Hacker News new | past | comments | ask | show | jobs | submit login
Datadog S-1 (sec.gov)
437 points by jonknee 59 days ago | hide | past | web | favorite | 96 comments



These are the two juiciest paragraphs for SaaS geeks:

> We have a highly efficient go-to-market model, which consists of a self-service tier, a high velocity inside sales team, and an enterprise sales force. As of June 30, 2019, we had approximately 8,800 customers, increasing from approximately 7,700, 5,400 and 3,800 customers as of December 31, 2018, 2017 and 2016, respectively. Approximately 590 of our customers as of June 30, 2019 had annual run-rate revenue, or ARR, of $100,000 or more, increasing from approximately 450, 240 and 130 customers as of December 31, 2018, 2017 and 2016, respectively, accounting for approximately 72%, 68%, 60% and 48% of our ARR, respectively. Further, as of June 30, 2019, we had 42 customers with ARR of $1.0 million or more, up from 29, 12 and two customers as of December 31, 2018, 2017 and 2016, respectively. As of June 30, 2019, our 10 largest customers represented approximately 14% of our ARR and no single customer represented more than 5% of our ARR.

... and ...

> Our business has experienced rapid growth and is capital efficient. Since inception, we have raised $92.0 million of capital, net of share repurchases, and we had $63.6 million of cash, cash equivalents and restricted cash as of June 30, 2019. We generated revenue of $100.8 million and $198.1 million in 2017 and 2018, respectively, representing year-over-year growth of 97%. Our revenue was $85.4 million in the six months ended June 30, 2018 compared to $153.3 million in the six months ended June 30, 2019, representing period-over-period growth of 79%. Substantially all of our revenue is subscription software sales. Our net (loss) income was $(2.6) million, $(10.8) million, $0.5 million and $(13.4) million for the years ended December 31, 2017 and 2018 and the six months ended June 30, 2018 and 2019, respectively. We generated operating cash flow of $13.8 million, $10.8 million, $10.6 million and $3.0 million in 2017 and 2018 and the six months ended June 30, 2018 and 2019, respectively. Our free cash flow was $6.0 million, $(5.0) million, $1.5 million and $(6.4) million in 2017 and 2018 and the six months ended June 30, 2018 and 2019, respectively.

Very impressive SaaS business, especially in such a competitive space. And I'm a happy customer for several years. This chart of their quarterly revenue run rate going back to 2016 is truly amazing and will make every SaaS founder green with envy:

https://www.sec.gov/Archives/edgar/data/1561550/000119312519...


So nice to see an S-1 from a SaaS company with actually solid financials and clear explanations of where the money is going. This is what IPO announcements should look like instead of being packed with hype, hopes, and dreams.


Agreed! For all the hype and noise out there, the combination of SaaS and business value has created some phenomenal companies, both in terms of growth, and in terms of unit economics. And, as a user, I can say Datadog solves a real (and burning) problem for product engineering teams, and, though expensive, is worth every penny.


I've had a different experience with them - difficulty with setup and tracking ability, including a much more resource intensive approach than newrelic for collecting execution traces. While trying to resolve this emails went unanswered until one of my co-workers reached out in the public spotlight of twitter.



That is really impressive revenue growth for a SaaS company that is roughly breaking even. Most SaaS companies growing that fast are operating at a substantial loss.


I'm pretty green but I love how they break things out comma-separated like "Z, Y, X from years C, B, A, respectively"

Feels slightly awkward as you have to make the associations in your head(Z to C, Y to B, etc.) but actually gives a way better sense of progress.


> As of June 30, 2019, our 10 largest customers represented approximately 14% of our ARR and no single customer represented more than 5% of our ARR.

This is really awesome. One single customer isn't calling the shots so they can build a great product how they want.


As a long time customer, I can attest that they thoroughly deserve their success. Their product has always been amazing. Absolutely reliable. Beautifully well designed. We never once flinched at our bill.


Efficient indeed - 1,212 employees equates to around $250,000 per employee per year based on their first half 2019 revenue, which I understand is great for SaaS.


Is it for a tech company? Sales & Engineering are not exactly cheap.


It's pretty close to Slack's revenue per employee (1,502 as of Jan 31, on $400 million estimated 2019 revenues).


