Catching up with the VP of Corporate Development at Hortonworks; a bit of historical perspective.
Last week, I had the chance to interview Brian Marshall, the VP of Corporate Development for Hortonworks (NASDAQ:HDP). Brian had been a partner and a senior managing director of ISI, one of the leading independent research-driven equity sales businesses.
I had initially written about HDP and its set of elephants back at the start of May, just before the company reported its Q1 earnings. The timing really might have been a bit better, it is fair to say. The shares plunged 19% in the two trading days after the earnings release and have never recovered to any significant extent. Since just after the earnings release, HDP shares have increased by 5% compared to the IGV performance over that same span of 5% appreciation. Overall, the shares are down 47% since the start of the year. Since the start of the year, the IGV software/tech index is up about 1%.
Since the start of the year, HDP has enjoyed two quarters during which revenues grew by 55% and then 85% respectively with billings growth in the two quarters of 148% and 122%. Q1 results showed revenues about 5% ahead of guidance and billings 8% ahead of guidance. Full year expectations were left more or less unchanged despite the Q1 beat.
The reason behind this significant underperformance has been that of cash burn and potential dilution. When trying to analyze the price movements of shares with high valuations, many factors can be responsible for their movement. By their nature, shares in small loss-making IT companies have been historically volatile and usually respond to the slightest deviation from expectations in a very negative fashion. And that is really what has happened to Hortonworks in the 1st half of 2016.
In the last quarter of 2015, HDP sustained an operating cash flow loss of $19 million, which left the company just $94 million of cash and equivalents. In conjunction with a pre-release of Q4-2015 results, the company also announced a secondary offering for $100 million. Analysts' comments at the time were that while a capital raise had been anticipated, the timing was far from optimal. At the time of the announcement, HDP shares had already declined by 29% from the start of this year by that point.
The shares plunged another 50% to around $7.75 in the wake of the announcement. (Q4 results were an upside of some noticeable amount compared to prior consensus expectations with regards to revenues and bookings.) Ultimately, HDP had to sell 10 million shares, or about 21% of its then current outstanding shares to raise almost $100 million. By the time the secondary was actually priced on February 2nd, the market had declined further and proceeds to the company were just $9.50/share. The shares actually fell further to a post IPO low of $7.74 at the trough of the market reached on 2/9.
HDP announced its Q1 results on May 4th. As mentioned above, the results were a visible beat on revenues and earnings were in line with the prior consensus forecast. CEO Rob Bearden reiterated the company's pledge to reach non-GAAP EBITDA by Q4. The Q1 non-GAAP EBITDA loss was $21 million. The saying that the devil is in the details was never truer than with this release. Investors saw an operating cash loss of just less than $36 million.
They saw the company's cash position at $124 million despite a capital raise of $91 million net of costs. (The company did purchase $22 million net of long-term investments that are not counted in the cash balances.) And in exchange for that, the company reiterated a promise to get operating cash flow to a break-even status (Adjusted EBITDA and cash flow track quite closely for this company.) by the end of the year and raised revenue guidance by a relatively modest amount.
The company's stock-based comp, in the wake of the prior share price performance, shot up from $5 million in Q1 to $29 million this past quarter. Of course, stock-based comp is not a cash charge. But the ultimate dilution is real enough. But another item that got the attention of investors and this writer was the $11 million increase in the quarter to a level that represents DSO of 142 days.
Needless to say, investors were not pleased. The shares melted like ice in the sun on a summer afternoon. The share price tumbled from $12.78 to $10.31 in two days. The short interest has increased 14% over the last month. At this point, the short interest is up to almost 21% of the float.
The Devil was in the details.
Overall, while revenues increased by about 10% sequentially in Q1, accounts receivable increased by more than 20% in the quarter. Management explained that the preponderance of the increase in receivables was a product of a large deal with a US Govt. customer who didn't pay its bills. The amount not paid by this one government customer was said to be in the range of $10 million. It would have made a significant difference, I believe, to the price of the shares when the quarter was announced if that installation had been paid for. It will make a significant difference to the price of the shares when Q2 results are released in about 6 weeks if the installation is paid for.
Brian declined to discuss whether or not the government customer had paid its bill thus far during the quarter. I would have loved to understand why and how it was that a government customer had been able to slide on a $10 million bill payment - but honestly I have seen and experienced worse over the years. Our government at work - or not work - depending on how you evaluate these situations.
