This can happen... but what can we all learn from this?

I have a lot of colleagues/ex-colleagues, gratitude, good memories, friendships, and respect for a company like Cloudera (and Hortonworks... and MapR, even MapR. Yes).

This can happen:

As an outsider and with some temporal space and distance, I can now express my own opinions on the underlying causes. It is never a single event or cause or a single factor.

One factor is certainly cloud. For any software vendor on this planet, it is super important to be able to provide their products on premise and on cloud, both as a product for self-managed installations and as Software as a Service for fully managed installations. Indeed some customers would like to have fully managed installations also on their own infrastructure and data centers. Hybrid is the must have now. Indeed cloud vendors that nobody would have though be able to launch an on premise product have done that. Vice versa, traditional on premise software companies know that, if they don't propose their software product on cloud and more importantly as Software as a Service with pay per usage / pay per query / pay per insight business model, they are destined to disappear.

Another important aspect to consider is the "one size fit all" approach vs. "best of breed" or "do one thing only and do it well" approach. I wrote another short article on why I believe there is no center in the IT universe. Big lesson learned for me. One intrinsic reason reason why this is true is simply because when you have a system or a platform with N components which all need to be integrated or communicate with each others there are N^2 point to point integrations or communication channels. In practical terms, this kills innovation and increases exponentially QA / integration tests costs and teams. This problem can be solved for data, but not so easily for software. Unless you have software which is driven by data or software and systems that are governed by a declarative approach, but this in itself is a log article (yet to be written).

Last but not least, to be successful with "data science", machine learning, AI, you need to be extremely good at failures. It is more important to fail fast because by doing that you will be able to run many more experiments and, hopefully, with a bit of luck also find something that's valuable. Many businesses, cultures, and individuals are not willing to accept failures or admit mistakes. Even worst, when sometimes a data scientist or a team of data scientists find some real nugget of valuable insights on the data, businesses are so bogged down by legacy systems and legacy way of thinking that they cannot in practice act on those insights or monetise them.

There are other aspects, including but not limited to: businesses and life do not happen in batches, not invented here syndrome, it's very difficult to maintain culture as you grow and scale really fast, US software companies look at the world just from their own US point of view, you cannot measure or control everything even you like or want to, less is more (but somehow many CEOs think the opposite, they like to have many products to sell or many versions/editions of the same product), unwilligness to listen to customers and act quickly upon their feedback, lack of standards among the IT industry around metadata management, computer science is still more like an art or craftsmanship than a science, innovation in the technology sector is accelerating at a pace which is not sustainable by humans or businesses (or our own planet), etc.

To conclude: cloud, "one size fit all" or the N^2, and the unwillingness to accept failures or make many experiments quickly are, in my humble opinion, the most important factors which have impacted the growth and expansion of a company like Cloudera.

There are many software companies and startups like Cloudera. Indeed, even a cloud vendor like AWS has some of these characteristics (at a completely different scale in terms of business): "one size fit all" attitude, N^2 integration testing nightmares.

Here is the check list for your next software investment or your next business venture or your next job (I write them down here also for my future self and Confluent):

• Is it "one size fit all" approach or "best of breed" i.e. "do one thing only, do it very well"?
• Is it available on premise, on cloud (both as self managed and fully managed service)? If it is also available on premise / private cloud as a fully managed service: bingo!
• Does it enable a fail fast, fail many times approach in a low risk and evolutionary rather than revolutionary fashion?
• Does it solve the N^2 problem?

In the end, it is business as usual, the world goes on... some data scientists or data engineers might complain. Nobody in the business or even IT, listen to them anyway, so... it's business as usual. A few, but it's only a very small fraction, will continue to apply technology to the right business problems, with the right people. These are only a few, they highly correlate with the leaders in every industry sector, and they are there because of how they apply technology to extract value from their data and apply those insights to drive their business.