Revenue, not cost :)


Wow, this is such a crowded market that I'm really surprised just how much money they are making!



I'd love to know their churn.


From Page 59

> Furthermore, our dollar-based gross retention rate, based on a cohort of all of our customers, has been in the low-to-mid 90% range as of the end of each of our last eight quarters.

> A further indication of the propensity of our customer relationships to expand over time is our dollar-based net retention rate, which compares our ARR from the same set of customers in one period, relative to the year-ago period. As of June 30, 2018 and 2019, our dollar-based net retention rate was 146%, and as of December 31, 2017 and 2018, it was 141% and 151%, respectively. We calculate dollar-based net retention rate as of a period end by starting with the ARR from the cohort of all customers as of 12 months prior to such period-end, or the Prior Period ARR.


Their financials on the surface and the growth look really good. Congrats to the whole DataDog team on this! I used it in my previous job and loved it. I love how quick they were to add support for new things and how easy it was to configure everything.

Since they are a SaaS startup which deals a lot with data and compute, I wanted to see their infra cost. I was looking through the consolidated financials[1] and I didn't find something that referred to infra explicitly. I am guessing "Research and Development" operating expense will cover employee compensation(stock and salary). So, would the infra costs be the one in "General and Administrative"?

[1] https://www.sec.gov/Archives/edgar/data/1561550/000119312519...


In this case, Infrastructure Cost is part of Cost of Revenue (similar to COGS or Cost of Goods Sold).

Cost of Revenue is defined on Page 61:

> Cost of revenue primarily consists of expenses related to providing our products to customers, including payments to our third-party cloud infrastructure providers for hosting our software, personnel-related expenses for operations and global support, including salaries, benefits, bonuses and stock-based compensation, payment processing fees, information technology, depreciation and amortization related to the amortization of acquired intangibles and internal-use software and other overhead costs such as allocated facilities.

> We intend to continue to invest additional resources in our platform infrastructure and our customer support and success organizations to expand the capability of our platform and ensure that our customers are realizing the full benefit of our platform and products. The level, timing and relative investment in our infrastructure could affect our cost of revenue in the future.

Cost of Revenue and infrastructure costs are discussed on Pages 64 and 65:

> Cost of revenue increased by $21.3 million, or 115%, for the six months ended June 30, 2019 compared to the six months ended June 30, 2018. This increase was primarily due to an increase of $17.9 million in third-party cloud infrastructure hosting and software costs, $1.0 million of depreciation and amortization, $1.6 million in personnel expenses as a result of increased headcount, $0.4 million of credit card processing fees and other fees, and $0.4 million in allocated overhead costs as a result of an increase in overall costs necessary to support the growth of the business and related infrastructure.

Tip: Search the S-1 for "infrastructure" or "cost of revenue" to find other mentions of this.


Hey, thanks for that. I started reading through it and it seems like they are largely on AWS and have a $225 million 3 year commitment with them.

> The table does not reflect the enterprise agreement and addendum for cloud hosting that we entered into with AWS in April 2019, or the AWS Agreement. Under the AWS Agreement, we are required to purchase an aggregate of at least $225.0 million of cloud services from AWS through April 2022. Except in limited circumstances, such as our termination of the AWS Agreement for cause, if we fail to meet the minimum purchase commitment during any year, we are required to pay the difference. Neither party may terminate the AWS Agreement for convenience during this three-year term. In addition to AWS, we operate on other cloud hosting providers.


That's an insane commit unless they have MFN prices with zero non-resale clauses.

Oh, my bad, I keep forgetting that the standard rules of "How not to go out of business by saying 'No'" do not apply to HN startups.


Infra costs would roll up into COGS since they're variable costs. So the line item for Cost of Revenue captures that I believe. G&A is for fixed costs.


Just to clarify, infra costs are under COGS because infra costs can be directly attributed to production of what is being sold, not because they're variable. You can have fixed costs in COGS as long as they're directly attributable to production.

G&A can also contain both fixed and variable costs. G&A is for operating expenses that can't be directly attributed to production.


Ah. Thanks, I somehow missed that.


Love their product, hate their shady billing.

