Why We’re Not Buying Appen Stock On the Dip

The difference between confidence and arrogance is whether or not you can walk the walk. Confidence is a highly desirable quality for managers because it’s married to another prized quality – decisiveness. And as a leader, you won’t be very persuasive unless you make decisions with conviction. In the world of investing, we say people with conviction have “strong hands.”

If you’re a confident investor, you’ll put your money where your mouth is. You’ll be less worried about being seen as some market oracle and more concerned about learning how to become a better investor. Investing with conviction means you trust your own judgment. Experienced investors know how important this is for your mental health. You will not sleep well at night if you cannot make a decision and stick with it, even when things appear to be going against you.

Today, we have a decision to make. We’re holding a miniscule amount of Appen (APX.AX) and our position is down -30%. We couldn’t be happier, as we had only just started buying it. (It’s the smallest position we have in our 27-stock portfolio). Before we take advantage of this great opportunity to purchase a growth asset at a 30% discount, we need to decide if this is a stock we want to be holding. The growth is there for sure:

Credit: Appen 2019 Annual Report

While the business looks extremely healthy, we’ve recently come across some concerns that make this asset much less appealing.

Investing in Appen Stock

The first time we wrote about Appen was in a piece titled Crowdsourced Big Data is Big Business for Appen. We all know about crowdsourcing, getting people to do menial tasks for peanuts. (In some countries it’s meaningful money, so let’s not immediately condemn it.) In Appen’s case, their team of over 1 million people are taking big data and then using it to produce metadata – data about data. It’s all about establishing some context around big data which is being used to train artificial intelligence (AI) algorithms. Appen operates in the following two segments (percentage of 2019 total revenues in italics):

  • Speech and Images (13%)
  • Relevance (87%)
Credit: Appen 2019 Annual Report

Let’s start with Speech and Images.

Speech and Images

Does anyone else find it odd that we still can’t figure out what 20% of the world’s population is talking about? We can create synthetic life, yet we’re still unable to translate a simple Chinese newspaper article such that it contains zero Engrish. The inability of humans to master language translation using traditional programming techniques is now being solved with artificial intelligence.

What Appen now calls “Speech and Images” used to be called “Language Resources,” and the name change may be associated with their March 2019 purchase of Figure Eight.

Credit: Appen 2019 Annual Report

In March of last year, Appen paid $175 million (plus potentially $125 million more) for Figure Eight, a machine learning software platform which uses highly automated annotation tools to transform unstructured text, image, audio, and video data into customized high-quality AI training data. It’s something we wrote about in an article titled Lots of Jobs You Can Work From Anywhere. By purchasing Figure Eight, Appen created more diverse revenue streams and acquired around 200 new clients. That’s a good thing, because language translation will eventually be mastered once and for all.

The Importance of Language Translation

Last year, we spent a month in Jakarta visiting with tech companies and spending some quality time recharging in the world-famous Central Jakarta health spas. During that trip, we found that the most well-funded startups in the country were chatbot solutions focused on Bahasa Indonesia. Many of the big tech players didn’t think it was a priority to teach computers this large lingua franca used to connect people who speak over 700 languages. There’s a reason for that, something the BBC wrote about in a piece titled “Why no-one speaks Indonesia’s language.” The rich complexity of Bahasa Indonesia is a good example of why solving the language-translation problem requires so much nuance that only native speakers can properly articulate it.

Appen has human resources that span 180 languages giving them a powerful advantage in being able to localize data for companies. There’s a big demand for that, which is why Lionbridge went from software localization to data labeling. They knew that offering translation-as-a-service wouldn’t be a thing once AI figures out how to speak any language.

Now that algorithms are learning how to fold proteins, it’s reasonable to think they may soon understand baiju-fueled conversations with Chinese mainlanders. Investing in a company that’s helping other companies to learn other languages only makes sense until the AI becomes clever enough to understand the languages itself. Whether we’re talking language translation or image labeling, what direction will this segment pivot into once the machine learning algorithms no longer need their hands held?


As Aspen’s stock took a nosedive on the news of revised guidance, they decided to let investors know everything was kosher in a 57-minute “trading update.” In that call, the CEO talks about how the biggest impact on their performance has been a slowdown in digital ad spending, the dominant source of revenues for Appen’s largest customers, and a good indicator of how these customers spend money with Appen. (With the number of ads media companies vomit onto your browser these days, it’s hard to believe there’s any real estate left for digital ad growth – aside from maybe starting to place ads within ads.)

In their 2019 Annual Report, Appen tells us their largest customers – the ones who rely on ad spending for a dominant source of revenues – pretty much have them by the cojones.

During the year ended 31 December 2019 approximately 88.2% (2018: 89.1%) of the Group’s external revenue was derived from sales to five major customers.

Though the Appen Eight acquisition may help with this, Appen’s revenues are extremely concentrated, which presents a great deal of third-party risk. They also derive 87.5% of their revenues from the Americans, and we all know that’s a rather dicey place at the moment.

The big takeaway from senior management is that Appen’s slowdown is temporary. Customers are changing their project prioritizations – benching older mature projects in favor of new projects that address emerging opportunities or threats. This means Appen is expecting a strong year in 2021, and investors have nothing to be worried about. For anyone who believes in the long-term potential of the company, this is just a buying opportunity.

A Buying Opportunity

There seems to be a pattern emerging where people just blatantly ignore the big elephant on the Wall Street trading floor right now – “The Rona.” When the trillion-dollar travel industry – and all the things that go with it – gets decimated by a pandemic that touches every corner of the globe, we’re fully expecting companies will put on the brakes and freeze expenditures. When companies formally announce this to investors, everyone acts completely shocked, and stock prices tank. We’ve seen this happen twice now, and both times opened new positions (Alteryx and Splunk). And now it’s happened with Appen.

Appen’s stock is down -22% this month on news that they adjusted their guidance. Long story short, Appen thought they could sneak by The Rona unscathed, and investors believed them. When things didn’t transpire as planned, Appen adjusted their guidance. Since we couldn’t care less about short term fluctuations, it’s almost non-news. We’re probably more concerned to hear that their client base is concentrated in California, a state that a growing number of tech giants are exiting in search of greener pastures. (Maybe enforcing hiring quotas based on gender wasn’t the best idea?) Today, we just need to decide if we’re going to open a meaningful position in Appen at a 30% discount.

To Buy or Not to Buy

Our recent piece on The Big Business of Big Data Labeling as a Service raised questions about how sustainable these data labeling business models have become. How long will training data be a viable business? If the demand for training data wanes, what takes its place to assure continued growth? Even if growth lasts for a while longer, it’s not what you might consider transformational. It’s a commodity offering that becomes less useful as time goes on.

There’s an opportunity cost associated with any investment you’re holding. Sometimes it’s useful to do relative comparisons. If we looked at every AI stock we’re holding right now, Appen is the least appealing because its value proposition diminishes as artificial intelligence becomes more intelligent. You don’t have to be Quasimodo to see where this whole thing is going.

The relevance of every investing decision comes down to how much skin you have in the game. We don’t have much. Our Appen position sits at about 0.65%. To put that number in perspective, if our total portfolio value was $100,000, then our position in Appen would be worth about $650. The reason our position is so small is because that’s how we approach purchasing stocks in volatile times. We buy small amounts at fixed intervals over extended periods of time.

The decision to make here isn’t whether or not to sell a miniscule holding. We need to decide if Appen is a company we want to hold in the first place. Based on what we’ve discussed today, the answer is no. We are risk-averse investors, and we believe that the ability for AI to teach itself to learn in any language will quickly make this 1-million-strong global workforce obsolete. (Our recent piece on artificial general intelligence shows how quickly algorithms are learning to learn.) We don’t like Appen’s revenue concentration, and feel the company should spend more time on conveying their business strategy and less time droning on about D&I initiatives, some of which are subtracting value from shareholders. If we’re wrong and the stock rises +1,000% over the next decade, we’ll live with our decision, and hopefully learn some lessons during the process.


We’re now actively managing 27 tech stocks in the Nanalyze Disruptive Tech Portfolio and we’re making that process more automated by refining our tech investing methodology to make it more rules-based. The first, and perhaps the most important rule, is not to become vested in a stock where the risks seem to outweigh the rewards relative to other potential opportunities in the same space.

Are we selling Appen? What other five AI stocks are we holding? Find out these answers and more in The Nanalyze Disruptive Tech Portfolio Report, now available for Nanalyze Premium annual subscribers.