As someone who is also sometimes underestimated, I sympathize with your situation.
Good on you though for being self-aware and perceptive enough to realize it is they who have the incorrect view and having the patience to let that view correct itself over time. Turning it to your advantage and taking joy in watching that correction take place is doubly good.
Most people in your position just assume, “people don’t think highly of me on first impression, and they must be right”
The re-focus on SaaS as the only class of investment that can reliably generate returns and avoid zeros or capital loss is a very real trend.
Would emphasize your point around services - SaaS investors are generally allergic to this stuff and prefer services to make up as little of revenue as possible. It's typical low margin and not seen to be very "strategic" (though this could be debated).
SaaS is also much easier to analyze and diligence than the typical non-SaaS software company or consumer internet business. I won't say it's dead simple, but it's very much not rocket science. In combination with excess capital, this leads to prices getting bid up as such ease of diligence leads many investors to throw in a term sheet. It's just so easy to get comfortable with this stuff.
One caveat to all this that ties to your last point around the market / economy is volatility. You can see in the data that companies that generate a higher amount of the their growth from SaaS-like retention/upsell see higher valuation volatility when the market turns for any reason. [1] The "best" SaaS companies in the eyes of later-stage investors are typically those with high net revenue retention - but these are also the ones that get whacked the most in corrections.
As far as a downturn taking down growth equity - time will tell.
Exactly, SAAS is much easier to analyze, that's why everyone moved to SAAS investing because it doesn't have many "flips of coin" in order to be successful.
> As far as a downturn taking down growth equity - time will tell.
As with everything. My prediction is, as the concentration around SAAS increases, more funds will be created (especially if no other instruments can produce such a high predictable growth) leading to more concentration.
If I didn't have a business to run, I would be already trying to raise fund for Series A and B to focus on non-SAAS business. You get quite heavy discount on them as there is little competition and and you again get the 1/3 business model of early stage VC, which can produce outsized returns. over just 3x.
Risk / return trade-off still holds - great insight!
Performance through-the-cycle is a big question. One can point to Salesforce (founded 1999, IPO 2004), which has been around for 20+ years... However, big sample bias here (ditto for my article, with sample n = 1). Salesforce, a big-category-defining company - may not be representative of moderately-sized businesses.
Whoisnnamdi - per your post, revenue retention looks like a key metric driving valuations!
My point wasn't that all SAAS business are going to be in trouble, rather than the growth equity is going to be impacted because they are funding only SAAS businesses. There is no hedging and as such, once business stop paying for certain SAAS services, they will in turn stop paying for others and so on and so on.
Just as an FYI - the statsmodels python package just released numerous new time series tools in version 0.11 rc1 [1] and also has functions for quantile regression [2]
Another point - a version of "do things that don't scale", sign up your first 100 customers door-to-door
I'm guessing this advice shouldn't be taken completely literally, but rather, the point is that one shouldn't fear signing up your first 100 customers in an unscalable, high-touch way. In fact, he seems to argue that by doing it this way you are more likely to get real, raw, and candid feedback from those users
So interesting to see something like this right after vscode added jupyter notebook support this past month, which I was excited to see given how poor the editing experience is in standard notebooks, especially around intelligent autocomplete.
There are lots of cool developments in IDE notebook support - IntelliJ just dropped a plug-in for it as well. To be honest I'd be thrilled if an IDE solution could fill all the gaps that Polynote's targeting (our work would be done!) so I'm looking forward to seeing what develops.
I might be confused, but isn't it the opposite here? Borrowing long (long-term leases from suppliers) and lending short (month to month leases to customers)?
You yourself said "long term promises to suppliers keep going", which sounds like borrowing to me?
I could just be dense here, feel free to correct me
Maturity transformation, to give it its proper name, isn’t inherently a bad business; it’s how your bank transforms your weekly or monthly pay packet into a 25-year mortgage. But - and here is the crucial point - it is not tech.
It is a bad business generally, because it requires an unfailing entity to exist as a stopgap ("lender of last resort") who will take the hit to prop you up as the short term payment becomes due, for the inevitable times when the payments don't line up. WeWork doesn't (yet) have that.
True, but that is a general property of fractional reserve banking, not specifically of maturity transformation.
There is a lot of ideology around homeownership both for and against, I am on the for side, so when asked about the social value of banking, that’s my go-to answer.
Stripe is absolutely a tech company. They're a B2B SAAS company. They provide software that connects e-commerce platforms with actual payment processors.
Uber isn't at taxi company, since it doesn't own or operate taxis. Conceptually it's a platform for connecting independent taxi operators with customers, like Expedia or Craigslist or eBay.