I'm in the middle of a bad experience with Datadog's billing/sales team. My monthly Custom Metrics usage wasn't being billed for because they said they didn't have a way to track it. Naturally my usage gradually increased because without the monthly bill increasing. When renewing a while back they went over my usage on phone and quoted prices, same usage profile as this month. Now this month my bill increased almost 400% and they said "oh we just soft-launched billing for this feature without notifying you". No heads-up when communicating with them that I wasn't actually being billed for my usage.

Still trying to get a refund for this curveball bill. Their sales team keeps forever escalating to management, leadership, and they say it has to be approved by a C-level. It's been over 20 days and they say it might take 2-3 more weeks before C-level makes a decision on credit or not. I've never heard of this many layers involved in something like this before with other companies.


Same. At several companies I've been at we've used or investigated Datadog. One place accidentally spun up a bunch of EMR nodes on AWS for a single run (less than 1 hour), to which we got charged their going rate ($15/mo?) for every instance - even though they were only up for an hour. DD refused to reverse this, though eventually did help up make sure those nodes got excluded going forward ... gee thanks. $15/hr/node for monitoring is too rich for my blood.

Most recently it was looked at to help monitor a small Kubernetes test cluster. 3 nodes. Now the base rate of $18/mo is just fine... except now they charge $1/mo/container past 10 containers per host. Because K8s (depending on how you install it) runs a bunch of little containers handling various back end things, you might not deploy anything to the cluster and still be WAY over that 10 container limit. In our case it came out to like $200/mo to monitor 3 nodes - that were no where near fully loaded.

They've got a great product but their billing just has not ever really made sense. While I haven't used them, Wavefront makes a lot more sense - pay per metric. Got a bunch of containers that don't need monitoring, then don't send metrics (or send them infrequently). Easy.


+1 for Wavefront. I've been a user since their pre-launch beta and have nothing but good things to say about the product and the team.

Their billing is based on rate not count, which is a better model IMHO.


All that you’ve said resonates with my experience, but:

“Because K8s (depending on how you install it) runs a bunch of little containers handling various back end things, you might not deploy anything to the cluster and still be WAY over that 10 container limit. ”

The way we manage that is to explicitly filter for containers. Don’t monitor all containers by default, make it opt-in and you’ll be fine.


Same boat here. We ended up going with an ELK stack instead. Muchm much, more control and worked fine for us.


Their billing is HORRIBLE. Its actually a big reason why we are investigating moving off.

Its insanely unclear. I get that their product domain is complex, and thus its hard to bill for, but... its things like, if you're container-focused, you still pay for nodes, and you get X number of containers per node that you're allowed to track. You wire up an AWS integration and it (1) counts as a node? I've heard different things from different salespeople about this, and (2) drives up your AWS bill like crazy because of the amount of data it pulls (our AWS bill is low-five-figures, and the Datadog integration represents 3-5% of it). The datadog-agent inside our kube clusters was constantly one of our largest logging services, until we started filtering those out. We had one of our services go crazy on logs one month, driving up our bill by 1500% until someone caught it (we're small. we make mistakes. we're learning). You can't delete the logs, they wouldn't, so we had to eat a 5 figure bill when we're used to, like $1000/month.

And its expensive. Maybe its comparable to something like New Relic, I don't know, but relative to StackDriver or CloudWatch? Its in a totally different league. Just on logging, its hard to do apples-to-apples because DD bills on lines whereas CloudWatch bills on bytes, but for our logging patterns, even with our annual enterprise discount on DD (and no discounts on AWS) DataDog is about 5x more expensive. CloudWatch (even with the new Insights product) is pretty hard to ask front-line developers to use, but StackDriver is pretty nice, and its also pretty cheap.

We joke about Datadog among our ops team. When we were investigating adopting them, I signed up for an account with my company email. Within a week, members of their sales teams were emailing our frontend developers, who I imagine they found by linking my company email to a LinkedIn company, about setting up a call. I'd get about one unprompted call a week from someone from their sales org.

Their product is great. Possibly even special, though I don't believe its miles better than StackDriver. The company is a nightmare. I tell everyone I can to stay away, at least until they sort out their sales and technical support organizations, if they ever do.


“And its expensive. Maybe its comparable to something like New Relic, I don't know, but relative to StackDriver or CloudWatch? Its in a totally different league. ”

To be fair to Datadog, their UI/UX is also in a totally different league. Stackdriver Logging is a total PITA to use when trying to investigate and correlate historical logs.


