anatomy of a thesis change

Investment Analysis Clubs Saul’s Investing Discussions
1

Cloudera reported yesterday, stock is down 40% as I write this. How can they be down 40% when they posted what is essentially a great quarter?
Revenue increased by 41% yoy,
subscription revenue up 50% yoy.
Net expansion rate 136%.
Gross margin 73% up for 69%

Their earnings call was a mess for one. In the prepared statements they tried to gloss over the fact that their revenue growth was going to decrease to 20%. Hidden in there they mention new leadership for sales and marketing, and then made a big hullabaloo about the next version of cloudera. They also focused on the fact they are going to moving away from land and expand and focus on their core customers.

Some questions I had below with my answers

How does a company with a net expansion rate of 136% only forcast revenue growth of 20%?
Buried in the analyst questions they drop the bomeshell that the 136% net expansion rate is only for their largest customers. WHAT? So basically they have been cherry picking the very best cohort to calculate NER on and reporting that information.

Why are they switching strategies?
Again in the analyst section they mention that their smaller customers have higher churn which in SaaS speak means their customer acquisition cost is too high and they aren’t getting enough revenue out of those customers. This is the case of selling something for a loss, then trying to sell as many of them as possible to make up that loss….you just lose more money. This HUGE bombshell basically means that they have been chasing customers, padding their revenue and customer counts but can’t do it anymore because they promised to be cash flow positive by 2020 and they won’t be able to do that with the cruddy growth they have been pursuing.

How can their growth be so low if they really have a such a huge TAM. (70+ billion estimated by cloudera)?
You guys may remember I didn’t believe Hortonworks TAM numbers and this really cements my opinion. Cloudera is saying they can’t make money on all but a few of the largest companies so in reality their TAM is much much smaller. So far hortonworks seems to be in a different boat.

Personally I find all of clouderas numbers to be suspect at this point. Cloudera finally had to pay the piper because they couldn’t hide their losses anymore and still reach some sort of reasonable cash flow by 2020 as they had promised. So what does this mean for hortonworks? HDPs cash flow is already better than clouderas. I’ll need to do some digging but HDP must not spend as much to acquire a customer hence needs less revenue to turn an operating profit on the company. HDP must also be more disciplined about which customers they are targeting so not wasting (again in the analyst section of clouderas call) their time on the really small customers. I’ll be very interested to see HDPs next year because if they aren’t lying then they have a dramatically better business than cloudera.

Long story short. Previous thesis, “huge TAM, huge tailwinds, huge growth ahead”. New thesis, “marginal growth, cash flow improvement in 2 years, no earning in sight, suspect management, PR department has been doing financial statements”

-Ethan

56 Likes
2

Ethan,

Altyrx had a similar question you pose at their earnings call, with 130% net expansion rate, how can they only be projecting 35% growth going forward for 2018! (I forget the specific growth, but was around that)

Management fumbled that question. However, with Altyrx it seemed like they were caught low balling their numbers.

We shall see if that is the case, I believe it is.

But almost all SaaS companies I follow have forecast reduced growth into the 30%-35% range for 2018. They may all be using the same source for making their projections for all we know.

But Cloudera, falling to 20% growth! Is that correct? That is an enormous sudden crash. And given how little penetration they have to give up on land and expand…yeah, that is a big issue.

Tinker

5 Likes
3

From Seeking Alpha, here is what an analyst said yesterday before earnings:

Needham raises its Cloudera (NYSE:CLDR) price target ahead of earnings from $22 to $26, a 22% upside to yesterday’s close.

Firm cites increasing confidence in Cloudera’s growth trajectory and margin leverage and expects a strong report aftermarket today.

Needham believes Cloudera continues to benefit from its unique solution that lets enterprises reduce model implementation time.

Firm maintains a Buy rating.

Source: Briefing.com.

Cloudera shares are up 1.8% premarket to $21.62 with a 52-week range of $14.50 to $23.35.

Firm cites increased confidence in Cloudera’s growth trajectory and margin leverage?

WOW

Jimbo

1 Like
4

CLDR and AYX formulas are listed here, including a brief definition.

