Snowflake’s Slowing Sales

GauchoRico summed it up very well…“SNOW is just getting started. Phase 1 of the journey for a customer is to get the data digitized on SNOW. Some customers are far along on that while many others are just getting started while others are not yet SNOW customers. It’s my view that SNOW’s best days are still ahead”

I had posted these sections just a day ago and ALSO year ago and am posting it once again… to emphasize why Snowflake is one of my top three investments along with Datadog and Zscaler.

“No one in their senses will pull out Datadog’s monitoring out of their development and IT environments or Zscaler’s network security out of their enterprise. Neither will they move data out of Snowflake to on-premises ( but a lot of data will move the other way)…

Explosion of Data: “How much is a “Zettabyte”?. Well its 1,000,000,000,000,000,000,000 bytes! According to IDC, the world’s digitized data will hit 175 Zettabytes in 2025! It’s probably around 50 Zettabytes today and was around 2.7 Zettabytes in 2012. So, you get the picture of the explosion of data. As mentioned above, IoT devices, biological sciences and many others are already contributing and will be producing even more data.”

Here are the two big problems with Big data ( without any technical jargon).

1. What’s the use of all that data in your organization?

Well, if you cannot get any meaningful insight from whatever data you have, it’s of no use; meaningless data is garbage. Ask any engineer/dev who works on data and they will tell you that getting those insights isn’t easy. Even if all that data is in your own organization, different teams in an organization can have their own data silos with different technologies and access rights.

2. Now what if all that data you need is spread across different organizations?

This makes it even more difficult to access that data and possibly join that data to get meaningful insights. That’s exactly the problem Snowflake ( SNOW) is solving with it’s Data Sharing! The applications that are being built leveraging on that shared data on Snowflake is a totally new category! That’s how TAM is increased! I think that many businesses focussing on ads and promotions are are going to rely a lot on Snowflake soon ( if not already). It’s no less than a network effect for data sharing when so many of the Fortune 500 businesses are already on their platform.

It’s worth watching the magic Big data is unfolding before our eye’s; the center stage for now is being held by Snowflake ( SNOW ) and MongoDB (MDB)!

Just my 2 cents.

Cheers!
ronjonb

p.s. Recently migrated some of my databases to MongoDB :slight_smile:

21 Likes

The NRR is not affected by new customers entering after the 24-months-ago starting point.

Thanks to Softie for the correction here, and WSM had also said Snowflake [with NRR] is telling us that their customers who consumed product from them in the first month of Q4 2020… So I think I bungled the explanation, but Softie made my point (we agree): Later, as the business is mostly established customers, the overall NRR will decrease as more of the customers are further in their journey. So basically, NRR won’t be a great leading indicator. The more I think about it though, so what? Ultimately we don’t care about NRR – we care about revenue, so let’s look at that directly:

Period 1 Fiscal Q4 2019 - Q3 2020: revenue was something like 200 - 225m (my guess), generated by a few hundred older customers and several hundred who were just starting to ramp up

Period 2 Fiscal Q4 2020 - Q3 2021: revenue was 490m, generated by:

  1. the 702 customers who had been around for 1 year or more, and
  2. the 1232 customers who were just starting to ramp up

Period 3 Fiscal Q4 2021 - Q3 2022: revenue was 1.027b, generated by:

  1. the 702 customers who had been around for 2 years or more,
  2. the 1232 customers who had been around for 1 year or more, and
  3. the 1620 customers who were just starting to ramp up

Period 4 So over the next 4 quarters, the revenue will be generated by:

  1. the 702 customers who had been around for 3 years or more,
  2. the 1232 customers who had been around for 2 years or more,
  3. the 1620 customers who had been around for 1 year or more, and
  4. the 1862 customers who are just starting to ramp up

In the last four quarters, period 3, we’ve seen a huge revenue increase vs period 2. But of course we have! The 1232 customers added during period 2 are finally getting to a much more normal level of spending. And period 3 is the spending from these, plus the 702 customers who have been around longer. When you compare that to period 2, which was the spending primarily from only those 702 customers, it’s obvious that revenue in period 3 was going to dwarf period 2.