>We joke about Datadog among our ops team. When we were investigating adopting them, I signed up for an account with my company email.

The exact same thing happened to me an employer or two ago. Their sales droids were so aggressive with the calls (one time, many of our people were on vacation, one guy answered, told them no thanks, and then that same caller proceeded to ring nearly every phone in the cube farm sequentially) that, to this day, that team refers to getting bothered by pushy vendors as "getting datadogged".


> The datadog-agent inside our kube clusters was constantly one of our largest logging services

This. Just this. Until I recognized this, these agent logs were making up more than 50% of our total log volume. These were, for the most part, a single log message that an APM was malformed.

I only noticed this when renewing my companies Datadog contract, and couldn't fathom how we had been generating so many logs.

I was pretty disappointed when their sales rep had been pushing us to increase our committed log volume, without mentioning that literally thousands of dollars per month was being spent for Datadog to send themselves their own logs. As far as I'm concerned these logs should either be filtered out of the index by default, or not count towards billing.


Had a similar bad experience. We had a situation where we were running instances with short durations; AWS bills by the second, but Datadog bills hourly. We ended up having a Datadog bill that was higher than our AWS bill!

They had the opportunity to help us out - they ended up using it as leverage in their sales process, only willing to work with us IF we signed up for additional services. (I think a better word fits, but I won’t go there)

We tried to work with them, but eventually just ate the expense.


Yea their pricing model charging per host is dangerous with many small machines. It will become more of an issue as more companies move to distributed micro-services.

How Datadog handled your situation is what happens when sales tries to take on customer support, as datadog is doing. Incentives are misaligned. Similarities to Yelp.


The fact that a billing issue like this gets escalated to C-level for some determination of account credit is, on its face, absurd. It suggests that billing practices at the company are so off the wall that they have a weeks-long backlog of billing disputes that can only be handled by a C-level, indicating a vacuum where policy and process might exist.

Either that, or they’re jerking you around. Both are pretty bad.


It probably means there are many instances of this particular scenario happening, so they need an exec to make some blanket decision for all of them.


Even the basic billing concepts are too confusing. You pay for some Host count. You open the billing and usage console, and you see about 6 graphs for Hosts under two tabs. What's an APM Host, an Infra Host, and Agent Host? Which are double-counted, and which are not? Answers can be found, of course, but it could be a lot more clear.

Also, as others have noted, they turn on new features that could incur more cost, but they don't really provide RBAC, so anyone on your (maybe huge) team can start using those features.


I run our customer success group at Datadog and that doesn’t sound right - let me follow up with the team and look into this for you!


I love how customer service is quick and great when it’s combined with PR.


I’m surprised they haven’t had more responses on here already.

We’ve been a customer for years, but the billing has made us pretty restrictive of how and where it’s being used.


> I’m surprised they haven’t had more responses on here already.

You're not supposed to comment on anything related to your filing until your registration statement becomes "effective". See details on the quiet period here: https://www.sec.gov/fast-answers/answersquiethtm.html.


Similar issue. We had to scramble off of Datadog quickly, as they adjusted our billing and it increased by a factor of 8 in a single month. This was due to our tag usage. Same pattern, previously unbilled, and next month bill increased 8 fold.

No slow adjustment, warned us about 2 weeks prior to the next month billing cycle. If they had given us one additional month, we could have re-worked our tagging but we were given no options at all.

Lost nearly 2 weeks of productivity out of a few team members as they needed to focus on getting us off of Datadog and onto a competitor.

Never going to consider using them again.


Vendor lock-in, difficulty getting a refund, Sales department doing support.

Investors will love this.


Is the output format still very nearly statsd/graphite? If so, the lock in is a good bit less than, say, New Relic


Datadog logging and Datadog APM are a huge part of our usage now.


Big companies seem to have bad practice when it comes to cancelling subscriptions.

We started out with New Relic but was unhappy with the customer support. We wanted to upgrade our subscription and reached out 3 times. We were ignored all three times. It took 3 tries to get our subscription cancelled.

Datadog seems a lot better. Their support team actually gets back to you at least. If you are a small customer New Relic will not care for you as much as the big customers. Even though we were using $500+ per month.