Public Saas Company Disclosure Metrics for Retention and Renewal Rates
mentions several board followed companies; from June 2017, but what I found interesting were the varying formulas used–I would have thought there was a more limited, specific measurement but the definitions are most helpful.
https://www.key.com/kco/images/Public_SaaS_Company_Retention……

CLDR: Dollar-based retention including the benefit of upsells, based on subscription revenue
“Our quarterly net subscription revenue expansion rate equals the
subscription revenue in a given quarter from end user customers that had
subscription revenue in the same quarter of the prior year, divided by the
subscription revenue attributable to that same group of customers in that
prior quarter. Our net expansion rate equals the simple arithmetic average of
our quarterly net subscription revenue expansion rate for the four quarters
ending with the most recently completed fiscal quarter. We have excluded
Intel from our calculation of net expansion rate, as it is a related party.
(424B4 filed on 4/28/17, Page 44)

1 Like
5

RE: Tinker Altyrx had a similar question you pose at their earnings call, with 130% net expansion rate, how can they only be projecting 35% growth going forward for 2018! (I forget the specific growth, but was around that)

Management fumbled that question. However, with Altyrx it seemed like they were caught low balling their numbers.
Yeah, I read that too. HDP also had a similar exchange. I came away with the impression from both of those that management was low balling.

RE: preferredinvesto Public Saas Company Disclosure Metrics for Retention and Renewal Rates
mentions several board followed companies; from June 2017, but what I found interesting were the varying formulas used–I would have thought there was a more limited, specific measurement but the definitions are most helpful.
https://www.key.com/kco/images/Public_SaaS_Company_Retention……
super useful link, thank you! Notice they don’t mention that they only included their best cohort. Dishonest to the point of me wondering if they will get a slap on the wrist.

re: Jimbo. I think that was released before earning right?

6

Yes, before earnings.

To recap:

Yesterday “Firm (Needham) cites increasing confidence in Cloudera’s growth trajectory and margin leverage and expects a strong report aftermarket today.” Raises target from 22 to 26

And then today they said this:

Needham lowers its target from $26 to $22 and says the lower FY19 guidance relates to the customer type Cloudera has added.

Firm says the new customers aren’t likely to match historical net expansion rates and the guidance hints towards a dramatic slowdown.

Why would anyone listen to analysts.

3 Likes
7

But Cloudera, falling to 20% growth! similar to ANET. Is it lack of visibility or is the Cloud overbuilt? I have nagging memories of the fiber optic build out.

2 Likes
8

The cloud ain’t overbuilt. Cloudera does not build out the cloud they are a Hadoop data analytical company.

https://en.wikipedia.org/wiki/Cloudera

I have yet to see any let up in AI, in fact just the opposite, AI is only going to increase as processing power/cost continues to improve.

Why cloudera is slowing down, why they are changing their strategy to focus on existing clients instead of land and grab, dunno.

If Nutanix or AYX were to do this their stocks would deserve to crash 40%. These two companies are still out there trying to conquer the world and succeeding. For some reason Cloudera is pulling back from such and it would seem focusing now on more immediate profits.

I don’t follow it, but it is most likely it is company specific, but companies like Hortonworks…are the affected?

Tinker

9

Wow, NER (or net revenue retention, NRR) being cherry picked. That’s worrying. I wonder if there were any clues in previous earnings calls/financial statements that this was happening. It makes me wonder about the Net Promoter Score as well. The NPS is arrived at from questionnaires/surveys performed by the company itself and is thus subject to bias or just poor survey design.

Here’s from Alteryx’ last EC, just to follow on from Tinker’s comment regarding Alteryx:

Bhavan Suri – William Blair and Company – Analyst
One, if I look at your [Inaudible] expense, it’s phenomenal, right? 130% has been very strong. And now I look at the guide of 35% growth and I look at sort of what your customers expand [Inaudible] and I obviously understand the lands are small, but they’re getting a little bigger, the expansions happening a little faster. Help me unpack the revenue guide a little bit to understand if I’ve got sort of 30%, let’s assume even if it slows down, growth in my existing customer base expanding, how we should think about sort of what that means for net new customer acquisition and sort of just how you guys are thinking through that part of how you built up the guide? I’d love to understand it a little more. Thank you.

