My open source product is due out early next year, after well more than a year in development. Whether it needs to be Open Source or not was a decision I considered very carefully. In the end, I chose Open Source because:
a) My product is a application development platform that benefits from people building on it and sharing the code.
b) I don't mind making money (at all), but I don't need _a lot_ of money.
c) I believe that the way we build companies today will change in future. We might move to a world where something _like_ Wikipedia or Redhat or Mozilla will be considered the benchmark for ultra-successful companies. Creators will provide leadership and benefit from community participation, while the community benefits from shared ownership of knowledge, code, ip etc. Financially, it will be a transition from the high-risk & very-high-reward system we have now, to a lower-risk, lower-reward system.
Your point b) resonates with me... But my experience with the dual-licensing model: it often took 3-4 months for companies to pay ($1,000-5,000 per license). Because they already had the software.
> it’s hard to make money when you give the shit away.
Well, it's also not easy when you charge for things. I guess the important thing is to have a business model, regardless of the openness of the software.
Between the two extremes of proprietary software and putting everything on a public Github repository, there's a middle term: your software is FOSS, but it's not publicly available - the source is only available to paying customers.
I never see this model discussed, but I've actually seen it work fine for many not very large software companies. They still get to sell the product, and the client gets the peace of mind of knowing that they have an alternative if the company goes under or discontinues the product. (This works better if the development is done on a common platform like Drupal, instead of being custom from the bare metal.)
Of course, the clients can still turn around and distribute (and even sell!) the source, but in practice the average SMB isn't particularly interest in going into the software reselling business.
It's valid, but not entirely in the spirit of open source software either, in my opinion.
There are certainly various open source business models, it's just that they require more thought and effort than simply selling the software, or access to it. For some companies, that extra difficulty might be what makes or breaks them.
"That’s why there is no Red Hat of Lucene or Solr, despite their enormous popularity in search"
You forgot ElasticSearch, which is not a Red Hat but seems to be doing fine and working on getting there.
They have some interesting takes on open source, they give the dough for free but the frosting like security and managing interface are charged/subscription, I think it can work.
Also their software is built on top of SOLR/LUCENE but with lot more enterprize keywords thrown at it (Distributed,Clustering....) So makes it easier to sell for the
Elasticsearch seems to have shifted their positioning from pure Search Engine to the log analytics with strong emphasis on the ELK stack.
They had early advantage over Solr by addressing the Innovator's dilemma and tackling things Solr did not touch (better clustering, JSON interface, smaller initial downloads, dynamic configuration, etc).
But the actual commercial entity on top of Solr (LucidWorks) has finally noticed that and has put some resources in closing the gap and maybe even getting ahead of Elasticsearch. And, of course, their own commercial product (Fusion) has all those enterprize keywords covered as well.
Also, LucidWorks is not the only commercial company putting money into Solr. Cloudera (yes, the one with 1.2 billion $ of investment) is fully committed to Solr and is contributing to it as well.
I'd say LucidWorks wants to the be The Red Hat of Solr, while Elasticsearch is trying to take over Splunk in that very specific (and, as Splunk proved, fairly lucrative) part of the log collection/analysis business.
log collection? eww.. that's so boring. For a company that's trying to focus on log analysis, they sure have a much better search engine than SOLR (see percolation, sharding, distribution capabilities, REST interface)
# The REST interface is nice but is also coming to Solr really fast
# The rivers are deprecated. The scripting (enabled by default originally) had to be reworked due to Amazon shutting down every EC2 Elasticsearch instance they detected.
# Some other features have been deprecated and now are basically the same as Solr (e.g. copy_to vs. copyField).
The production recommendations from Elasticsearch specialists are to disable dynamic schema, _source, and _all fields.
Not that Elasticsearch is bad. I just feel it is not fair to compare a commercial solution (that does not even give you admin console for free) to a fully open-source solution (that does). If you are planning to pay money for commercial solution, then compare it with another commercial solution (e.g. LucidWorks Fusion). Or at least compare latest Elasticsearch (1.4) to the latest Solr (4.10.2 currently).
$33M of revenue is unsuccessful? That should be a very comfortable revenue stream for a small-medium software company. Not every company needs > 200 employees, not every company needs to IPO for billions.
