Why Veritone Stock Is An AI Company We’re Avoiding
Warren Buffet famously said to invest only in what you know, something that especially rings true in today’s world of ICOs and NFTs. No matter how complex a technology is, the purveyor should be able to explain it simply. The inability to clearly articulate a value proposition is a big red flag, and one company that epitomizes that is Veritone (VERI).
About four years ago, we published a piece titled How Much “AI” is there in the Veritone IPO? About five months later, we asked the question, Are Veritone and ITUS Corp. AI Pump and Dumps? based on the fact that Veritone stock soared +357% in a single month. It turned out to be a pretty timely call as shares cratered not long after, falling from $56 a share to $10 a share in a year’s time. By early 2020, the share price has fallen below $2 a share. Today, shares trade at around $20 a share, which means we’re dealing with an extremely volatile stock.
Our original criticism of Veritone was that they talked a lot about artificial intelligence (AI), but their revenues didn’t seem to reflect that. That’s now changed. On the heels of some recent news surrounding their acquisition of PandoLogic, we decided to take another look.
About Veritone Stock
Of all the companies we’ve looked at, not one has proved to be more confusing than Veritone. At the highest level, the company says they’ve built “an operating system for AI” called aiWare which has “over 300 AI engines in six cognitive categories, including text, speech, vision, speech, data, biometrics, and audio.” They’re helping companies make sense of all the unstructured data in the world, something that seems useful enough, but the whole thing goes downhill from there.
What Does Veritone Do?
Attempting to understand what the company does based on the 15 applications they list out is futile. There’s an app that helps police officers in the State of California collect stop data information, and an energy offering that helps buy, sell, and dispatch energy. There’s a smart contact center solution, and a tool that helps redact sensitive information within audio, video, and image-based evidence. At its best, the list is a poor attempt at demonstrating various use cases with no common theme being apparent – except, of course, the AI operating system. Then, we came across this slide which breaks everything down by industry vertical.
That makes a bit more sense, but not when we compare the above to the revenue segmentation provided in the company’s latest 10-K.
We’re now given three new categorizations to figure out which means we’re back to square one.
In 2016, the majority of Veritone’s revenues came from “media placement services performed under advertising contracts with our media clients,” and it remains that way today. During 2020, Veritone placed approximately $257.8 million in media advertising for their advertising clients. Their use of AI to scour video content for placements seems intuitive. Says the company, “during 2020, ten advertising clients accounted for approximately 58% of the total revenues from our advertising services.” Advertising contracts generally can be cancelled by their clients upon 30 to 90 days’ prior written notice, which means these revenue streams can change very quickly. Even worse, that customer concentration risk extends across their other lines of business.
aiWARE SaaS Solutions
Veritone lists out a number of KPIs they track for their SaaS business, but fails to include the most important SaaS metrics one expects to see such as “annual recurring revenues” or “retention rate.” One of their metrics is “total accounts on the platform” which they say is 1,896 accounts. They then go on to say, “during 2020, ten customers accounted for approximately 53% of the total revenues from our aiWARE SaaS solutions.” It makes you wonder just how the remainder of their SaaS revenues are distributed across the other 1,886 accounts. Not including typical SaaS metrics is a big red flag, not to mention when they say, “our aiWARE SaaS solutions revenue from customers in certain markets, particularly in the government, legal and compliance markets, is often project-based and is impacted by the timing of projects.” That’s not how SaaS business models work.
aiWARE Content Licensing and Media Services
Customers include major sports networks, advertising agencies, and film production companies that require high-value content for their broadcasts and projects. During 2020, ten customers accounted for approximately 38% of total revenues from content licensing services. The decline in 2020 was related to the cancelation of sporting events because of The Rona.
There’s always a temptation to ignore what a company says they do and focus on one of the most important metrics that demonstrate traction – revenues. For Veritone, their annual revenue growth and steady quarterly revenue growth don’t look half bad.
The above revenues will be increasing sharply as Veritone recently announced the acquisition of a recruitment technology company.
Veritone Acquires Pando
Back in 2017, we speculated that Veritone would likely start acquiring companies to boost their AI capabilities, and indeed they have. The same year they acquired some “intellectual property and technology for data analytics” from a company called Atigeo, and then in 2018, they acquired three more companies.
- Performance Bridge – A podcast advertising agency that’s become one of the world’s largest full-service performance-based audio advertising agencies. Total funding and acquisition price not disclosed.
- Machine Box – Pre-trained, production-ready machine learning models inside Docker containers. Total funding and acquisition price not disclosed.
- Wazee Digital – Digital asset management content licensing services. Raised over $70 million and was acquired by Veritone for $15 million.
The biggest acquisition so far for Veritone was announced just last month. Founded in 2007, New Yawk’s own PandoLogic raised around $22.7 million in disclosed funding to build “programmatic recruitment advertising” solutions that have been used by over 125,000 different employers. Their fully automated platform serves job listings across all major search recruitment sites including Indeed, craigslist, LinkedIn, Monster, CareerBuilder, ZipRecruiter, and others. The advertising aspect of PandoLogic meshes with the advertising activities that Veritone conducts, though it’s unclear how they plan to segment their revenues following the acquisition which completed using a combination of cash and stock (Total consideration will dilute existing shareholders up to 7.8%.)
The below table shows how PandoLogic will bring in an additional $50 million in SaaS revenues for Veritone.
Perhaps now they’ll begin including the sort of metrics we’d expect to see from a SaaS company. We’d also reiterate our earlier concern about how their existing “SaaS revenues” are often “project-based.”
With a market cap of $682 million, Veritone falls under our $1 billion market cap threshold, so the stock would be off our radar just based on size alone. Unfortunately, the SaaS part of their business – even following the PandoLogic acquisition – will only account for around half of their revenues, there’s a great deal of customer concentration risk, and the business strategy seems scattered and difficult to follow. Veritone isn’t a company we’re going to spend our limited resources trying to monitor going forward unless they can better articulate what exactly it is they do.
Biometrics is one part of the broader cybersecurity domain which we've invested in with the best ETF out there. Become a Nanalyze Premium annual subscriber and get immediate access to our portfolio of ~30 tech stocks and 3 disruptive tech ETFs.