Dean Stoecker – Chairman and Chief Executive Officer

Bhavan, thanks for the question. I think we’ve talked on previous quarters about the lands being better-performing cohorts, that we expect our net retention numbers to be where they’re at. There’s no assurance that that will continue. While we haven’t seen a flattening of our S-curve, we continue to see this $29 billion market opening up.

We’re seeing some of the lands, like SP Innovations, happen at a larger scale. We’re seeing a bigger percentage of our customers expand and having those expands larger. The long-term sustainability of that, we don’t actually know at this point. What we do know is that our playbook, in both the land and the expand motions, are working around the globe.

So we’ll continue to double down on making sure that that continues.

Brent Bradford – KeyBanc Capital Markets – Analyst

Thanks for taking the question here, one for Dean and one for Kevin. Dean, let’s start with you. Revenue growth here accelerated to 55% from 52% last quarter. We don’t see that often in subscription models.

So what were some of the specific factors that drove the acceleration in the subscription revenue this quarter? Do you think there’s this broader market awareness of this problem that you’re addressing? Is it just improved sales, productivity, seasonality? Walk us through what were the specific factors that you think contributed to the upside in the quarter and the acceleration in the overall business.

Dean Stoecker – Chairman and Chief Executive Officer

There’s no one issue. There’s a number of issues. I think part of it is the exposure we’ve gotten from going public. It’s taken away some of the friction in the selling process.

The market clearly understands the value of higher-order analytic outcomes. And then we see some people moving on from first-generation visualization products to higher-order tools that do everything from spatial to predictive machine-learning. We’re seeing that. We’re also seeing the CDOs have an impact for larger organizations.

We did have a number of larger-scale lands, and whenever there is a CDO (Chief Data Officer) or a proxy for a CDO, we see expansions happening, landing a bit bigger and expanding a bit faster.

----------------------

I’m getting the same impression from these guys as with the Arista folk. They’re being blown away at how well they’re doing, all visible signs point to this phenomenal growth continuing for at least the immediate future, but they have no idea for how long it can be maintained, so they’re settling for a more ‘realistic’ target. Arista’s long-term target is 25%, Alteryx 2018 guidance is 35% (down from the 50s) It’s almost like they can’t believe their own eyes.

Now, a company boasting a phenomenal NPS and NRR looks like a fantastic prospect. But these are scores provided by the company to ourselves. Without committing fraud, these numbers can be shown to us in very different lights, so we need to be careful at taking them at face value.

Cloudera has shown how the NRR can be cherry picked.

The NPS can also be influenced. If the question was asked of customers at the start of the business relationship, it’s different from being asked a year into it, after they’ve used the product extensively. What % of customers respond to the survey? With Nutanix, are only truly excited customers responding, whilst the majority of mildly content customers not bothering to? Are at least 50% of customers replying to these surveys? In regards to Nutanix, they’ve received an award from Omega Northface, which has a leaderboard of businesses based on customer satisfaction and NPS. Sounds great, but they’re still getting the values from the companies themselves.
I’m by no means anywhere near an expert, merely an amateur investor. I’m happy to be corrected and educated in regards to how the NPS is obtained. But I’m seeing the NPS being thrown around more and more and don’t want people to mistakenly take it as gospel.

Summary: Although both NPS and NRR numbers are a fantastic starting point, they’re not totally reliable. As with all analysis, we take the entire picture, not just two numbers. The revenue numbers of Alteryx and Nutanix look great, and these are the numbers we can trust as much as is possible (requiring actual fraud in an audited USA-based company in order to be false).

I never followed Cloudera, but for those who have, can you recall any clues regarding the net expansion rate being for their top customers only?

17 Likes
10

Ben-

In reviewing the last quarter transcript for AYX, I also didn’t see any signs of the same issues as Cloudera.

One thing that gave me pause, was the continued talk of increasing sales and marketing spend with the hiring of additional team members.

I agree they are being careful with guidance like Arista, but if the growth does slow down from 50%+ to mid 30’s, with the increased spend in sales and marketing, that would be worrisome to me.

One of the reasons I like AYX is they have been able to grow 50%+ and improve their profitability at the same time. Now they are cash flow positive and it appears they are going to invest more into growth.