It also seems obvious that they'd be making far less than $33M in revenue if they weren't open source. If Hadoop was not open source, everybody would be using the open source competitor that would magically spring up, not Hadoop.
That's what should be questioned in the article, not the open source nature. Hortonworks is in the consulting business, but they don't seem to suffer from the perennial consulting problem of variable revenue. So why aren't they printing money?
It looks like they're also spending money developing stuff. If you want to optimize for profits, you need to let other people actually work on the free code and crank up the billable hours.
Of course, then the actual products don't get developed as quickly, so it's a bit of a paradox.
From what I've seen though, consulting is not necessarily a very good way of producing open source software.
Many times the difference between success and failure is your definition of success. VC and an exit after years of grueling work is only "success" for a small number of people who are disproportionately represented here on HN.
FOSS has value in other places too. In particular it has value when software isn't your core business. As such, you develop instrument software, analysis tools, support tools etc. which supports your core business and give it away for free.
This does 2 things:
1. If your open source package gains widespread adoption, it gives you control of the tools that most users in your area use.
2. It levels the playing field and takes software out of the equation to a degree. Competitors have little reason to develop their own platforms when they can leverage yours.
----
In addition to that, I think FOSS software is valuable almost everywhere else, but as you say, does lead to different models.
Data analysis software is probably a good example of this. You write the analysis software, it gains widespread adoption. You provide a data analysis as a consulting service. You can claim to be the leading experts in the leading package (because you wrote it). I think this model also works for RedHat and others.
Yes, I have forgotten about apps behind pay walls.
As for native apps, good luck with the business if the full code is available and then you have to start fighting against clones that just relabel your work.
It also doesn't help that the Hadoop ecosystem is unstable, insecure, and overkill for the vast majority of the use-cases out there. Cloudera can also be thrown into the same bucket when they will no doubt file next year.
IMHO Hadoop just doesn't have a viable business model behind it. The people that really need a big data solution will run it themselves, and the people that don't (who will have it shoved down their throats all in the name of big data anyways) will unsuccessfully run it for a year or two before scrapping it for a better, more appropriate solution.
Part of the added value of open source is that you are giving away part of the product for free. It's intrinsically less profitable than closed source presuming the consumers are willing to buy the closed source alternative.
But if consumers start deciding that closed source software is unacceptable then the situation will change.
Figure out
A) is there something customers value
B) would they pay money for it
C) how much and for how long
D) is the value something you capture (how much would YOU keep)
E) would the money you capture allow you to survive for a while based on your capital and shareholders and give them (or yourself) something back
Otherwise you are not building a business. You are just keeping busy or worse wasting someone's money.
How does that screw the employees? The S-1 linked from the article talks about absolute dollars to be raised, not share count. The employee-held shares should just now have twice the value that they had before the consolidation. Am I missing something?
If I have 10 options in a company that has 100 shares outstanding (including my options for the sake of easy math), then the reverse-split will take this to 5 options out of 50 total. Now, imagine that the company is valued at $1M. This valuation is independent of the share count.
So, in the scenario before the reverse-split, each of my options would be worth $1M / 100, or $10k. And I have 10 of them, so that's $100k.
In the post-reverse-split world, the company is still worth $1M, but now there are only 50 shares outstanding, so now each option is worth $1M / 50 each, or $20k. And that comes to the same $100k total.
Now, I believe that typically, the option pool is distinct from the issued shares. However, that shouldn't really change the math, unless the reverse split was applied only to issued shares and not to options. If that was the case, then yes, the option holders would be screwed.
In any event, it seems like pretty much everyone uses RSUs these days instead of options. Does Hortonworks still use options? With RSUs, the above math definitely works out, as the RSU is backed by an issued share, not a yet-to-be-issued one.
Running an open source company myself (also somewhat overlapping with the hadoop ecosystem). First of all, we can all be armchair analysts so take this with a grain of salt.
I think Horton IPOed too early.Their alternative may have been an acquisition. I would say wait till cloudera IPOs before we really judge the market on this one. As I said before though: I'm biased.
I work with a number of the big open source players and I'd say there's real revenue here, especially in machine learning.
Software at the end of the day isn't about whether it's open or not, it's how you sell it and package it.
