A Coda On WeWork’s Odd S-1

Morning Markets: Among the many failings of WeWork’s attempt to go public, the sheer complexity of its S-1 filing is perhaps my favorite. Here’s a small anecdote as illustration.

The companies which have gone public this year and done well, the Cloudflares and Datadogs, if you will, had reasonable S-1 filings. They listed revenue, cost of revenue, and then the usual mix of operating expenses before reporting GAAP results. It was all quite pedestrian.

Subscribe to the Crunchbase Daily

The two filings (here and here) were not perfect. Each contained different classes of stock that will bend internal power towards early shareholders into the future, but, all the same, their finances were pretty easy to get. Investors, liking what they saw, priced Cloudflare above its raised range, and Datadog the same (its final price was also above its raised range).

WeWork went the other direction, seeing its initial pricing hopes dashed, only to see reduced price targets also fail to generate interest. The firm’s value fell until its IPO was retired.

Unlike Cloudflare and Datadog, however, its S-1 was opaque to the point of obfuscation. Some of this we’ve touched on, some of which we haven’t.

Among the odd was the fact that WeWork didn’t report a traditional cost of revenue expense. The company’s use of both “Contribution margin including non-cash GAAP straight-line lease cost” and “Contribution margin excluding non-cash GAAP straight-line lease cost” as financial metrics were strang, its charts without labeled axes felt like jokes, and its insane and complex leadership planning structure didn’t leave much to be desired.

But, and I am sorry to say this, Crunchbase News did not catch all the lunacy. No. There was even more to be found in the various versions of the company’s S-1 filings than we thought. Here’s the Wall Street Journal:

The document leaves unanswered some basic questions about the company’s finances. For example: How many new workstations did We deliver in the first half of this year? The prospectus filed in August said 273,000. Barely a month later, an amended version said 106,000. What was the total gross cost? In August, We said $1.3 billion. In September: $800 million. The reason for the dramatic changes is that the first version was wrong, people familiar with the matter said.

This was a shocking thing to get wrong. So, I did what anyone in my seat would do and went to fact check the Journal. Not that I don’t trust its reporting. But, instead, so that I could learn where I missed the change, and how I can do a better job reporting on such changes in the future.

Oddly enough (you can try yourself here with the original S-1 and a later version here), fact-checking this claim wasn’t perfectly simple. You see, WeWork actually employs the 273,000 number a few times. Five times, in fact. Here they are:

We estimate that with every $1 billion in net capital expenditures we invest in opening new workspaces, we can create workstation capacity of approximately 273,000 workstations, based on our net capex per workstation added for projects completed in the first half of 2019. […]

The table below demonstrates the potential annual revenue and target contribution margin that could be generated over the lifetime of a lease from incremental capacity of 273,000 workstations. […]

This presentation is illustrative and not necessarily indicative of the actual annual membership and service revenue or the actual annual contribution margin that would be generated from an investment of $1 billion in net capital expenditures or from the addition of 273,000 workstations. In addition to assuming 100% utilization of workstation capacity, this illustration assumes no maintenance capital expenditures, no pre-opening location expenses, no sales and marketing expenses, no general and administrative expenses and no growth and new market development expenses. […]

Represents implied annual membership and service revenue based on illustrative workstation capacity of 273,000 multiplied by illustrative average revenue per WeWork membership of $6,320. […]

Net capex per workstation added for the first half of 2019 reflects gross capital expenditures of approximately $1.3 billion, less tenant improvement allowances of approximately $0.5 billion divided by 273,000 workstations delivered, in each case in connection with projects completed during the period from January 1, 2019 to June 30, 2019. […]

We’ve included these in the order that they’ve been mentioned in the S-1. Now, you’ll note that the first entry notes a spend rate and a resulting workstation add number. But it doesn’t fully imply that 273,000 is the number that were added in H1 2019 (though it is hinted at).

The next occurrence makes up some math regarding what those workstations could generate in positive contribution margin. The proximate decouples the $1 billion sum from the 273,000 workstation figure that the first use of the metric raised (it also goes on to show why the math is trash, which is why I included a longer quote). Penultimately, we see the 273,000 used again to explain some mathmagic.

And then, finally, in the fifth entry, we see that the 273,000 figure is the number of “workstations delivered” in H1 2019. Why it was so hard to get that number is, in my view, illustrative of what was wrong from the start with the WeWork S-1. It shot for excessive cleverness, which, according to the useful cliche, usually fails.

In contrast, the 106,000 workstation figure that the Journal cited from a later WeWork S-1 comes up just once in that document:

Net capex per workstation added for the first half of 2019 reflects gross capital expenditures of approximately $0.8 billion, less tenant improvement allowances of approximately $0.4 billion divided by 106,000 workstations delivered, in each case in connection with projects completed during the period from January 1, 2019 to June 30, 2019.

Whoops.

Why does this matter? Because WeWork had negative investing cash flow in H1 2019 of $2.36 billion. How many new, possibly revenue-generating workstations the company got out of that expense really matters. The later, worse numbers did not help the company’s case that its “model” works.

You’ve done well getting this far so early in the morning. The lesson from all of this is that if you have to dance around your company’s actual results, they probably aren’t very good.

And please report your numbers in a way that every investor can understand. No business is so unique that cost of revenue, gross profit, and the like aren’t useful metrics. Even Google, a highly non-traditional company that went public with a reverse dutch auction (remember that?) listed its revenue costs in its S-1 in a manner that was grokkable. It isn’t impossible.

Illustration: Dom Guzman