To add another example Postman keeps charging my card after my account was deleted. I reached out and they said Rey couldn’t find any previous record from my email and requested I forward the email I got. But I can’t because ProtonMail doesn’t allow forwarding. To this day Postman is attempting to charge my card every month. Lucky for me I usually don’t have money on that account but once it does come in Postman can expect a chargeback and formal complaint to the authorities of their shady business practices which violates EU regulations.

TLDR; big companies suck when it comes to canceling subscriptions.


We had a different experience with New Relic - call, don't email, and you get better results.


Their sales teams are ridiculously aggressive. I've received 9+ calls from them this month, have told them that my org isn't interested, and won't be revisiting the decision for a few years. It's soured me on them altogether.


Seems like more S-1 filings have been making it to the front page of HN. Is that because more people on HN are interested in them, or because there's been a recent uptick in S-1 filings? And if the latter, is there any reason?


Investment bankers talk about “The Window.”

When The Window is open, you can “File.”

When the window is closed, you can’t file. (General rule)

The window can slam shut at any time. It is usually declared open after 1-2 obviously strong filers actually Make It Out and Trade Up, after which more marginal players can start gingerly to try going out.

Once it’s fully obvious that the window is open, IPOs are a super hot commodity and bankers will work themselves to death to take out any promising deal. These are some of the best transactions for bankers and historically were 7% headlines in fees and a lot of extra opportunity for enrichment and influence through the green shoe and the ability to dole out allocations to private clients.

When the window is closed, there might be an equivalent flow of IPOable companies ready, but they won’t make it out. They’ll wait for the window, sell in strategic M&A, or occasionally even opt to stay private.

But the reason we have so many S-1s right now is that the consensus of the animal spirits is that The Window Is Open.


A lot of tech IPOs recently after a decently long drought. There is conjecture that people are rushing to get them filed before a recession hits and it's more difficult to go public.


And more difficult to continue to secure funding for companies that have yet to make a profit.


HN is always interested in tech-related S-1s, but there has been a recent uptick in filings. The most common explanation I've seen for the increase is economic uncertainty -- everyone is trying to get their S-1s in while investors are available.


> And if the latter, is there any reason?

Recession fears


This plus a lack of IPOs in general. People are trying to find "deals" in tech especially those who have seen their peers make fortunes on the FAANG's.


Money is cheap, downturn coming — good time to have liquidity event.


A large generation of the non-investing class in the tech sector has been convinced to covet IPOs and consider them as the ultimate form of validation and success.

This non-investing class derives entertainment from seeing the lottery tickets they missed out on.

There are a few other personalities here too, such as a wealthy and now older tech sector that likes to see market trends.

But a drought of IPOs makes this of public interest.


Their sales/marketing teams are way too aggressive, IME. They cold called me on my private number, despite being on the do-not-call list. At several conferences they've the one booth one had to be careful to stay away from, lest one get into pretty aggressive sales talks (as in, "not interested" and starting to walk leading to booth folks walking with me for a bit, and talking to me).


Their product is great.

But the people I met there last year during a maybe-but-not-quite acquihire adventure we're also great. I wish them all the best with this IPO.

For those interested, I wrote a blog about what it's like for a solo founder to (almost) be picked up by company like Datadog.

https://blog.checklyhq.com/my-acquihire-adventure-with-a-lar...


Presumably you would have received options - in hindsight does any percent of you wish it would have happened now that they’re going public?


When I was there talking it was already abundantly clear an IPO was in the works. So, no. I don’t regret anything and am totally fine with the outcome.


This company has a shite hiring process. Ask you to submit pull request on github with answers to a challenge (and you can publicly view other people's answers as it is a public repo).

The answers were straight forward and I didn't pass. And I could see my answers were the same as other people submitting. Very odd company.


I've never quite understood the rationale behind the 'public PR' method of submitting coding challenges.

I understand it's much lower friction for the recruitment team, but having everyone else's answers public too kind of defeats the point entirely.


What role were you interviewing for?


My service contract expired with them this Friday. Opted not to renew because of price change in synthetics product 100x initial release. The product is reliable, but billing is troubling.


The Synthetics pricing does seem unreasonably high.

We were using Synthetics during beta alongside Pingdom - doing pretty much the same on both. Beta period ended and sales rep said our Synthetics bill would be something crazy, like 10x our Pingdom bill, and it’d start next month (they later said that was a mistake and it would actually start in 2 months, when I pushed back).

