I like Splunk. I like Godfrey Sullivan and what he’s done with the company. Steve Sommer is a great marketing guy and I think he’s done a superb job with Splunk’s marketing, particularly in imbuing the company with a hip, fun, consistent corporate personality, making them the Virgin Americas of log file analysis.
I also like Splunk because many months ago, they let me riff with Godfrey and many members of the e-staff about marketing and strategy. They were smart and it was fun. They even gave me a superb bottle of wine for my troubles.
I like Splunk because, unlike Jive, Godfrey hasn’t turned the e-staff into a crony club. Building great teams is about finding the right people for the right job, not just carrying around an entourage. Exercise: search Spunk’s management page for Hyperion and then search Jive’s for Mercury. (Answer: 2 and 6.)
I also like Splunk because they pivoted. When I first heard of them, they were positioned as “IT search.” I had no idea what that was or who would buy it. When we met for the strategy riff, they were in middle of re-positioning around machine-generated data, a message that I liked.
Most folks make software that analyzes human-generated data; we focus on machine-generated data.
Clear, simple, and true. Instead of piling on as a YABDW (yet another big data wannabe), Splunk built a message that was sexier than “log file analysis” but still true to their essence and still generalizable to a broader vision of “operational intelligence.” A+ marketing. Bien fait.
Finally, I like Splunk because they haven’t burned through lots of cash. They’ve raised $40M. They have $23M in the bank. $17M net burn isn’t bad for what they’ve created.
Splunk’s fact sheet does a great job of telling their story in two pages.
When I heard that Splunk had filed their S-1 to for an initial public offering, it was no surprise. I’ll spend the rest of this post analyzing it and pulling some highlights. Those looking for controversy will not be happy. I’ve read the S-1 over the past few days and found few surprises. Overall, the company looks pretty clean from where I sit.
Let’s start with the income statement:
- FY11 revenues of $66M, up 88%
- Trailing 9 month (T9M) revenues of $77M, up 78% so there’s a slight deceleration in growth
- FY11 gross margin of 89%, on the high side for an enterprise software company and reflective of the high license revenue mix (75%)
- S&M expense in FY11 of 60% of sales, which rose to 62% for the T9M period. They’re not afraid to spend on growth.
- Healthy R&D expense of 21% of sales in FY11, which stayed roughly constant.
- Small operating loss of 5% in FY11, rising to 10% for the T9M period, probably due to costs associated with the offering.
In my opinion, this is a VC’s dream income statement (with one notable exception that we’ll cover in a minute).
- High revenue growth = big opportunity
- Small operating loss = sustainable, but spending it all on growth.
- Small net loss = nowhere to go but up in profitability
- High license mix = software-focused
The only part of that VC formula I dislike is the license mix. Boards like high license mix because market share is measured in license dollars, license dollars are seen as the engine of a software business, license dollars drag other dollars with them (e.g., maintenance), and finally because boards get tired of companies missing their high-margin license target and covering the gap with low-margin services. They see it as soybean filler in their hot dog.
I think that view is myopic because it is not customer-oriented. To the extent your software is truly easy to use and requires few services, then I guess it’s great. But to the extent you are selling typical enterprise software, it can be hard to use, setup, and configure. In that case, keeping your services org tiny may win you cheers at the board meeting, but jeers at the user conference.
Personally, I’d like Splunk even better if they had the same license revenue and more service revenue on top, even though it would reduce the license mix and gross margin percentages. (Note: not gross profit, both revenue and gross profit would be higher in my ideal company, but the license mix and gross margins would be lower.)
I worry that Splunk could end up in No Man’s Land on the services issue: too small an opportunity to entice big consultants to build serious practices around the product, but too great a need to be satisfied by a small (6% of sales) professional services organization. For example, we are customers in my current job, and — far as I can tell — we get some good value from the software, but could probably get a lot more.
Now, if you’ll pardon the pun and the mixed metaphor, let’s find the cloud in the silver lining: Splunk is not a SaaS company. OMG.
For example, we typically enter into perpetual license agreements, whereby we generally recognize the license fee portion of the arrangement upfront, assuming all revenue recognition criteria are satisfied.
Personally, I’m OK with it. In some ways, I admire Splunk for swimming upstream on this issue. While SaaS is wonderful and has many advantages, not all customers in all categories want to buy on a SaaS basis. Splunk has evidently decided that in machine-generated data, people primarily want perpetual (yet they also offer term as an alternative).
Splunk’s sales are backloaded:
As is typical in the software industry, we expect a significant portion of our product license orders to be received in the last month of each fiscal quarter.
