Description

Alteryx, “SaaS” with a giant asterisk.

AYX Is a $7bn+ market-cap on-premise software company regarded as a SaaS-darling by the market. We initiated our short on a competition thesis, as we believe the shift of analytics workloads to the cloud could reduce the need for licenses by a factor of 10:1. However, our on-going work revealed accounting that we believe is the most aggressive we have seen in years and widely misunderstood by the market.

We believe that reported revenues are over-stated by ~50% when compared to ratable SaaS peers, that revenue growth overstates the true organic business momentum by a factor of at 1.5x, and that the aggressive accounting makes it more likely that growth will cliff in 2020. We wonder why AYX parted ways with their auditor in January of this year, in a manner for which we can't find a public company precedent, the very quarter they switched accounting standards (ASC 606) that enabled them to aggressively pull revenue forward.

Fundamentally we think this is an architecturally challenged on-premise software company trading at 14x forward revenue fully diluted, and we believe at more than 20x pro-forma forward ratable revenue if its accounting was apples-to-apples to the high-flying SaaS companies the market has placed in its comp-set.

We believe the revenue pull forward borrows from future periods, setting the stock up for a spectacular fall as the company faces increasing competition and an architectural crisis as analytics workloads move to the cloud. Despite the fact that AYX is down from all-time highs as high-growth software has sold off in recent months, we believe that AYX is a more compelling short today, now that the competitive and addressable market issues are becoming near term, and as the market is just barely starting to realize how aggressively AYX has pulled forward revenue using ASC 606. 

There are three principal ways we believe AYX is using ASC606 to over-state revenue relative to ratable SaaS comps and the true organic growth rate of the business: (i) They are recognizing significant portions of 3-year contracts upfront immediately as revenue, vs. ratably over the course of 3 years, (ii) they are selling an increasing mix of 3-year contracts instead of 1-year contracts, which allows them to pull even more revenue recognition up-front, and (iii) not only are they pushing for longer contracts, they are recognizing an increasing portion of long-term contracts up front.

1)  AYX is recognizing a significant portion of 3-year contracts upfront, vs. ratably over the course of the deal:

a)  As an example, SaaS company X signs a 3-year contract for $100/year on Jun 30th of 2019. A typical ratable SaaS company recognizes ~$50 of revenue in 2019 (6 months of revenue recognition) but we believe that AYX is recognizing $150 of revenue in 2019 in this example, 3x higher than the typical SaaS company.  This is in effect borrowing from future periods, as for the remaining 2.5 years of the contract post-2019, AYX will only recognize $150 of revenue vs. SaaS company X recognizing the remaining $250.

b)  To quantify the degree this up-front recognition benefits near-term revenue, the day that AYX adopted ASC606, they restated 4Q18 revenue from $61mm to $89mm. We can't think of another software company with this material of a change to revenue from the stroke of the ASC606 pen where management refuses to disclose ARR and instead explicitly instructs investors to look at revenue to measure the momentum of the business. But this is just the tip of the iceberg. 

c)  We believe the market is complacent because ASC606 is a complex topic and many investors incorrectly assume that ASC606 was applied evenly by all software companies (making rev rec to comps apples to apples) but AYX’ accounting treatment stands out by orders of magnitude no matter how we look at it.  

i)   For most SaaS companies, revenue changes under ASC606 were minimal as rev rec remained ratable before and after. There was a quirk in ASC606 for on-premise term-licenses that allowed AYX to pull rev rec up-front that wasn’t available to most ratable SaaS companies.

ii)  Even for the subset of software companies we track that sell on-premise term-licenses, AYX stands out:

(1) Under 606, there is a new balance sheet item called a "Contract Asset". It is the opposite of deferred revenue. It represents "revenue that has been recognized before the customer has been billed". You might need to read that definition twice. It is not a typo. 

(2) Contract Assets under 606 (i.e., revenue recognized "early")  as a % of quarterly revenue for on-premise term-license software peers we cover average 2%. For AYX this number is 49%.  