However, in period 4 there are already 1934 customers 2 years older or more. The 1620 that are getting up to a more normal level of spending will make a big difference, but it won’t be as massive as in period 3, because:

1620 / 1934 = 84%
whereas
1232 / 702 = 175%

Maybe the 1620 are larger. That will help. But I still foresee a rapid deceleration. Wouldn’t surprise me if SNOW revenue grew less than 80% in the next 12 months.

Bear

19 Likes

In the last four quarters, period 3, we’ve seen a huge revenue increase vs period 2. But of course we have! The 1232 customers added during period 2 are finally getting to a much more normal level of spending. And period 3 is the spending from these, plus the 702 customers who have been around longer. When you compare that to period 2, which was the spending primarily from only those 702 customers, it’s obvious that revenue in period 3 was going to dwarf period 2.

However, in period 4 there are already 1934 customers 2 years older or more. The 1620 that are getting up to a more normal level of spending will make a big difference, but it won’t be as massive as in period 3, because:

1620 / 1934 = 84%
whereas
1232 / 702 = 175%

Maybe the 1620 are larger. That will help. But I still foresee a rapid deceleration. Wouldn’t surprise me if SNOW revenue grew less than 80% in the next 12 months.

We can’t (shouldn’t) assume that each customer’s journey is going to be the same. I think SNOW has a highly heterogeneous install base of customers. We also can’t (shouldn’t) assume that customers who have been customers for 2-3 years will slow their rate of spending increases going forward. We know one customer just signed a 3-year $100M spending commitment. Was this customer amount the 702 cohort? We also know that 40% of SNOW’s top 10 customers grew greater than SNOW’s 110% Q3 revenue growth rate. Were these 4 customers in the 702 cohort? The thing is that if customers who have been around for 2 or 3 years can still grow their spending by >100% per year, then the hyper growth can be sustained going forward. This seems to be happening as evidenced by the revenue growth numbers we have seen and by what management is telling us on the earnings calls.

So this begs the question, once a customer has reached a more “normal level of spending” for how much longer will the growth in spending continue to be huge? If it’s 5 years then looking at customer additions doesn’t foretell an eminent slowdown (as of today).

GauchRico

38 Likes

Hi WSM.

Thanks for the analysis and thanks to those who have critiqued it.

One point I would add is that I don’t think the customers per 100 sales force added is useful as you’ve done it. What I’ve heard from management teams generally – although not Snowflake’s in particular – is that new sales people generally take two quarters to ramp (occasionally I hear three, but let’s stick with two). So I think you want to use lagging data for this. For example, instead, I believe it would be better to use the sales force in place as of two quarters ago to compare against the current quarter customer adds. That probably paints a clearer picture. I think you’re on the right track by looking at that – we definitely want to measure sales force effectiveness when the company is accelerating adds to the sales team. But putting new hires immediately into the denominator is likely to paint a bleaker picture than the underlying reality.

Fool on!
Thanks again for the analysis, and best wishes,
TMFDatabaseBob (long SNOW)
Advanced Research Fool (formerly called “Maintenance Coverage Fool”)
See what a “Coverage Fool” does here: http://www.fool.com/community/community-team.aspx
See my holdings here: http://my.fool.com/profile/TMFDatabasebob/info.aspx
Peace on Earth

Please note: I am not a member of any newsletter team. My opinions are my own and do not necessarily reflect those of the TMF advisers. I am not an investment professional, merely an investor.

21 Likes

new sales people generally take two quarters to ramp (occasionally I hear three, but let’s stick with two). So I think you want to use lagging data for this. For example, instead, I believe it would be better to use the sales force in place as of two quarters ago to compare against the current quarter customer adds. That probably paints a clearer picture.

TMFDBBob - good point; you’re right. Here it is if we use two quarters ago employees (i.e. assuming that the S&M employees added in the last 2 quarters produce zero incremental sales. I would hope they contribute at least a little in that period, but let’s be generous here. I think the answer lies somewhere in between tbh.