Brian said that the company hadn't released any information thus far about the payment and he declined to tell me if HDP had received its $10 million. Short of pliers, a dentist drill and some wires hooked up to a generating source, I wasn't getting any more specifics. My best guess - and that is what it is - is that HDP has gotten its payment but hasn't been able to issue a substantive press release.
As I explain below, Brian is well aware that one of the major issues regarding share valuation is cash burn and he seemed quite confident that the company was on track to reach its objective of more or less eliminating cash burn by year-end. I interpreted his comments on the subject of cash burn and the current state of the business positively and inferred that the $10 million payment issue has been resolved.
The other major reason that I think payment was received is that Northland Securities chose to initiate coverage on Hortonworks this past Friday with a buy rating and a $24 price target which is more than a double from the current price. Given that this payment (the ten million owed by the government customer) is really the equivalent of a 600-pound gorilla in a pretty small room, at least in terms of short-term share price performance, I have to imagine that the analyst must have been reassured to some extent, in order to make such a call.
Some significant other topics during the interview
We spoke about competition in the overall Hadoop market space and why HDP seemed to be gaining market share in the last couple of years. Brian said that he believes that HDP had passed Cloudera in terms of installations and in terms of market share sometime earlier this year. At one point dating about 2 years ago, Cloudera was as much as 40% larger than HDP.
Brian believes that the success of Hortonworks in the market has been because it has remained more committed to the "open source" philosophy in providing tools for users to build Hadoop applications than the competitors. He feels that the relations that Horton has with Apache, and the use that HDP makes of the data management layer of Hadoop called YARN has allowed the company to capture mindshare amongst users looking for the most advanced big-data technology.
Needless to say, there are arguments that the enhancements that both Cloudera and MapR have made to their versions of Hadoop have sufficient value that users have been willing to abandon the purity of their approach to the open-source paradigm. I do not propose to recapitulate that discussion and it really means very little for HDP's share value at this point.
When I wrote about Hadoop a few weeks ago, some commentators on the Seeking Alpha site had significantly different views on the future of the technology. I really do not purport to be a data scientist or someone who has some deep technical understanding about all of the emerging technologies that are meant to handle the storage of big data. At the time, I simply referred to all of the many articles regarding the prospects for the Hadoop market from the Gartner IDC and many other research analysts on the size of the Hadoop market.
Of course, that is a debate that no one will ever be able to win in prospect. I don't purport to have an independent view on the subject, but the results of the major participants in the space seem to suggest that Hadoop will be a large market and will continue to enjoy significant growth rates at levels almost unique in the IT space.
But I wanted to see how Hortonworks' management might respond to criticism about the prospective growth of their market. So I put the question to Brian as to how he would answer doubters regarding the future of Hadoop. He told me that he thought the success that Hadoop was having with its global summits was some indication of the interest that developers have in trying to understand the technology and use it in big data projects.
The Hadoop Summit for the US is scheduled for the end of this month in San Jose, CA and attendance is forecast to exceed 5000 professionals. A summit was held earlier this year in Dublin that had significant attendance. These are very expensive events compared to most other industry conferences. The admission to the conference in Dublin was $1100 +VAT. The cost of the conference in San Jose is $1750 including attendance at Hadoop meetings and conferences or $600 for entrance to the exposition hall - at least there is no further sales tax to worry about.
Brian suggested that a conference that was to be attended by 5000 professionals paying an aggregate of $5 million in admission fees + costs for training was an indication that corporate IT departments certainly believed that Hadoop would be a significant component of the big data landscape with a broad set of solutions/applications built on the open source platform and the ancillary technology. In fact, he told me that one of the items that related to the negative operating cash flow that was reported in Q1 was the more than $2 million of deposits that had been made by Hortonworks to cover their booths at the exposition.
He directed my attention to Hortonworks' plan to extend and refine the company's Global Professional Service (GPS0 in Hortonworks' speak) in order to keep up with demands amongst HDP users to successfully deploy their solutions. According to Brian, users are looking for both measurable and achievable results and they are looking to their data software vendor to help them implement and to measure their Hadoop projects. No lack of industry buzzwords in this discussion.
I asked Brian about what appeared to be conservative guidance by management regarding Q2 expectations. To a certain extent, guidance seemed to be unrealistic to me based on the results of the prior quarters and the supposed momentum of the technology in the market.