4 Likes
11

I’ve been following this thread and I’m sorry I haven’t had more to contribute thus far. Here are a couple things I would like to point out:

  1. I wouldn’t compare CLDR to any company other than HDP. They are both very weird in that they lose a lot of money, and thus trade at a lower PS, but are growing very fast. HDP’s losses are shrinking. I’m not sure about CLDR’s. I have never followed CLDR, and back when I looked at them I chose HDP instead.

I don’t know why CLDR would think it ok to say they plan to grow at a 20% pace. If true, this reaction is reasonable.

But this has NOTHING to do with Alteryx. That’s not a good comparison.

  1. HDP gives out a ton of SBC – $110M last year, but CLDR gave out $290M. Wow.

  2. The net expansion rate thing Ethan mentioned is a big deal and sounds scary as hell! I haven’t had a chance to look into what he’s talking about, but if indeed they gamed it, that’s a big problem. I would love to see a post further exploring that specifically, but I don’t think I’ll be the one to write it.

  3. Even after this huge drop, CLDR is not cheap. Because of the “very weird” business model I mentioned above, they trade at a lower PS than most companies we follow, but it is now about the same as HDP’s I believe. So don’t try to buy on the dip. In fact, it sounds like there are currently many more reasons to be skeptical of CLDR. Perhaps they are simply losing to HDP.

  4. I am still confident in HDP. I may add to my position on the dip they’ve taken (I assume in “sympathy” to CLDR’s huge drop).

Bear

6 Likes
12

Cloudera built their world around Hadoop. I’m not sure Hadoop is going to be all that significant in this rapidly evolving space for much longer.

Hadoop is being eclipsed. Companies like LinkedIn, Yelp and other high traffic information businesses are moving to more advanced data streaming technologies like Amazon Kinesis, Apache Kafka, Flume and Storm, delivering data direct to their stakeholders, without landing it into a data lake or Hadoop. I’ve personally been driving a similar project, making our own Cloudera Hadoop ecosystem obsolete.

This is just one man’s opinion. Do your own analysis and see how these companies are able to achieve scale and deliver data near real time, without ever landing in Hadoop.

-----
Invest wisely my friends

6 Likes
13

Cloudera built their world around Hadoop. I’m not sure Hadoop is going to be all that significant in this rapidly evolving space for much longer. Hadoop is being eclipsed.

Well, yes, but pieces of Hadoop will remain significant - for instance HDFS (Hadoop Distributed File System) (and yes I’m aware that private MapR doesn’t actually support that either). I think it’s more accurate tot say that Hadoop is changing: for instance, I don’t think anyone today is implementing new Map Reduce based applications. That said, I think the Hadoop distribution companies are offering things extending well past traditional Hadoop technologies.

This is a confusing space technology-wise. I don’t think anyone here has real insight into who has the best offering, technology-wise. I also don’t think anyone here knows if the salesforce at Hortonworks is better than that at Cloudera. The business models are different, with TMF copping out on actually describing the differences and simply referenced HDP’s deals with IBM and MS (https://www.fool.com/investing/2017/07/09/better-buy-hortonw… ).

Let me try: Cloudera is the older of the two companies. It has a suite of software that work on top of Hadoop. While Hadoop is open source, Cloudera’s software suite isn’t and that’s how they make money. Hortonworks’ business model is different - it distributes the full open-source Hadoop stack and makes most of its money from training and services. Here’s a comparison article: https://www.springpeople.com/blog/hortonworks-or-cloudera-wh… Then again, Cloudera

One data scientist I recently spoke with thinks that Hortonworks is selling their stuff to people who won’t really benefit from it, or would benefit more from other solutions. Here’s an article analyzing what just happened at Cloudera: https://digitizingpolaris.com/cloudera-fails-investors-but-s… : For background, Cloudera sells via a “land and expand” strategy; first it aims to get its foot in customers’ doors and then, when the customer finds value in its product, it looks to sell more. Or as Cloudera CEO Tom Reilly put it, “Our lands are an investment, our opportunities are in expand.” In other words, it is money-losing proposition to land the wrong kind of customer?—?the kind who doesn’t buy more. That’s what happened in Q4.

Cloudera’s cure for this problem is to: A) Hire a new head of sales and a “logo hunter” B) Stop rewarding salespeople who sell outside of its target market of Global 8000 enterprises and C) Bring on “expansion reps.” While that strategy might work, the analysts on the call weren’t convinced. JP Morgan, Deutsche Bank, Needham, Mizuho and others all downgraded the stock the next day.

That author’s view was different back in September, that Cloudera is pivoting its business away from Hadoop and towards AI/Machine Learning: https://digitizingpolaris.com/cloudera-buys-ai-company-disse… and followed that up in Dec: https://digitizingpolaris.com/reality-check-is-cloudera-real… (with mentions of Saul favorite, AYX, btw).

Is picking winners in the Big Data space space obvious to you? If you don’t understand the technicals well enough to choose one for without consulting your DBA, then do you at least understand the differences in business models or at least what will make your pick successful? If you do, then please tell me how you’re choosing which of these to invest in, because I’d really like to hear why.

7 Likes
14

Good feedback smorgasbord!

From what I’ve been observing, it feels like the market and attitudes are shifting. The batch ETL process that feeds hadoop is in need of innovation. What most companies do is extract data from transactional stores, transform it into a format suitable for analytics, and then load it into secondary stores such as a data warehouse - typically on hadoop. But these warehouses are huge and cumbersome, so more ETLs are implemented to move subsets of topical information from the big data lakes into tertiary stores or data marts for individual business stakeholders. It is easy to see why this is inefficient and costly. While HDFS solves some big data analytics challenges, the cost of ingesting data, and the timeliness of putting actionable intelligence in front of business stakeholders are big drawbacks. In addition, hadoop is not ideal for many interactive business workloads. SQL on Hadoop variants such as Hive and Impala haven’t been able to address this gap.

The main threat to hadoop, as I see it, is a lot of R&D is going into bypassing the redundant step of landing data into a big data lake. Instead pioneers like LinkedIn and Yelp are investing in streaming smaller sets of timely intelligence directly to business stakeholders in near real time. IMPO we’re not far from some big commercial breakthroughs in this regard. Hadoop may still be in their landscape, but with shrinking strategic importance to the business.

I don’t know who the winners and losers are going to be, there is a lot of innovation in this space, and it is evolving very rapidly. Apache, Cloudera, Spark, Databricks and others are investing in this research. Tech titans like Google, AWS, and others are working furiously on these problems, and they will eventually overcome them. Until then, I’m sure there will be lots of fits and starts along the way.

-----
Invest wisely my friends

1 Like
15

The main threat to hadoop, as I see it, is a lot of R&D is going into bypassing the redundant step of landing data into a big data lake. Instead pioneers like LinkedIn and Yelp are investing in streaming smaller sets of timely intelligence directly to business stakeholders in near real time.

Well, I think that as IoT grows, more data will be coming in from small connected devices. That has to go somewhere right off the bat since the device itself will have very little processing power. And then the data will usually get processed before being analyzable. With millions/billions/gazillons of devices there’s a data quantity problem. At some point, companies will decide to throw away data rather than pay to keep it. But first, of course, they’ll want to extract from it what they think they might ever need. To throw an invented scenario out there in an area in which I have no expertise, think about pedometers. The data they gather would include the time and gps location of each step, your speed, the temperature - and your temperature, your heart rate, etc. For each step. That’s probably too much data to keep when you have millions of subscribers taking tens of thousands of steps hundreds of days a year. So, you’ll want to summarize to something like per hour or per day rather than per step. And you may decide to throw up something like body temp, or just record trends or variants.

And while we’re focused here on smaller companies, don’t forget the big gorilla, Oracle, not only had relational offerings, but also NoSQL and Hadoop-based offerings of its own. Heck, even Bert thinks that Oracle’s 18c “autonomous” database may turn out to be big winner (see https://seekingalpha.com/article/4159958-oracle-art-making-a… ).

2 Likes
16

At some point, companies will decide to throw away data rather than pay to keep it

Yeh - like my bank can’t even keep my frikking bank account statements available beyond 6 months and that ain’t exactly gzillions of terabytes of big data.

A

2 Likes