The combination of this backloading with the more volatile revenue stream associated with perpetual model should make Splunk’s earnings more volatile than its peers. We’ll see if that turns out to be the case. (See here for my generic analysis of SaaS vs. perpetual businesses.) But they do have some ability to manage it:
We typically ship products shortly after the receipt of an order. We may have backlog consisting of product license orders that have not shipped and maintenance, professional and training services that have not been billed and for which the services have not yet been performed. Historically, our backlog has varied from quarter to quarter and has been immaterial to our total revenues.
The astute reader will notice they’re saying that they may choose to not ship all orders at the end of quarter. This is common in perpetual software businesses both due to order volume and because experienced managers know that if you’ve hit this quarter’s targets it’s time to slow down the order processing desk. You’ll never know if you’ll need those orders next quarter, so let’s not put them on the midnight truck. And while a few million dollars here or there may be immaterial to revenues, in the case where expenses approximately equal revenues, these orders can have a big impact on earnings.
I always read but rarely analyze the risk factors. For Splunk, there are about 25 pages of them, in which the only tidbit I found was this:
We employ a unique pricing model which subjects us to various challenges that could make it difficult for us to derive expected value from our customers.
We charge our customers for their use of our software based on the customers’ estimated daily indexing capacity. As the amount of machine data within our customers’ organizations grows, we may face pressure from our customers regarding our pricing, which could adversely affect our revenues and operating margins.
I wasn’t aware of this, but it makes a lot of sense. Increasingly, software vendors want to sell the copier machine priced by the copy. The desire here is always to hook your pricing to “something that goes up” (e.g., the old MIPS-based pricing model). Splunk is betting that data volumes will go up and ergo that customers will need to buy more licenses as they do. This should help offset the volatility argument above — while existing customers aren’t setup on renewable contracts, if they have the perpetual right to analyze only half their data, I suspect they’ll be back ordering more software.
Let’s talk about equity now.
- It appears to me that there are about 1o2M shares outstanding before the offering based on 79.4M outstanding on 10/31 plus 23.2M shares issuable upon exercise of stock options.
- If they raise $125M in the IPO with a typical share price of $15, then that’s another 8.3M shares, so that means about 110M shares outstanding after the offer.
- This implies a target valuation of around $1.6B which I find high, so high, that I’m wondering if I’ve made an error. This article says the valuation may be $1B. Either I’ve mangled the math or they will reverse-split their way out of the problem. Either way, let me assume they’re targeting a ~$1B valuation after the IPO even though I can’t yet see how that pops out from the math.
Here’s a graph of Splunk’s common share price as seen by the strike price of options during the year.
While there are literally pages of math explaining how they calculate the fair market value (FMV) of the stock to me this curve looks a big flat and a bit low. The point of periodically performing section 409a valuations is to ensure that boards didn’t hand out in-the-money stock right up before the IPO. If the company goes public at $15 and even just stays flat, I’ll let you explain to the IRS and the SEC how the stock was really worth $4 in October and $15 in February.
The pages on compensation and the fees paid to compensation consultants always make me ill. The CEO loses a lot of control in the IPO process, consultants make a lot of money, and executive pay is not constrained in the process because the exercises are based on benchmarking. By the way, despite those general objections my reactions to Splunk executive compensation are:
- The bases seem reasonable
- The on-target earnings (OTEs) seem reasonable
- They have a highly leveraged compensation plan (Godfrey is a salesman, after all.)
- They blew out their numbers
- Ergo, they made a lot of cash compensation
Let’s look at the cap table.
The VCs own 70% of the company, which had raised a total of $40M in venture capital. If the company ends up with a $1B valuation, then the VCs will on average — which is both interesting and misleading because different people bought in at different valuations — get a 17.5x on their money. Not bad.
Godfrey Sullivan owns a hefty 8.1% of the company — quite a lot for a hired CEO (as opposed to a founder). This is because Splunk is successfully running what I call the “new VC play.” Because:
- Consolidation means there is an oversupply of very senior executives
- There are relatively few portfolio companies with the potential to go public
- It now takes 7-10 years as opposed to 3-5 to go public
- There is ample venture to fund the promising companies through IPOs that now happen closer to the $100M revenue bar than the $30M bar of the 1990s
You then go get the biggest guy you can find, load him up with options so a $1B CEO will run a $20M company, and then fund him for high growth over the long haul to get to the IPO. This is true of Jive (Zingale) and Splunk (Sullivan). It is true to a lesser extent at Lithium: Tarkoff was a billion-dollar GM (which isn’t quite the same league) though the VCs are certainly backing him with money.
Hence, if things go I think, my guess is that Sullivan’s share will be worth $80 to $100M after the IPO. Nice work if you can find it. Or, as I believe was the case in this instance, have the vision to see the potential and pick it.
I’m fizzling here about page 120 of what looks to be 175 or so pages. If you find anything interesting in those pages or have thoughts on what I’ve presented here, please share them.
And, in conclusion, congrats to the Splunk team and best of luck with the IPO.
(Be sure to read my disclaimers.)