A great alternative example to this is docker. Their registry is a great way of monetizing open source while working towards a scalable business.
I think we will see innovation in the open source business models come out here in the next few years. The best is yet to come imo. I wouldn't be doing an open source business if I didn't see the potential.
32 comments
[ 2.8 ms ] story [ 68.1 ms ] threada) My product is a application development platform that benefits from people building on it and sharing the code.
b) I don't mind making money (at all), but I don't need _a lot_ of money.
c) I believe that the way we build companies today will change in future. We might move to a world where something _like_ Wikipedia or Redhat or Mozilla will be considered the benchmark for ultra-successful companies. Creators will provide leadership and benefit from community participation, while the community benefits from shared ownership of knowledge, code, ip etc. Financially, it will be a transition from the high-risk & very-high-reward system we have now, to a lower-risk, lower-reward system.
Well, it's also not easy when you charge for things. I guess the important thing is to have a business model, regardless of the openness of the software.
Between the two extremes of proprietary software and putting everything on a public Github repository, there's a middle term: your software is FOSS, but it's not publicly available - the source is only available to paying customers.
I never see this model discussed, but I've actually seen it work fine for many not very large software companies. They still get to sell the product, and the client gets the peace of mind of knowing that they have an alternative if the company goes under or discontinues the product. (This works better if the development is done on a common platform like Drupal, instead of being custom from the bare metal.)
Of course, the clients can still turn around and distribute (and even sell!) the source, but in practice the average SMB isn't particularly interest in going into the software reselling business.
http://en.wikipedia.org/wiki/Cygnus_Solutions
It's valid, but not entirely in the spirit of open source software either, in my opinion.
There are certainly various open source business models, it's just that they require more thought and effort than simply selling the software, or access to it. For some companies, that extra difficulty might be what makes or breaks them.
You forgot ElasticSearch, which is not a Red Hat but seems to be doing fine and working on getting there.
They have some interesting takes on open source, they give the dough for free but the frosting like security and managing interface are charged/subscription, I think it can work.
Also their software is built on top of SOLR/LUCENE but with lot more enterprize keywords thrown at it (Distributed,Clustering....) So makes it easier to sell for the
Elasticsearch seems to have shifted their positioning from pure Search Engine to the log analytics with strong emphasis on the ELK stack.
They had early advantage over Solr by addressing the Innovator's dilemma and tackling things Solr did not touch (better clustering, JSON interface, smaller initial downloads, dynamic configuration, etc).
But the actual commercial entity on top of Solr (LucidWorks) has finally noticed that and has put some resources in closing the gap and maybe even getting ahead of Elasticsearch. And, of course, their own commercial product (Fusion) has all those enterprize keywords covered as well.
Also, LucidWorks is not the only commercial company putting money into Solr. Cloudera (yes, the one with 1.2 billion $ of investment) is fully committed to Solr and is contributing to it as well.
I'd say LucidWorks wants to the be The Red Hat of Solr, while Elasticsearch is trying to take over Splunk in that very specific (and, as Splunk proved, fairly lucrative) part of the log collection/analysis business.
# The percolation apparently does not scale: (as compared to https://github.com/flaxsearch/luwak - Not in Solr yet, granted ),
# The sharding and distribution has (in ES 1.3) serious issues due to the in-house implementation (http://aphyr.com/posts/317-call-me-maybe-elasticsearch)
# The REST interface is nice but is also coming to Solr really fast
# The rivers are deprecated. The scripting (enabled by default originally) had to be reworked due to Amazon shutting down every EC2 Elasticsearch instance they detected.
# Some other features have been deprecated and now are basically the same as Solr (e.g. copy_to vs. copyField). The production recommendations from Elasticsearch specialists are to disable dynamic schema, _source, and _all fields.
Not that Elasticsearch is bad. I just feel it is not fair to compare a commercial solution (that does not even give you admin console for free) to a fully open-source solution (that does). If you are planning to pay money for commercial solution, then compare it with another commercial solution (e.g. LucidWorks Fusion). Or at least compare latest Elasticsearch (1.4) to the latest Solr (4.10.2 currently).
But of course being commercially successful is not the same as being ready for an IPO.