2)  AYX is selling an increasing mix of 3-year contracts instead of 1-year contracts, which allows them to pull even more revenue up-front (even if longer-term contracts are sold at a discount) as they recognize a substantial portion of the total contract value (not annualized contract value) immediately as revenue.

a)  Management acknowledges that duration of contracts has trended up in recent quarters but insist that contract term remain "around 2 years". AYX is the only software company we know impacted by contract duration that provides contract terms rounded to the year.

i)   What does "around 2 years" mean? Moving from 1.9 to 2.1-year average contract duration could be a meaningful (~10%) tailwind to revenue. From 1.8 to 2.2 years would be an even bigger one. 

ii)  All other software companies that we track provide duration in either months (e.g, PANW, VMW, etc will tell you duration for billings was 22 vs. 24 months) or months rounded to 1 or more decimal points (e.g., NOW, etc.  will disclose duration was 11.9 months vs. 11.7 months)

b)  AYX' accountants/lawyers appear to agree that contract duration could be a material impact on reported revs.

i)   In the most recent 10-Q filed November 1st, the risk factor that used to state that AYX might not be able to attract new customers or retain existing customers, now adds as a risk to revenue " and maintain the subscription amount and subscription term to renewing customers" and further added  "New customers may enter into license agreements for lower subscription amounts or for shorter subscription terms than we anticipate, which reduces our ability to forecast revenue growth accurately".  We wonder why the auditors/lawyers felt strongly enough to add this new language around subscription term (i.e., duration) if the impact is immaterial.  

3)  Not only is AYX selling more long-term contracts where they can pull forward more revenue, we believe AYX chose an accounting treatment that is increasingly recognizing a larger portion of those long-term contracts up front. 

a)  The % of long-term contracts that is recognized on Day 1 as revenue has moved from ~35% last year to ~40% this year, based on management’s disclosures on recent calls. This represents a nearly 15% tailwind to reported revenue simply due to management’s choice of accounting treatment. 

Putting this all together, we believe that on an apples-to-apples revenue basis to SaaS peers, the upfront rev rec issue inflates reported revenue by ~20% vs. comps, and then in addition the issue of increasing contract duration and recognizing a higher percentage of the long-term contracts inflates revenue by another 25%+ relative to SaaS peers. We believe the market is just starting to ask questions about AYX' true organic growth rate and what revenue multiple is "fair" for that growth. But investors should also be asking what revenue base the multiple should be applied to. 

We believe the way AYX is recognizing revenue like a classic license/maintenance software business rather than a ratable SaaS business. Historically, investors have placed an order of magnitude lower revenue multiple (e.g., 6x vs. 12x, all else equal) on high-growth license/maintenance revenue because of the difficulty in managing the "growth treadmill" of license sales. The appeal of ratable revenue is that you start with the next year with that base of recurring revenue (less some churn), and new business is additive.  Under a license revenue model, you first sell the same number of "widgets" that you did last year, and only then is the next sale additive. We think that the market is starting to recognize that, perversely, the bigger the revenue beats are today the more revenue that AYX is borrowing from the future.  

The AYX setup reminds us of its predecessor analytics darling Tableau in 2015, which at the time reported license and maintenance revenue. We could not understand why it seemed like we were the only ones scratching our heads that Tableau circa 2015 traded at 12x upfront recognized license/maintenance revenue. All it took was for Tableau’s quarterly beats to narrow and growth to slow from 64% to a still respectable 42% for the stock to promptly de-rate to 2-3x sales when it reported 4Q15.   

Why did AYX fire their auditor in January 2019, the very quarter they adopted such a material change in rev rec?

AYX abruptly fired PwC as their auditor in January of 2019, the first quarter they were to report under AYX' interpretation of ASC606. Over the last 5+ years, we can't find another example of a $1bn+ public company with a December year-end that fired their auditor in January.  If simply firing your auditor is a red flag, I'm not sure what analogy to draw with a company that fires their auditor the quarter they undergo such an extensive and aggressive change to revenue recognition. 

AYX' explanation was that PwC was so enamored with AYX' product internally and had such high hopes for the consulting arm to re-sell AYX product that they notified AYX in mid-January that they may lose their independence as an auditor. AYX accepted this explanation and fired PwC a week later. We find it puzzling that PwC's audit division counts CSCO and DELL as customers, as we find it hard to believe that PwC doesn't internally use CSCO routers, or that their consulting arm does not re-sell significant amounts of DELL servers or PCs.  