Airbnb isn't a hotel company, since it doesn't own or operate hotels. Conceptually it's a platform for connecting independent hotel operators with customers, like Expedia or Craigslist or eBay.
SpaceX isn't a tech company any more than NASA is. They require science and technology but they sell rides on space ships that they own and operate.
>Uber isn't at taxi company, since it doesn't own or operate taxis. Conceptually it's a platform for connecting independent taxi operators with customers
Uber is a moonshot bet on re-imagining transportation when autonomous vehicles become a reality. If that reality never materializes, Uber either raises prices and sees its growth rate and valuation crash back down to earth or it goes out of business.
It’s also a moonshot bet that when autonomous vehicles become a reality, Uber will wind up on top. I don’t know about that at all.
IBM owned everything in sight in computing, amd when PCs cam along, it improbably wound up owning them for a short while too...
Then MS took over. They owned everything in sight, but stumbled when the Internet came around, and then again in mobile. They’ve roared back, but they don’t own everything in sight.
Uber’s valuation doesn’t leave room for stumbles and settling for roaring profitability. It’s a bet that when the inflection point comes, they will own everything and keep owning everything.
It could happen, but it takes more than just autonomous vehicles becoming a reality for their valuation to become real.
There are bets stacked on top of bets. Let’s say that Tesla does deliver a viable self-driver, something presently only seen in sci-fi, like fusion power has been just around the corner for 50 years. Why would they cede the relationship to Uber when they already own the relationship with the end users? They already bypassed conventional dealer networks. Other car makers in this space do own their own dealers. Companies like Hertz own customer relationships and have experience managing fleets. Where does Uber fit into this world? They don’t.
At least nobody said Uber is a "data" company like they used to. Five years of Uber telling me the driver is "5 minutes away" for 20 minutes straight when I request a pickup at LAX tells me they don't even look at their data!
I have no evidence, but I believe the five minute thing is deliberate - to prevent users from cancelling taxis because it will take too long to get there.
When you get to the point of saying space rockets are not technology, it's time to acknowledge that somewhere in the construction of your ontology, something has gone wrong.
In this context I take "tech" to mean software, as in arbitrarily scalable with an exponential growth curve. (Not that that can last forever in software, either, but it is growth potential that drives valuations.)
Here's an even better proposal - lets try immeasurably hard to stop re-inventing words and maybe use words that people would generally understand without an accompanying definition. Where, for instance, software actually means software and tech actually means technology. :p
That's very nice, but "Tech" sometimes means an industry now, not the technology itself. Same for software. There is software the code, and Software the industry. Nobody invented a new word, but we have to understand what we're talking about in context.
I think this is a reasonable comment to make (perhaps unworthy of downvotes, which it seems to be accumulating), but I will note that the article defines "successful modern tech companies" as generally having five principal features:
- Low variable costs
- Low capital investments
- A lot of customer data and customer intimacy
- Network effects
- Ecosystems that boost expansion with little cost
Based on those features, of the companies that you listed here, Uber and Airbnb are definitely tech companies. So is Stripe, I think most of us would agree, although I'm not sure whether there are many network effects with their platform. SpaceX is not.
As such, the only way your point really stands is if one can make a valid argument that WeWork is doing something material and tech-focused that makes it different from a standard commercial real estate company. If it were leasing space and successfully executing on a model with low variable costs, low capital investments, lots of data, network effects and expansion-boosting ecosystems, that would be one thing. That doesn't appear to be the case.
> They all need technolgy and came about because of the powerful handheld mobile device.
I'm not sure why Uber is a tech company because you can order a taxi using your phone, but WeWork is not ... because you can reserve a room or space using your phone?
Not all members pay for an office. Some of us are nomads and stay at whichever WeWork is convenient. We make reservations for space using our phone.
You've been able to book a taxi on your phone for decades - with your voice. Just like if WeWork had existed 60 years ago, you would've been able to book a space in your newest nomadic locale with a phone call.
The value prop of Uber was that you could be instantly and automatically matched with a nearby taxi (minimizing cost and time) and see those costs / timings in advance. That required[1] the internet and software, i.e. "tech".
[1] Ok, so hypothetically, you could've implemented Uber 60 years ago if the taxi company called every single one of their cabbies every time they got a phone booking, figured out which one was closest, sent them to your location, and then called you back to tell you "your cab will be there in 5 minutes". But tech made it VASTLY more efficient. Tech makes WeWork only marginally more efficient. Hence, not really a tech company.