We immediately suspended all our Synthetics tests. It’s nice having them with the rest of our Datadog stuff, but it’s not 10x better than Pingdom.

Datadog is still a great tool overall, but experiences like that make me wary of putting all our eggs in the DD basket and getting locked in.


I wrote an article summarizing the S-1 and contextualizing Datadog's metrics against its competitors (New Relic, Elastic, Splunk):

https://medium.com/@jonbma/datadog-s-1-teardown-35934690e76f

A beast of a business indeed


Love the product through and through, my only annoyance is "live tail" and log lines in general show up from the top scrolling down, aka the opposite of how regular logs scroll.


I've been meaning to learn how to read and understand an S1. Does anybody have any recommendations?


There are two interesting parts to an S1 (or a 10K):

* a qualitative description of the business, as the management sees it. I find this super interesting and it's usually in the middle of the document - look for "management's discussion and analysis ..", or go here: https://www.sec.gov/Archives/edgar/data/1561550/000119312519...

* the quantitative part. There are three key, but interrelated concepts to learn. The income statement, balance sheet, and cash flow statement. A intro course to accounting should be great, I can recommend this one: https://www.wallstreetprep.com/self-study-programs/accountin...


Thanks. These look very helpful. Going to set some time aside this weekend to get to it


Hijacking this to maybe also get a recommendation -- For the purposes of better understanding S1s, I've been trying to find a good non-textbook to understand corporate finance in general. Anyone?


I've similarly wanted to start learning some basics of reading financial statements. A friend of mine in a financial role recommended the book _The Basics of Understanding Financial Statements: Learn How to Read Financial Statements by Understanding the Balance Sheet, the Income Statement, and the Cash Flow Statement_. When asking how to "read and understand an S-1", I assume you mean the financial statement figured within, so I think that book may be beneficial to read, although I've not yet read it. If someone has further recommendations geared towards S-1's or other financial statement reading that would be great!


Take an introduction to accounting course (https://www.coursera.org/learn/wharton-accounting is very good, and it's free to audit), and you should be able to understand the meaty part of S-1 (and 10-Q/10-K)


Does anybody publish "cliff-notes" versions of S-1's? Seems like a great business idea.


For starters, don't take any advice from Hackernews.


that piece of advice is a paradox

if I choose to ignore it, I have actually followed it!

heheheh


Learning how to read a document and interpreting what it means are two different things. For the former you can learn how to do so from (almost) anyone who already knows how to do so. Your interpretation of the document might also vary from whomever taught you how to read it.


Why is that?


One of the top 10 metrics they choose to highlight is operating loss in the past 6 months. They just knew what everyone was going to look for anyway!


For those interested in the financials or model building, you can grab all the tables from the S1 in a single excel file here: https://get.sentieo.com/datadog-ipo/


I got to briefly/tangentially Work with Ilan at Ooyala long time ago. Brilliant guy... very well deserved this outcome for the DD team.


This was already submitted at https://news.ycombinator.com/item?id=20781524, not sure why it didn't get marked as a duplicate.


Are there any hints on what will their market cap be on the first day of trading?


Congrats to the whole Datadog team! I have been an avid user of it at a previous job and loved the product. Not surprised to see them have a solid business. When the product is good, people buy it.


Just for curiosity, what's the legal process to go from being a French startup to a public US company? I guess there is some paperwork.



It looks like it was founded in New York, but in general the process is just making a new American corporation and transferring the French corporation (or its assets) to the American one, and dissolving the French one. Also, getting work visas for your employees and hoping that the bureaucracies don't affect you/deny you/remove your work visas after being accepted.


1. Be a smart cookie

2. Move to US on a visa (typically H-1b)

3. Get green card (hope politics don't kick you out with immigration bullshit bureaucracy)

4. Start a company

5. Hire smart solid people

6. Work at your product for years

7. Become a public US company.


Well done Datadog! Couldn't be happier for them—great product, great company!


What do their margins look like? Is it worth investing?


test1


test


I interviewed with them and it was one of the worst experiences I've had interviewing. Really shabby company.


Can you elaborate on your experience rather than just dropping a hate-bomb? (I’m actually curious). What was so bad about the interview? Why did that make the company shabby?




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: