Chegg: The Case of its Revival.
Deep Dive
Last week I was reading Brad Jacob’s book called How to Make a Few Million Dollars. In one of the chapters about AI, he cites Chegg as an example of how fast a business can be disrupted by AI models.
Since I run an education business and am actively developing my Duolingo thesis, Jacobs’s case study piqued my interest. Turns out it’s not so black and white.
Edited by Brian Birnbaum.
Chegg is priced for bankruptcy, but it may have a bright future ahead. Chegg also teaches tech investors a terrifying and valuable lesson: disruption comes when you least expect it.
ChatGPT has disrupted Chegg’s business by reducing the willingness of customers to pay and becoming top of funnel. Customers just aren’t so keen to pay when they have a free study copilot. Before ChatGPT’s arrival, Chegg spent over a decade building an SEO (search engine optimization) traffic moat by indexing expert content. Now, that moat is being eroded as students ask LLMs (large language models) questions instead of using traditional search. As of Q3 2024, Google is answering student questions in-situ, featuring Chegg’s links below said interface.
As a result, Global nonsubscriber traffic to Chegg was down 37% YoY in October.
However, a glance at the financials reveals that the stock market may be overreacting to this development. As you can see in the graph below, cash from operations (orange line) has remained relatively stable as compared to the stock price (blue line), which has plummeted. While ChatGPT does present a headwind to the business, it’s likely not as much as the market currently believes. Hence, Chegg may be a highly asymmetric opportunity at present.
At just 0.38 times sales, Chegg is priced for bankruptcy. However, some quarters before the departure of the previous CEO, Chegg started working on building out a Duolingo-like infrastructure that leverages their proprietary data to deliver an AI-enhanced student experience that’s more accurate and reliable than generic LLMs. Chegg has decades of data on what works for students and what doesn’t, which positions them to potentially outperform.
In Q2 2024 CEO Nathan Schultz said that, per their research, receiving incorrect information is a top concern for 47% of Generative AI users and that 75% are willing to pay for a service that’s designed to produce better outcomes.
Although subscribers are down meaningfully YoY, per the graph below, the number of questions asked is up 79% YoY as of Q3 2024. In H1 2024, the number of questions asked was up a notable 109%, coming in at a total 16.2M questions. Further, as of Q2 2024, 70% of Chegg’s students were participating in conversational instruction.
These two metrics indicate that Chegg’s new approach is resonating with consumers. The feedback gives Chegg license to rapidly build additional features that improve student outcomes per dollar spent per vertical. These features are yet to be built, but I believe the company is now in good hands.
Former CEO Dan Rosensweig served from Q1 2010 to Q1 2024, taking the company public in November 2013. In Q1 2024 current CEO Nathan Schultz took over, having spent 16 years working at Chegg. He is therefore familiar with the education market and Chegg’s mindshare. He currently holds around 0.5% of total shares outstanding, which is perhaps not a sufficient stake. I would much rather prefer Schultz to be a founder-operator, which I have found is an essential component of a multibagger thesis.
Though we have little to go on, Schultz’s vision is one of rapid iteration. The students of my Tech Stock Goldmine course know this is a leading indicator of future financial outperformance. We’ll need to see evidence of execution to validate his leadership. Still, philosophically, he’s on the right path. Per the Q2 2024 earnings call, he is making strides to infuse Chegg with a focus on iterative product development as an organization.
In the Q2 call, Schultz shared some insight on how this approach is already yielding features which set Chegg on a path to be competitive in the AI race:
On the global product side, we are well underway in implementing our iterative approach to product development.
This fall, we will be testing a variety of innovations.
As an example, we have developed a feature internally referred to as Starting Point, which is meant to address the common issue of students simply not knowing where to start, whether they are studying for a mid-term or writing an important paper.
This introduces a whole new way for students to leverage Chegg on their learning journey.
In addition to Starting Point, we have developed two new applications, one that keeps students on track and another that organizes students' notes and turns them into study tools.
As we get more products into students' hands through iterative development, you are beginning to see the evolution of Chegg from a Q&A platform to one that delivers 360 degrees of support.
Chegg is currently pointing its AI infrastructure to what management terms as the two ‘most relevant use cases’, which are practice and solution comparison.
By gathering data on these two use cases on what works and what doesn’t, Chegg has a clear path ahead to train increasingly better AI models. This is the same path that Duolingo is on, except that they are far ahead with 8.6M paid subscribers and 37.2M DAUs (daily active users).
As of Q2 2024, Chegg has added three key components to its AI architecture:
A mechanism that seeks to understand questions in depth, so as to personalize the learning pathway for the student in question.
A multi-source approach, leveraging generic and proprietary LLMs as well as subject matter experts.
An ‘air traffic controller’ that selects the best approach to assist each student, screening for quality of content displayed to students simultaneously.
As of Q3 2024 there is some evidence that these features are driving value. In Q3 2024 Chegg Study global retention rate came in at 81.8%, up from 78.5% in Q2 2024, meaning subscribers are more engaged and thus more likely to stay. I suspect that the market, rather than (correctly) focusing on engagement, will instead be looking at overall subscriber numbers. However, higher engagement means that the AI architecture is working and that Chegg is likely on a path to being competitive again.
I also wouldn’t be surprised to see the market fixating on the declining ARPU (average revenue per user). As you can see in the graph below, ARPU has been declining steadily with the onset of subsequent LLM generations. However, AI is inherently deflationary to the education industry and this means ARPU is a less useful metric in the context of the turnaround. What matters is free cash flow per subscriber, which is a function of the engagement per subscriber and the leverage the AI infrastructure yields.
In the Q2 call, Chegg CFO David Longo shared some thoughts on how Chegg is thinking about this.
And thankfully, with our automated solutions rollout and our further integration of AI, we can serve those customers still profitably bringing in a lower ARPU.
In terms of profitability, a superficial look would reveal that Chegg has taken a considerable hit, with net income plummeting in the last two quarters as you can see in the graph below. However, this is largely due to non-cash expenses. In Q3 2024 Chegg registered an impairment charge of $196M, as a result of the continued ‘industry pressure’ and the declines in the market capitalization. In the same quarter, Chegg also reached a settlement agreement to resolve Leventhal class action securities lawsuit, recording an estimated $55 million contingent liability for the loss.
In Q2 2024, Chegg completed an impairment test on their goodwill too. The test resulted in $481.5 million of non-cash impairment. This impacted their Q2 income tax provision, as at that point they’d recorded three years of cumulative pre-tax losses in the U.S. This triggered a further $141.6 million non-cash valuation allowance recorded on all U.S. federal and state deferred tax assets.
This is why, when dealing with growth companies, watching the cash flow statement is paramount. Net income can decline drastically with the introduction of non-cash expenses, while cash from operations tends to be a more accurate reflection of a company’s earning power. And indeed, Chegg’s cash flow statement that the company is not doing as badly as the income statement suggests.
Per the aforementioned increased retention rates from Q2 to Q3 2024, it seems that Chegg is making good progress in terms of its value proposal. The bad news is that, as previously mentioned, its legacy customer acquisition channel is in dire straits. As LLMs continue getting better, SEO will likely continue losing market share. The 37% YoY decline in non-subscriber traffic in October is likely only the beginning. Per the Q2 and Q3 2024 earnings reports, Chegg seems to be making progress in developing new acquisition channels. They now have over 53,000 followers on TikTok and in Q3 2024 they launched a Discord app and a Chrome extension, in order to meet their customers inside their study workflow.
Over the coming quarters, I will continue to study the company closely. Specifically, I’ll be looking at the engagement levels to gauge the effectiveness of AI infrastructure. And I’ll also be looking for signs that Chegg is able to reduce its reliance on SEO traffic, ideally diversifying into a set of acquisition channels. However, I walk away with a good impression of the management team. Chegg has non-zero odds of beating generic LLMs at delivering better student outcomes per dollar spent over time and the market is not discounting this possibility.
Chegg shows tech investors why investing in tech can be so dangerous. And why it’s so important to invest in companies that are actively looking for ways to disrupt themselves.
In Q1 2021 Chegg was valued at $12B (46 times more than today) and back then the business was highly successful. The Q1 2021 earnings call was a victory lap, with then-CEO Dan Rosensweig comparing Chegg to Google:
And I've never really seen [sic] with the exception of maybe Google search a [sic] business like this one.
Meaning its growth rates, its gross margins, its EBITDA margins, its ratio of EBITDA margins to free cash flow.
