It’s been nearly 20 years, but Palantir, a tech company founded in 2003 that helps commercial enterprises and the government manage, analyze, and act on their data, is finally going public. Palantir filed its required S-1 financial papers with the SEC earlier this week, and as I was looking through it, one question stood out to me in particular: Is Palantir a Software-as-a-Service (SaaS) company? My answer is no, and I think Palantir is currently over-valued.
Let’s begin with a few properties of SaaS companies:
SaaS companies incur high up-front cost to develop cloud software that they then sell to other users, generating recurring revenue, usually on a monthly or yearly subscription basis. Because software, once it’s developed, has pretty much 0 marginal cost of distribution, the marginal cost essentially devolves into the customer acquisition cost (CAC), which, as you might guess, is one of the most important metrics for SaaS companies. Andrew Chen has a great post on the complexities of calculating CAC.
While SaaS companies face high up-front costs, the recurring nature of revenues eventually (hopefully) kicks in and more than compensates for those initial losses, making SaaS companies the VC poster child of investments. Thus, another important SaaS metric is the customer lifetime value (LTV), the amount of revenue a customer will generate across its entire lifetime. The success of a SaaS companies depends on keeping its LTV higher than its CAC. If your LTV is higher than your CAC, then from a purely profit perspective alone, you’re incentivized to sell your software to as many people as possible, until your CAC rises to meet your LTV. VCs typically want an LTV:CAC ratio of 3:1.
One of the ways you can value a SaaS company is through a revenue multiple. A quick rule of thumb I use when guessing the market cap of a SaaS company: Multiply its yearly revenues by 10. In fact, the largest and most successful SaaS companies have been generating even larger revenue multiples. For comparison, Sprout is at 22x, Cloudflare is at 18x, Datadog is 36x, Crowdstrike is 27x, Slack is 21x, and Zoom is at 37x.
On the surface, Palantir seems like it’s a SaaS company: It offers software products that customers can access on the cloud. Palantir also seems like it’s being valued like a SaaS company. In 2019, Palantir generated revenue of $724m and now reportedly has a valuation of about $20bn. That’s a yearly revenue multiplier of ~27x, worthy of mention in the major leagues of publicly-traded SaaS companies.
But I can’t find any other evidence in Palantir’s S-1 that points to it being a true SaaS company. While Palantir does offer its products on the cloud, many customers also deploy Palantir in their data centers,laptops, specialized hardware, on-premises, and more. This means that the cost of distributing the software isn’t 0 in all cases. Furthermore, while Palantir does generate a portion of its revenues from subscriptions, it also sells “license[s] and maintenance and service agreements.” While subscription revenue is probably recurring, licenses and service agreements probably aren’t.
Even more interesting is Palantir’s unprofitability. It lost about $580 million in 2019, and about $165 million in the first half of 2020. Palantir is even repeatedly warning investors that it “may never achieve or maintain profitability.” Most healthy SaaS business models reach profitability at some point (remember that VCs like seeing LTV:CAC of 3:1), but not Palantir.
One of the reasons Palantir may not ever achieve profitability is that its cost of sales and marketing (and therefore its CAC) is astronomical. From its S-1:
[O]ur sales model often requires us to spend months and invest significant resources working with customers on pilot deployments at no or low cost to them, which may not result in any future revenue.
“Pilot deployments” kinda’ suggests that Palantir is developing custom software in the customer acquisition process. This is antithetical to the typical SaaS business. As I mentioned, SaaS businesses typically incur 0 marginal cost of distributing the software, since customers are all using the same version on the cloud. Here, Palantir incurs nontrivial cost in tailoring its software for every marginal customer.
We can in fact try to estimate Palantir’s LTV and CAC (they’re not explicitly in the S-1). Let’s stat with LTV. According to its S-1, Palantir brought $5.6m average revenue per customer in 2019. Now we ask how long each customer stays with Palantir. Palantir’s top 20 customers have remained with Palantir for an average of 6.6 years. We can probably assume that the average customer has remained for less time than 6.6 years, but since we don’t know the exact number, we can stick with 6.6 as an upper bound. Finally, Palantir says that it expects its customers to remain on average for another 3.5 years. The LTV is therefore approximately $5.6m per customer per year * (6.6 + 3.5) years = ~$57m per customer.
The CAC is a bit harder to estimate, and I admit I could be way off here. Palantir discloses that it brought in $5.6m of revenue from new customers in H1 2020. If we continue with the assumption that each customer brings in $5.6m in revenue per customer per year, then Palantir brought on 2 new customers during H1 2020. During this same period, Palantir spent $201m on sales and marketing. The CAC is therefore $201m / 2 = ~$100m. The LTV:CAC ratio is therefore < 1!
[Of course, disclaimer: This LTV:CAC ratio may be wildly off. I’m pulling numbers based entirely on what Palantir has given in its S-1. If I had more data, I’d try to do a cohort analysis since Palantir segments its customers into three separate cohorts (“Acquire,” “Expand,” and “Scale”). Palantir undoubtedly has different LTVs and CACs for customers in each cohort. To illustrate, Palantir’s top three customers together represented 28% of the company’s revenue for 2019. Its top twenty customers represented 67% of total revenues, with each one of those customers averaging $24.8 million in revenue.]
In conclusion, I’d probably classify Palantir more as a hybrid services business that also sells cloud software services. The are very comparable to an IBM or a Cloudera: they have a proprietary analytics platform that they then customize specifically for the workload. They then charge implementation fees, on-going maintenance, and a subscription fee for the software, which is, again, customized to the user. For reference, Cloudera’s revenue multiple is ~5x, and IBM’s is ~1.5x.
Based on this information alone, I’m skeptical that Palantir should be valued as a SaaS company at $20bn. That said, I want to be fair to Palantir. Just because Palantir isn't a SaaS company doesn’t necessarily mean that its current valuation is wrong. Perhaps private and public investors place higher emphasis other salient features (e.g., customer stickiness? average contract value? government contracts? competitive landscape?). If you’re an investor who disagrees with me, I’d love to chat!
I will be releasing Part 2 of my Palantir S-1 analysis next week. That analysis will be less finance and more strategy / vision.
Kamala Harris’ track record on tech policy.
Recapping Elon Musks’ Neuralink presentation. The eventual goal: Putting chips in our brains; solving all sorts of neurological conditions; uniting us with AI. The current status: Implanting chips in pigs’ brains. “Cyporks,” if you will.
Venture capitalist Bill Gurley on SPACs.
The philosopher who is trying to explain COVID-19. I disagree with nearly everything he says, but I appreciate the thought-provoking take nonetheless.
Uber, Lyft, and AB5. California is a mess. Bad, unclear regulation is not conducive to innovation. I’m interested in learning more about startups that have consciously chosen to avoid California for this reason.