It also seems obvious that they'd be making far less than $33M in revenue if they weren't open source. If Hadoop was not open source, everybody would be using the open source competitor that would magically spring up, not Hadoop.
It is if you need over 500 employees and lose $54M while making that $33M revenue.
Of course, then the actual products don't get developed as quickly, so it's a bit of a paradox.
From what I've seen though, consulting is not necessarily a very good way of producing open source software.
If you spent $114 million, could you make $33 million in revenue? I hope so.
It's definitely not an easy road, but then again neither is starting a funded startup.
[1]: http://mikeperham.com/2014/10/01/the-path-to-full-time-open-...
[2]: https://www.payola.io/pro
I think that is a pending discussion here on HN because it is considered a lifestyle business instead of a startup that grows fast.
Many times the difference between success and failure is the size of the business you have in mind.
This is why I moved from being a FOSS zealot in the mid-90s during the .com boom to a consultant in whatever software our customers ask for.
FOSS is nice when one is an university student with parents paying for bills.
It is a complete different matter when one tries to make money out of free stuff and VC that give money for whatever idea are gone.
FOSS only brings money home in consulting, support and teaching, everything else requires a second job to keep money around.
This does 2 things:
1. If your open source package gains widespread adoption, it gives you control of the tools that most users in your area use.
2. It levels the playing field and takes software out of the equation to a degree. Competitors have little reason to develop their own platforms when they can leverage yours.
----
In addition to that, I think FOSS software is valuable almost everywhere else, but as you say, does lead to different models.
Data analysis software is probably a good example of this. You write the analysis software, it gains widespread adoption. You provide a data analysis as a consulting service. You can claim to be the leading experts in the leading package (because you wrote it). I think this model also works for RedHat and others.
No, it's perfectly possible to make money hosting FOSS SaaS apps, selling FOSS apps on various App Stores, selling copies to SMBs, among others.
Evidence: my paycheck working for an (almost) 100% FOSS company.
As for native apps, good luck with the business if the full code is available and then you have to start fighting against clones that just relabel your work.
IMHO Hadoop just doesn't have a viable business model behind it. The people that really need a big data solution will run it themselves, and the people that don't (who will have it shoved down their throats all in the name of big data anyways) will unsuccessfully run it for a year or two before scrapping it for a better, more appropriate solution.
But if consumers start deciding that closed source software is unacceptable then the situation will change.
It's not.
Figure out A) is there something customers value B) would they pay money for it C) how much and for how long D) is the value something you capture (how much would YOU keep) E) would the money you capture allow you to survive for a while based on your capital and shareholders and give them (or yourself) something back
Otherwise you are not building a business. You are just keeping busy or worse wasting someone's money.
This applies whether you open source or not.
If I have 10 options in a company that has 100 shares outstanding (including my options for the sake of easy math), then the reverse-split will take this to 5 options out of 50 total. Now, imagine that the company is valued at $1M. This valuation is independent of the share count.
So, in the scenario before the reverse-split, each of my options would be worth $1M / 100, or $10k. And I have 10 of them, so that's $100k.
In the post-reverse-split world, the company is still worth $1M, but now there are only 50 shares outstanding, so now each option is worth $1M / 50 each, or $20k. And that comes to the same $100k total.
Now, I believe that typically, the option pool is distinct from the issued shares. However, that shouldn't really change the math, unless the reverse split was applied only to issued shares and not to options. If that was the case, then yes, the option holders would be screwed.
In any event, it seems like pretty much everyone uses RSUs these days instead of options. Does Hortonworks still use options? With RSUs, the above math definitely works out, as the RSU is backed by an issued share, not a yet-to-be-issued one.
I think Horton IPOed too early.Their alternative may have been an acquisition. I would say wait till cloudera IPOs before we really judge the market on this one. As I said before though: I'm biased.
I work with a number of the big open source players and I'd say there's real revenue here, especially in machine learning.
Software at the end of the day isn't about whether it's open or not, it's how you sell it and package it.
A great alternative example to this is docker. Their registry is a great way of monetizing open source while working towards a scalable business.
I think we will see innovation in the open source business models come out here in the next few years. The best is yet to come imo. I wouldn't be doing an open source business if I didn't see the potential.