Still, this is what I get then:

Customers added per 100 last Q-2 S&M employees:


#	Q1	Q2	Q3	Q4
2021	36.2	40.1	40.4	51.3
2022	33.4	36.4	29.4	

YoY	Q1	Q2	Q3	Q4
2022	**-8%	-9%	-27%**

F500 Customers added per 500 last Q-2 S&M employees:


#Q1	Q2	Q3	Q4
2021	9.4	7.6	6.0	9.2
2022	1.7	7.6	2.8	

YoY	Q1	Q2	Q3	Q4
2022	**-82%	0%	-54%**	

So a very steep deterioration in efficiency even when calculated like that. That is on top of the absolute decline in total customers added as well as F500 customers added - both qoq and yoy for both metrics in the last reported quarter!

I think the point has been made in as many ways as practically possible; I specially liked Bear’s revenue calcs above, but I’ll nevertheless try one last time.

Snowflake has got a fantastic EXPAND model going. And most everyone with a differing view to what I tried to sketch above keep on articulating different reasons why this will continue to be strong or even very strong.

However their LAND motion, the key first step of a land and expand model for a successful software company has been stuttering of late and has downright taken a dive in the last reported quarter. This will surely flow through into their numbers, irrespective of how impressive their expand drivers are, but will take more time with Snowflake vs our other companies because of the long ramp-up time of their customers.

I want to be out long before that happens and believe that the writing may already be on the wall for those willing to read it.

-WSM

13 Likes

WSM,

https://investors.snowflake.com/events-and-presentations/eve…

Have you watched the Barclays webcast with Snowflake’s CFO? I’m pulling the paraphrased notes I took from it when I watched it last month and apologize I may be butchering the webcast here but this is what I got from it (please watch it if you haven’t!):

“We have seen product revenue growth rate accelerate due to diverse broad based consumption on platform…NRR is fueled by the largest most mature customers…high NRR is very durable and will remain best in class…”
“Takes 9 months to see customers begin to consume at the contract rate…”
“Average customer consumption is increasing as they continue to move workloads”
“Many customers’ growth is dependent on consumption of SNOW, like lacework or coinbase”

“Never had a data science team in a finance department before. My data science team builds models using Snowflake to project customer by customer revenue on a daily basis, based upon the prior day’s customer consumption. We dive deep into the top 50 customers with our sales team for a sanity check on these projections”

“Our sales reps are increasingly motivated for consumption of existing customers. Some reps have quotas on growth (acquiring new accounts) vs consumption (from existing accounts). Some are 90% growth, 10% consumption. Some are 10% consumption, 90% growth. We rebalance sometimes so reps might keep an account for one year or longer…we want reps that understand each customers’ business, in order to translate values to that business. We are focused on the vertical clouds to generate new use cases that our reps are focused on”

"We have over 300 customers built directly upon Snowflake. But the vast majority are on-prem, and they will all eventually find new use cases, which is why our NRR is so strong and durable. Only thirty of our customers have completely shutdown Teradata. The vast majority will take many years to migrate all of their workloads. That’s why you will see us have very high NRR that will stay best in class."

"The founders always had data sharing in mind and it is part of the architecture from day one…we are seeing investment companies, hedge funds, financial service customers are asking their vendors to be on snowflake to facilitate data sharing. For example our media cloud data cleanroom (NBC, disney) - video streaming customer and retail customer can share data to help target specific advertisements and monetize their data. Data sharing is very much early innings and not yet a driver of revenue, but it is driving customer lands as they all very much want to data share.

“More and more companies want to replatform onto Snowflake, like Blackrock, or to build new offerings on Snow. We see lots of small tech companies want to replatform onto Snow. I deal with these companies on a daily basis. It takes time for the small companies to be meaningful, but it all opens the opportunities for data sharing.”

“We don’t want sales productivity to go too high too quickly, because it means you are missing out on opportunities… We see successful enterprise sales reps that come out of onboarding into our inside BDR sales team program for a year or two, then go into inside sales for a year or two, and then they become enterprise sales rep. Some of the strongest enterprise sales reps come out this way. We have good muscle for onboarding new people. The reps dealing with our major accounts are those with twenty plus year experience.”

What I’m seeing:

1.) Most customers are still on-prem, and these will take years to migrate more and more workloads onto SNOW. The NRR is also said to be driven by large, mature customers. Combine these two, and you could have product revenue growth sustained for an incredibly long time, possibly for years, from existing customers alone. The CFO sees their NRR remaining best in class for a reason.

2.) There are an increasing number of customers landed where their business growth is directly tied to (dependent upon) the use of SNOW. Think about AMPLITUDE or BRAZE, which are hypergrowth companies talked about on this board. https://www.braze.com/resources/videos/the-power-of-braze-an…

3.) Data sharing looks like it will be a huge driver of future product revenue through increasing use cases by existing customers, and also driving future customer lands

4.) Using sales headcount to measure incremental customer lands is not going to be very useful.
First, some sales reps don’t even go into enterprise sales until 1 to 4 years out from onboarding.
Second, you have no idea what mix of sales reps are directed toward something like 90% consumption quota versus 10% growth quotas.
While number of customers added are something to pay attention to each quarter, I won’t sell out of SNOW unless I see actual product revenue growth decelerate too quickly.

5.) And as many other posters have already pointed out, the customer mix is very heterogeneous. One large customer land could be driving 8 or 9 figure contracted minimum spend deals.

6.) Others have also pointed out the comparison to hyperscalers. SNOW could very well be like one of them - the market has seen how AWS can grow at ridiculous rates for years (and years more to come), and I believe the market will value SNOW accordingly. What I mean is, when SNOW does decelerate from triple digit growth rates, its high growth durability is likely to remain intact, and that might be much, much more highly prized than any of us can anticipate.

Data growth is exponential, use cases growing constantly. SNOW is consumption based. This is vastly different from other hypergrowth companies like CRWD. CRWD’s growth is heavily dependent on customer lands - there’s only so many modules they can compel their customers to add in the long-term. So when CRWD has persistent deceleration, the market will not assign large forward multiples anymore - it believes high growth can’t be durable for years to come.

110 Likes

Awesome post JonWayne! It explains so clearly why Snowflake’s NRR will remain huge for years and years. Greatly appreciate your help to my understanding.
Saul

16 Likes

Thanks for posting this John. A question that I’ve been pondering for a while now has just been answered. The question was how does Teradata play with Snowflake? I recently took a quick look at Teradata (TDC) to see if it might be an investment candidate. It’s not.

For those who don’t know, Teradata is a hardware/software company that sells software optimized for SQL queries of a normalized database. The primary use case for Teradata is business intelligence analytics. It is generally not employed as a transactional database, but there is no particular reason that it couldn’t be (or so it was told to me by a Teradata sales representative several many years ago). The software will run on most any box, but it runs best on a Teradata box.

“Normalized” refers to a relational database design that eliminates redundancy. Theoretically, each data item is stored only once irrespective of how and where it might be used. As the name implies, Teradata is designed to accommodate very large data volumes. As business analytics very frequently involves cross-functional data (for example, engineering, manufacturing, procurement, finance and marketing data may all come into play for certain analytical routines), the transactional data from all these functional areas may populate a Teradata database. Teradata sells their products primarily to large enterprise customers. So far as I know, Teradata is strictly on-prem, but of course that may have changed since I retired.

From this post it appears that Snowflake is progressively, albeit slowly, eroding Teradata’s customer base. As noted by the Snowflake’s CFO, most of there customers are still on-prem, presumably using Teradata for their business intelligence analytics.

2 Likes

Just a few days ago in this thread I wrote …“I think that many businesses focussing on ads and promotions are are going to rely a lot on Snowflake soon ( if not already).”

Well incase you haven’t noticed here’s some positive development in that direction…

Privacy-Safe Data Clean Room on AWS with Snowflake

"The deprecation of device identifiers and third-party cookies has accelerated the need for advertisers, data providers, and media companies to collaborate directly as they seek to understand customers, plan media campaigns, and analyze path-to-purchase. At the same time, privacy regulations, consumer trust, and intellectual property protection are driving the need for increased data security and governance when data is shared between parties. These two industry trends – increased data security and multi-party data collaboration – are driving customers to invest specialized workloads called data clean rooms. Clean rooms act as collaborative workspaces where multiple parties can run analytics together with restrictions that ensure data security and governance for underlying raw data."

Read more in this blog here if you’re interested:

https://aws.amazon.com/blogs/industries/deploying-a-privacy-…

Cheers!
ronjonb

19 Likes