Revenues in Q1-2016 were $41 million, up from $37 million in Q4-2015. The forecast for revenues in Q2 is for $45 million, or 10% sequential 4th quarter growth. Although, HDP is a small, development stage company, where seasonality is not a significant factor, it has always been the case, that the Q4 to Q1 comparison for all companies in the IT space, not just Hortonworks, almost invariably show the lowest percentage growth of the year, Normally, Q2s show substantially greater sequential percentage growth in revenues than are seen in Q1. Further, Q1 billings showed only a minimal percentage increase compared to Q4. This quarter, bookings are forecast to show a 17% sequential increase from Q1 to Q2.
Brian spoke about how Hortonworks had guided to 18 numbers since it became a public company. The company guides to revenues, bookings and Adjusted EBITDA for each quarter. Of those 18 numbers, the company had beat 17 in the time it has been a public company. Brian said that company management felt that under-promising and over-delivering was basically the company strategy with regard to guidance. That is not too dissimilar to many other companies who attempt to cater to investor sentiment by attempting to insure that each quarter's results are reported as a beat. I think that Brian's long-time background as an analyst is part of the company's thought process regarding guidance.
Brian made mention on several occasions regarding the enthusiasm engendered by the upcoming Hadoop Summit where 5000 attendees are to come to a multi-day event centered entirely on Hadoop technology. Such events are often a venue in which large deals are done and this one likely will be no exception. Based on all of the commentary that Brian made about the conservatism of the guidance and the current state of business, the pipeline and his expectations regarding the signing of large contracts during the summit, I would be more than a bit surprised if Q2 were not at least a noticeable beat in terms of revenues and bookings.
Adjusted EBITDA is forecast to be an $18 million loss in Q2 and was a $21 million loss in Q1. Given that the company has estimated a full year EBITDA loss of $52 million, then the Q3 EBITDA loss would need to be $13 million followed by the forecasted Q4 break-even. I would guess that such a progression is an extremely aggressive ramp and so I imagine that Q2 EBITDA results will also be significantly greater than expectations.
One of the more significant components of Hortonworks' financial statements is the rather substantial variance between billings and reported revenues. Last quarter, the increase in deferred revenues of $11.7 million was 28% of reported revenues. Overall, the deferred revenue balance is $119 million or some 62.6% of this year's anticipated revenues. Both of these numbers are significantly higher than is the case for most other software vendors and speak to a better quality of operating results than many might imagine.
Brian said that the company had never sought to get what is called a vendor-specific objective evidence. He said that the principle reason that HDP was not trying to get a VSOE policy established is because there is going to be another major alteration in the FASB standards in 2018 regarding the elimination of VSOE criteria. For readers interested in the specifics of the new revenue recognition standards that eliminate the VSOE criterion, please go to the Bloomberg article that is referenced here.
The major change for many vendors, and for Hortonworks in particular is "that the inability to establish VSOE of fair value will not preclude an entity from identifying separate performance obligations in a software arrangement and allocating a portion of the transaction price to each obligation." Further, "a single contract may contain a different number of performance obligations than under current U.S. GAAP."
Last year, HDP reported gross billings of $166 million and GAAP revenues of $122 million. That difference of $43 million increased the net loss significantly compared to what it would be if a greater proportion of billings were recognized as revenue. Because the HDP portfolio is growing and because users are buying a more complete suite of solutions, the percentage different between revenues and billings continues to grow. Last quarter, revenues were $41 million, while billings were $54 million.
The lack of accepted VSOE standards is significantly increasing the reported loss. Last quarter, 20% of the GAAP net loss was a function of the lack of VSOE standards. Brian suggested that when the new draft goes into effect in 2018, the difference between billings and revenue will substantially narrow and which will, thereby, significantly add to revenues.
Many, many readers dislike (dislike is hardly a strong enough word but one that will have to suffice in a professional publication) the use of non-GAAP and want to see analysis based entirely on a GAAP earnings presentation. And for HDP, stock-based comp increased substantially in the wake of HDP's share price decline. The disease of stock-based comp is firmly entrenched in employment costs in the IT space. Most start-ups and young firms can only attract the talent they need by generously providing some form of options or stock appreciation rights. And if the rights or options are underwater, most companies are going to have to "fix" their compensation scheme by changing the option strike price.