Not only was the timing suspect because of the switch to 606, but PwC was fired just as they were obligated to file their first 404(b) letter. Section 404(b) is one of the cornerstones of the 2002 Sarbanes-Oxley act, where auditors must attest to adequate and effective internal control structure at the audited company for purposes of financial reporting. The reason we bring this up is that according to AYX’ prior 10-Ks, it noted that it identified a material internal control weakness in 2016. This took over a year to be addressed and the management team was ultimately the arbiter of success as "our Chief Executive Officer and Chief Financial Officer, concluded that we have remediated the material weakness described above as of March 31, 2018." We note that at that time PwC did not need to issue a 404(b) assessment as AYX still qualified as an emerging growth company under the JOBS act. AYX no longer qualified for such status for purposes of the 2018 10-K, and PwC was obligated to submit their first 404(b), which would have greatly increased PwC’s financial and criminal liability to the adequacy and effectiveness of AYX' internal financial controls and reporting. 

To recap, AYX fired their auditor in January, a move for which we can't find a public company precedent in recent history. The timing coincided with the first time PwC had to issue a 404(b) letter, which under Sarbanes-Oxley would have opened PwC up to financial and criminal liability in attesting to there being no material deficiencies in AYX' financial reporting controls, an area where the AYX has a history of identified issues, in the very same quarter where AYX adopted a 606 rev rec interpretation as described earlier that we believe is orders of magnitude more aggressive than any other software company we cover. 

If it was any one of these items in isolation, we might raise an eyebrow and move on. But the totality of the circumstances that surround PwC's dismissal make us question if there is a different reason why PwC could not file the 10-K or attest under 404(b) to no material deficiencies in AYX’ accounting. We will leave drawing conclusions as an exercise to the reader. 

Catalysts for AYX’ growth rate to cliff in 2020

Now that we covered accounting, let’s discuss how we believe the ‘story’ is likely to collapse even if auditors do nothing about the aggressive accounting. We believe that competition is increasing, and that AYX faces a massively misunderstood architectural crisis as analytics workloads move to the cloud – one that based on our channel work, could reduce the need for licenses by a factor as high as 10x.  

Traditional competition

Bulls and sell-side will cite attractive leverage to AI/ML use-cases but we believe AYX is primarily a self-service data-prep tool for business analysts. Our channel checks consistently tell us that AI/ML drives the minority of the use cases, that the tool is too simple for PhD caliber data scientists (they tend to stick to R or well-funded startups such as DataRobot). The use cases we hear most from users involve business users cleaning and “prepping” data, for instance when dollar values are listed in text format, or when dates are listed in European vs. US order, etc. and then offloading that ready-to-consume data to a visualization tool such as Tableau or PowerBI.

We note the AYX list price of more than $5,000/year/user, which is quite expensive compared to the MSFT Office Suite at ~$240/year/user, PowerBI at $120/year/user or Tableau at ~$850/year/creator.

Tableau launched a similar feature, Tableau Prep, which was an overhang on AYX’ stock for a few quarters, but we believe this fear largely dissipated in the market as AYX’ reported revenue growth started to accelerate after they adopted ASC606.  Conventional wisdom now dictates that Tableau Prep is not competitive because it only exports to Tableau’s visualization tool whereas AYX, acting as ‘Switzerland’, can connect to any visualization tool. We believe this is not an AYX architectural or technical advantage but rather a strategic decision made by Tableau management that can be easily reversed under Salesforce ownership.  For instance, AYX doesn’t have a direct connector to PowerBI, but allows the user an “export to .csv” equivalent function, one that is trivial to replicate. We note that Salesforce just this month cleared global anti-trust approval for its acquisition of Tableau, which now enables them to integrate and market the Tableau Prep product across its user base, likely at minimal if any cost to the customer. For Microsoft-heavy shops, you have Power BI doing very well, and an improving Power Query service that preps data at a dramatically lower price than AYX, also often bundled in for free.

What we see is a high-end AI/ML market where competition is heating up from well-funded private companies such as DataRobot or tried and true options such as R. And we see mass-market use cases increasingly getting crowded by incumbents such as Salesforce and Microsoft, who are likely driven by ubiquity of their respective data products to create stickiness for their bundled offerings, rather than stand-alone ASP.

But this form of traditional competition is not AYX’ biggest problem. We believe that Analytics workloads moving to the cloud are a misunderstood architectural crisis for the business as it can greatly reduce the need for AYX licenses.  

Analytics workloads moving to the cloud and “the Snowflake problem”.

First a bit of context, where we acknowledge that AYX brilliantly used two factors to its advantage from 2015-2018.

The first is that the company was positioned incredibly well for a  three-year rush of fervid increases in software spend, particularly in Analytics; one where G2000 companies were seemingly willing to throw any buzzy software startup at the wall and only now are starting to decide what will stick.

Second is that the “self-serve” sale to the business user has always faced competition from an internal alternative -- the IT department has had a way to address this problem of “dirty data” for decades. But the historical way IT solved this problem, through a variety of legacy tools such as ETL and data warehousing (e.g., Informatica and Teradata), was slow and frustrating, and not up to par as business user expectations increased due to the consumerization of enterprise software. From 2015-2018, we saw enterprise IT teams collectively place their eggs in the on-premise Hadoop basket (i.e., CLDR and HDP) to try to improve the user experience of accessing data for purposes of Analytics. Unfortunately, for anyone that has followed the space, on-premise Hadoop just did not work. If multiple teams tried to access clean and prepped data, resource contention grew exponentially resulting in slow and often unusable load times.

Not only was 2015-2018 a golden age of seemingly unconstrained self-serve Line of Business software budgets, it coincided with a period where the alternative offered by IT went from slow and inconvenient to unstable and unusable.

So what has changed? Snowflake, “the hottest US Startup of 2019” according to LinkedIn data scientists, is a cloud-based data warehouse that is gaining widespread adoption in the enterprise. Tech investors have likely heard the name more this year because it poached ServiceNOW’s rockstar former CEO and CFO (and many senior sales staff).  The checks and surveys on Snowflake adoption trends have been out of this world for a while now but have recently hit an inflection point. Snowflake stores data in a way that is prepped and ready to consume. We believe that in a Snowflake world, instead of 10 different teams at the same company looking at similar datasets, buying 10-20 AYX licenses (1-2 per team), and independently duplicating the cleaning of that data 10 times, now you need 1 or 2 licenses to load that data into Snowflake once, where it is stored in a prepped and consumable format, ready to be directly ingested by Tableau or PowerBI. Industry experts we speak to believe you should actually use something less fully-featured but much cheaper like a Matillion or something with cloud-native support for data compliance and governance such as Talend to load data into Snowflake, which would further reduce the need for AYX licenses to 0.

We think it will be incredibly difficult for AYX to navigate this pivot to the cloud. It is a desktop product that owned its product category because most startups entering the space in recent years were spending their R&D dollars to build for a cloud-native architecture rather than an on-premise one. The former employees we have spoken to believe that Ned Harding, cofounder and CTO at AYX for 23 years, who personally coded AYX’ original core, was the technology visionary at the company. He left his post as CTO over a year ago, in our view further lowering the probability of that AYX can re-invent itself successfully to a cloud-native architecture.  

And this brings us full circle to where we started the writeup on accounting. We believe that AYX has been able to pull forward a significant amount of revenue relative to the SaaS peers to which it is comp’ed, due to a quirk in 606 accounting for on-premise term licenses, one not available to ratable SaaS companies, precisely because AYX is an on-premise license and maintenance desktop application.  We believe competition is intensifying on-premise but that the real challenge for AYX is that analytics workloads are moving to the cloud. We believe the closest analog is the very prior analytics darling stock, Tableau, which tumbled from 12x license/maintenance revenue to 2-3x when growth slowed abruptly from 60%+ to 40%+.

We believe that AYX is an architecturally challenged on-premise software company trading at more than 20x pro-forma forward revenue if its accounting was apples-to-apples to the high-flying SaaS companies the market has placed in its comp-set. We see 50% or more downside to the stock when the market begins to fully digest the accounting treatment and changing competitive environment.

Catalyst

.