couple hundred clicks. And spread over hundreds of thousands of pages with 99% getting 20 clicks or less. We were playing at the margin of Google’s index with little competition.
Numerade had digitized thousands of STEM textbooks coupled with millions of human educator videos. Each one explained step-by-step how to solve these complicated homework questions. These textbook questions totaled over 10M URLs.
We also had a user-generated content flywheel where students could “ask” our educators any STEM question they liked and in time, would get answered. This added tens of millions more questions to the content portfolio. We quickly indexed tens and tens of millions of questions and the clicks flooded in.
In August of 2021, we were bringing in around 20,000 clicks a day. By November it was 200,000 clicks a day.
In 2022 as spring semester finals approached, we were regularly over 425,000 clicks per day. Some 200k pages would get at least one click from search according to Google Analytics. Traffic came from students all over the world and about 40% of this traffic came from the US. In 2022 we had over 70M clicks and over 3 billion impressions.
Easy game.
About 75% of all clicks were in this QAPage carousel format.
80% Revenue
Numerade used a traditional subscription revenue model. If a student wanted to watch a video solution, they had to enter a credit card. Unlike Chegg or other monetized edtech platforms, Numerade offered students a 7-day free trial. When I started at Numerade, we had yet to monetize but as the fall 2021 semester approached, our small, scrappy, startup team decided to put up the paywall to see if students would enter their credit cards.
They did – and far more than anyone at the time imagined. At the peak in 2023, Numerade was converting 3k-4k registered users a day. This made the company’s trajectory flirt with unicorn startup status. $100M annual recurring revenue was our goal and by early fall of 2023 semester, growth rates suggested this was possible. This was growth hacking at its most extreme.
AI & LLMs
We had millions of questions without educator video answers. Before AI, we matched unanswered questions using a similarity score to existing videos in hopes it was close enough to help a student understand the concepts. This created friction and understandable frustration for our subscribers.
When ChatGPT released in November 2022, we immediately put it to work. Now we could answer the millions of unanswered questions in text form. This helped plug a big content gap and provided some quality answers to well-formatted (transcribed) questions. But the model was new and not accurate enough in early 2023. In time, the accuracy improved and Numerade was able to re-apply text answers with more accuracy across the entire question base.
Then US STEM students began to use LLMs. Chegg blamed ChatGPT usage for slowing subscriber growth as early as their Q1 2023 earnings call. The Numerade team suspected this was just an excuse for what was becoming a failing business. After all, Numerade had yet to detect an impact from student LLM usage. Numerade had over 25M unique visitors in Q1 2023 according to Google Analytics. This was 40% more than in 2022 YoY.
Easy game.
2023 Helpful Content Update
We were vulnerable and knew it. Vulnerable to what brought our only real success – Google’s algorithm. We gamed the system to our advantage and it was our only source of quality traffic. We had few referrals and brand demand was non-existent. Nearly 80% of our visits were from organic search.
As students returned to class for fall semester, the cracks emerged. We felt immune to Google’s fickle algorithm updates until mid-September – Google found our content “unhelpful.” Clicks dropped by half from the US and Canada.
Our strategy worked brilliantly – until it didn’t.
We called our question & answer page templates PDPs (product description pages) because they were the brand’s products. We did everything we could think of to make the PDPs “helpful” to satisfy the supposed “EEAT” criteria. We modified our website in dozens and dozens of ways. By December and early 2024 we just about recovered.
Then came a series of spam and core updates Google periodically released over the course of 2024. Every update peeled more and more search visibility away from the Numerade domain. Traffic from the US fell from once was 200k visitors a day to 20k by February 2025.
What was once a 10x increase was now a 10x decrease from our most valuable cohort.
AI Overviews
After their annual May I/O 2024 summit, Google hastily released their Search Generative Experience. Numerade’s traffic followed student demand and search traffic drops in May as expected – students wrapped up the semester and went home. We couldn’t detect an impact.
Following the embarrassment of the initial SGE rollout, Google rallied in the fall. Having rebranded to “AI Overviews,” the impact was almost immediate but also undefinable. We couldn’t determine what the AIO impact was among the many simultaneous applied Google algorithm updates.
By 2025 the average query rank was decent and more or less stable, but demand had fallen drastically. Few Numerade link citations were visible in the AIOs. The available tools confirmed what we already could see. Google was solving the bulk of our queries directly in the search results. Impressions dropped by half, click through rates plunged, and our once mighty SEO program crashed.
Product Market Fit Collapse
Read Reforge’s excellent blog post: Numerade fell even faster than Chegg. By Product Market Fit Collapse – abrupt and disruptive technology shifts. And established, informational edtech was one of the first verticals to go. The target demographic for edtech is what Reforge calls “tech forward.” Fareed Mosavat:
“We’re seeing the real disruption today at the tip of the adoption curve (code, design, tech, students). Businesses that cater to less savvy customers are likely less susceptible.”
Google is for older people and they know it. The younger student demographic are not habitual Googlers. The gen Z cohort spends their time discovering content on TikTok, Instagram, and YouTube. LLMs are abundant and free, while Chegg and Numerade are not. Edtech companies tend to have low-frequency use without an ingrained habit.
LLMs are easy for younger, tech savvy audiences to pick up. And they did. No habit to break.
TL;DR
Below is a list of SEO and business reasons the company is struggling to survive. First by the algorithm, then AI Overviews, and finally user behavior shifts. I added a few business things that contribute to the brand’s unfavorable status in the market.
- SEO traffic grew too fast. Skyrocketing traffic growth, with no brand or channel growth, made Numerade.com an easy target. We gamed the system and were called out on it by coming out of nowhere. Not officially with a manual penalty or anything official, just throttling.
- Perceived as a spam website. By ingesting content at such a massive rate without any type of moderation (dups, malformed questions) made Google’s job easy to apply incremental downward pressure after every algorithm update.
- Google AIOs. Solved questions directly in the SERP, few clicks, fewer citations.
- LLMs and technology forward students. Poor, savvy students live on free stuff.
- Monetized too fast. We needed to create a better product first before throwing up the paywall. I understand why revenue was a focus but in my opinion we needed to create a loyal user base, a fan base – a habit for students. Instead, we frustrated users with a substandard product and they let all the forums know about it.
- Brand. “Numerade” was in low demand and never really gained much popularity over time as per impression data. Growth hacking was the sole focus. Branding and creating a superior product wasn’t the company’s top priority.
- No user verification process. Students could sign up with any email address – valid or invalid – and their account would get lost in the system. Customer service would have to hunt down abusers and issue refunds to angry (and abusive) users who forgot which email they used to sign up and could not cancel on their own.
Post Mortem
Unless Numerade secures entry into schools or acquired by a company with vast distribution channels, it will likely fail due to Product Market Fit Collapse.
Due to the subsequent loss of revenue, in February myself and many other fellow Numeraders were laid off. Heartbreaking but nothing personal. The founder’s fiduciary responsibilities are to its investors.
I’m a fan rooting for Numerade and want them to survive and thrive. After three and a half years at Numerade, I am emotionally and financially invested.
Easy game.