In fact, I think "efficiency gains from developing tech" is probably the best way to determine if something is a tech company or not. SpaceX is a tech company because they're using advances in technology over the past 60 years, combined with their own advances in technology, in order to deliver a product (re-usable rockets) that is far more efficient than what we had before (and brings down cost appropriately).
Uber is a tech company because their primary business is providing an app. If they owned or leased a million cars to provide their service then they would not be a tech company.
If WeWork was AirBnB for work spaces then I'd call them a tech company. But they're not.
So you're saying Uber is a tech company for being what it is. But it wouldn't be a tech company if it is exactly what it is today but also owns the cars the driver's use?
So by that logic if Uber ever does migrate completely to a self-driving taxi system, it is no longer a tech company.
> So by that logic if Uber ever does migrate completely to a self-driving taxi system, it is no longer a tech company.
That seems fair to say. If Uber transitions to self-driving taxis the whole business model will be radically different. They'll have to employ hordes of mechanics, and lease huge garages to store vehicles and run repair bays. They'll have massive capital expenses, a complex supply chain to manage, and a large distributed and probably unionized workforce to negotiate with. Their geographic distribution, legal exposure, exposure to economic shifts like tariffs, reliance on capital markets, number of employees, etc... will all be radically different. It seems fairly obvious to say that pre-self-driving and hypothetical post-self-driving Uber are totally different kinds of companies.
Yeah, pretty much. They would be a fleet management company at that point. They might be more high-tech than most fleet management businesses, but their core business would be fleet management, not tech.
You could get a taxi with you phone. You could fax in a lunch order too. But that's not how the market is doing business.
Early on, one of the UX joys of Uber was being able to see where your driver was and how soon before they would arrive. SMS wouldn't be quite the same, would it?
If you're calling SpaceX a defense contractor, then nothing is tech. So what's your ACTUAL definition of "tech" like Apple? Or are you going to say - No, Apple is a fashion company?
Even if AirBnB isn't "tech company" it definitely is NOT a hotel.
I wouldn’t call Uber especially high tech but it’s ok to consider it “tech”. While WeWork is literally just a landlord with no technical value ad. The closest it is to “tech” is its customers.
Did you all consider using Double Selection [1] or Double Machine Learning [2]?
The reason I ask is that your approach is very reminiscent of a Lasso style regression where you first run lasso for feature selection then re-run a normal OLS with only those controls included (Post-Lasso). This is somewhat problematic because Lasso has a tendency to drop too many controls if they are too correlated with one another, introducing omitted variable bias. Compounding the issue, some of those variables may be correlated with the treatment variable, which increases the chance they will be dropped.
The solution proposed is to run two separates Lasso regressions, one with the original dependent variable and another with the treatment variable as the dependent variable, recovering two sets of potential controls, and then using the union of those sets as the final set of controls. This is explained in simple language at [3].
Now, you all are using PCA, not Lasso, so I don't know if these concerns apply or not. My sense is that you still may be omitting variables if the right variables are not included at the start, which is not a problem that any particular methodology can completely avoid. Would love to hear your thoughts.
Also, you don't show any examples or performance testing of your method. An example would be demonstrating in a situation where you "know" (via A/B test perhaps) what the "true" causal effect is that your method is able to recover a similar point estimate. As presented, how do we / you know that this is generating reasonable results?
Thanks! Yes, the concerns you mentioned would also apply to PCA. What we've actually done to help alleviate this is a union of components from y-aware[1] and normal PCA to capture variables that are correlated to both the dependent variables and (hopefully) most of the treatment variables. This seems similar to the double selection approach you mention - the difference being that since we are trying to run this at scale for 1000s of treatment variables, running a feature selection with each of the 1000 treatment variables as the dependent variable isn't super feasible, so the normal PCA acts as proxy for this part of the double selection.
Regardless, we're never going to completely remove omitted variable bias, as we're never going to capture 100% of relevant variables. One way we monitor our model's bias is by looking at the error distribution between users in the treatment vs control. If these aren't similar, there's too much bias in our estimate of the treatment effect, so we wouldn't want to serve an estimate of the treatment effect for this variable to our customers.
The current product is in beta and we're working with some of our current customers to try to re-create our results with A/B tests. I'm hoping that by our GA release in the fall we'll have some case studies with specific examples!
Good on you though for being self-aware and perceptive enough to realize it is they who have the incorrect view and having the patience to let that view correct itself over time. Turning it to your advantage and taking joy in watching that correction take place is doubly good.
Most people in your position just assume, “people don’t think highly of me on first impression, and they must be right”