Reading through this earnings call’s transcript is terrifying. Before the release of ChatGPT 3.5 in November 2022, there was no way to foresee the disruption that lay ahead unless you had a visionary understanding of LLMs. In Q1 2021 Chegg management was laser-focused on putting more content out there to drive growth, with great financial success. Meanwhile LLMs were a year and a half away from presenting Chegg with an Innovator’s Dilemma, whereby what previously seemed like a toy quickly came to disrupt the business in meaningful ways.
In the earnings calls after Q1 2021, CEO Dan Rosensweig began to gradually talk about AI and LLMs. In the Q2 2022 call Rosensweig started to address AI as a friend and not a foe, stating that he believed it would improve Chegg’s ability to “provide support with grammar, paraphrasing and set structure.” What’s even more terrifying is that his view made perfect sense, in that he believed AI was going to decrease costs and enhance value delivered to students. He was totally right about this.
He just didn’t foresee that the combination of LLMs reducing the willingness of customers to pay and disruption at the top of the funnel would lead to quite a bit of pain for Chegg (and I say this with all due respect, because being CEO isn’t easy). Here’s Rosensweig’s view about the matter in Q4 2022:
But our expectation, and I think you're going to see it from a lot of companies -- remember, Chegg, they've already said that they do not plan to keep it free. They can't run it if it's free. So it's going to be an API-based business where we will be participating and using it to enhance our product.
The argument was only valid for a few months, because there soon came a non-linear increment in LLM performance and versatility. In Q1 2023 ChatGPT-4 launched and by March 2023, Chegg was already feeling the pain. Chegg saw a ‘spike in student interest’ for ChatGPT, with customers starting to be a bit reluctant to pay. Here’s what Rosensweig said during the Q1 2023 call about this:
It's just on the margin that based on our research that people normally who would have pay for us around mid-terms are closer to finals that were reluctant to pay or be longer term subscribers are now have a new free site to go try.
By the end of Q1 2023, ChatGPT and other LLMs had started to disrupt Chegg’s customer acquisition funnels and this trend is still unfolding today. Students are increasingly turning to these models for answers, with the models’ answering in-situ without taking users anywhere else on the internet. This is a spectacular and highly unforeseeable form of the Innovator’s Dilemma–LLMs seem like a silly chatterbox at the start, but, over time, have considerably disrupted Chegg’s core value proposition and the way they acquire new customers.
That’s disruption at two different levels of Chegg’s value chain.
This is a reminder of the dangers of investing in the tech space. For Chegg, disruption came at a moment of maximum complacency and delight when the stock was trading at over 18 times sales. Although dealing with this disruption was very hard in real time, a company with a culture of constant-self-disruption would have higher odds of thriving. Per management remarks in the quarterly earnings calls at the time, the company was uniquely focused on enhancing financial results by executing its old playbook more intensely.
I found this statement from Rosensweig in the Q1 2021 Q&A section fairly insightful:
We are not trying to manage to any particular number except growth. So when we see an opportunity to invest, we do it. We're not holding back anything.
Meta is a great example of a company trying to disrupt itself by investing in the Metaverse. Should virtual reality become the next frontier for social media, not being number one in the space would likely mean obsolescence for Meta. Rather than risking that possibility, Zuckerberg decided to invest heavily in virtual reality technology–as I explain in my November 2022 Meta deep dive. The market put tremendous pressure on Zuckerberg to stop doing so, including a 70% decline of Meta stock in 2022.
Still, Zuckerberg has continued to invest.
Going forward, Schultz’s management will be the defining factor for Chegg’s turnaround. However, what enables them to orchestrate a comeback in the first place is their proprietary data. But they will also have to train AI models that others can’t, which I believe will be the primary moat in the AI-driven economy–and which I teach students in my Tech Stock Goldmine course.
Tracking Chegg over time will be fascinating for two reasons: it may yield a highly lucrative investment opportunity; and it will reveal to what extent a proprietary dataset is a moat or not, when competing with generic AI models.
Stay tuned and until next time!
⚡ If you enjoyed the post, please feel free to share with friends, drop a like and leave me a comment.
You can also reach me at:
Twitter: @alc2022
LinkedIn: antoniolinaresc
Insightful and Thoughtful. The question is: Is an answer a commodity or not, to me? I believe an answer is a pure commodity particularly with students. As long as, ChatGPT and Gemeni are Free and Google Search is pushing for disintermediation, Chegg is a loser. Coursera and Udemy are in a much better competitive advantage. Due to not being a pure answer commodity.
Great read! But wouldn’t say I’m 100% convinced!