I asked Brian about the company's concerns on the subject of dilution. Last quarter, stock-based comp was $29.4 million. That was no less than 71% of revenue. Stock-based comp grew by 71% sequentially and it increased by 6X year on year. I asked Brian how the company plans to deal with the ballooning of this expense. He said that the company has been assiduous in insuring that stock-based comp represented a maximum of 5% dilution in any given year. I am not exactly sure as to how that calculation works. There are, and will continue to be, different points of view on that subject. At the moment, there is a GAAP calculation, which may or may not reflect the real cost of the options to existing shareholders. Unfortunately for me, my allotted time ran out at that point in the discussion.
I asked Brian, who as I had mentioned had been a well-known analyst before he took his present position with Hortonworks, what he thought was the bearish case for the shares. He said that many current investors had told him that the short thesis was that cash burns would continue unabated for the balance of the year. The current cash and equivalents balance including long-term investments is $149 million. The cash burn in Q1 was $35.7 million. So, the theory is that HDP will run out of cash and needs to raise more money in about 3 quarters or so.
Obviously, Brian vehemently denied that could or will actually happen. The cash burn ought to drop by something like $15 million in Q2 representing both operational improvements and also representing what Brian describes as one-time impacts on cash flow such as the unpaid government invoice and the deposits for the Hadoop Summit. The cash burn is expected to drop further in Q3 and then more or less disappear entirely in Q4. If it happens that way, HDP will have minimum cash balances of greater than $110 million and it ought to start generating operating cash flows next year.
Some key takeaways from the interview:
- HDP shares have been pummeled because of the company's Q1 cash burn in conjunction with a raise that diluted then current shareholders by more than 20%. It seems likely that 1/3rd of the cash burn related to one-time and indeed the growth in DSO and hence receivable balances should be reversed this quarter.
- The competitive position vis-a-vis both MapR and Cloudera is quite favorable. The company has enjoyed significant traction that win the argument that it is the only Hadoop vendor that completely conforms to real "open source' standards as defined by the Apache Foundation.
- Management believes that the future for Hadoop, which includes very high growth rates for the foreseeable future is confirmed by the exceptional level of attendance at this year's European Hadoop Summit in Dublin and next week's summit to be held in San Jose, CA.
- Management reiterated that it looks to achieve break-even adjusted EBITDA by the end of the year. Adjusted EBITDA and cash flow are generally interrelated and the company expects that this close to 1 to 1 relationship will be maintained.
- The company generates a very high proportion of deferred revenue as a proportion of total revenue. This is mainly a function of the fact that Hortonworks has never done a VSOE analysis that would allow it to more readily recognize the revenue from multi-product deals based on the acceptance by its customer of individual products. There will be a new FASB policy that in effect does away with VSOE standards and this will accelerate revenue recognition for HDP in 2018.
- The company has deliberately chosen to provide essentially conservative forecasts with the expectation that they will be well beaten. So far, the company has issued forecasts on 18 metrics over its 18 months as a public company. In that span, it has beaten 17 of the forecasts. I drew a strong inference that Q2 results would continue that pattern.
- The company's stock-based comp. relative to revenues is enormous and stock-based comp has risen rapidly in the last couple of quarters. Management maintained to me that even at current levels of stock-based comp, the dilution to present shareholders would be limited to 5%. I am not too sure how that calculation works. Unfortunately, my time ran out before I was able to understand the calculation.
- The short case for the shares is a function of the Q1-2016 cash burn and the company's then current cash balances. Management suggests that cash burn will be greatly reduced during the next two reported quarters and will be more or less extinguished by Q4 this year.
HDP shares have not been one of my better recommendations this year. I think that the upside potential is enormous as the growth expectation for this no-Sequel, open-source database that enables real big data applications is realized. In the shorter term, shares are going to rise or fall significantly based on Q2 cash flow. My belief is that the elements that lead to the $35 million of negative Q1 cash flow are partially one-time and partially a function of a level of investment that has been growing faster than the very high revenue growth rates this company has enjoyed. If the company beats, as I expect, if it achieves a much smaller cash burn as I expect and modestly raises guidance about which I can't give odds, I expect the shares to be a strong performer and to produce significant positive alpha.
Disclosure: I am/we are long HDP.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.