Segment raises $64 million because marketing analytics is a truly massive space

Segment.io announced a $64 million Series C investment round on Thursday, and has now raised nearly $
ARCHITECHT
Segment raises $64 million because marketing analytics is a truly massive space
By ARCHITECHT • Issue #112
Segment.io announced a $64 million Series C investment round on Thursday, and has now raised nearly $110 million since it was founded in 2012. If you haven’t heard of Segment, it’s a platform for collecting digital customer data from sources like websites, chat, email, SaaS services, etc., and feeding it into any number of marketing analytics tools or data warehouse products. You can get more info in the blog post linked to above, or by listening to Segment co-founder and CTO Calvin French-Owen on the ARCHITECHT Show podcast in April. 
I’ll be honest: customer marketing analytics is not a space about which I care too deeply, but it’s a space that has become hugely important over the past few years. You could chalk that up to many things, but I think it boils down to the fact that the messages around big data and SaaS sunk in. And thus sprung thousands of companies and services collecting everything they can about customers, and thousands of companies willing to pay for them.
I spoke with Segment co-founder and CEO Peter Reinhardt earlier this week, and he pointed out that there currently are more than 5,000 companies listed in the MarTech Industry Index. Mature Segment customers might connect their data to dozens of different destinations (services or databases), and Segment itself connects to 20 or 30.
The result, Reinhardt explained, is that the amount of data Segment handles for customers has been doubling every six months, and the company is presently processing 50,000 data points per second. Every one of those data points is attached to somebody clicking, typing, or otherwise interacting with a customer data source.
In fact, one of the main reasons French-Owen was on podcast in April was to discuss a Segment blog detailing how it cut more than $1 million a year off its Amazon Web Services bill. It’s a really good case study in how fast a company, its customer base and its AWS bill can grow—the latter of which can be full of opportunities for optimization that require a focused effort on watching it, troubleshooting it and getting clever about how you manage resources.
Segment is also an omen for what’s about to hit (or is already hitting) companies trying to offer any sort of meaningful analytics around data derived from sensors or other previously non-digitizable interactions. I would put various AI applications—digital assistants, facial-recognition apps, chatbots, etc.—in this bucket, as well. Helping customers take advantage of these new types of interactions and integrate them with existing data is going to take even more investment in architectures, data storage and processing.
For more on this general topic, I would suggest listening to a handful of ARCHITECHT Show episodes from earlier this year. Aside from French-Owen, interviews with IBM’s Adam Kocoloski, Cloudera’s Mike Olson, Confluent’s Jay Kreps and Honeycomb’s Charity Majors come to mind off the top of my head.
In other news, Google and Alibaba are doing so much in applied artificial intelligence …
Here are links to some good blog posts and news articles from this week about what they’re building and how they’re applying AI in some really interesting ways:
P.S. The ARCHITECHT Daily newsletter, ARCHITECHT Show, and ARCHITECHT AI and Robot Show are growing fast! If you’re interested in reaching the smartest audience in IT as a sponsor, please contact me at derrick@architecht.io.

Sponsor: Bonsai
Sponsor: Bonsai
Check out what's new on the ARCHITECHT Show
In this episode of the ARCHITECHT Show, Buoyant co-founder and CEO William Morgan talks about reasons for building cloud-native applications and how the barriers to adopting those technologies have fallen away over the past few years. He also discusses Linkerd, Buoyant’s flagship technology; the company’s recent $10.5 million funding round; his time helping scale Twitter to overcome its infrastructure woes; and how Buoyant plans to monetize its open source foundation.
Highlights from last week’s ARCHITECHT Show podcast, in which IBM Fellow and Watson Data Platform CTO Adam Kocoloski, who also co-founded Cloudant, talks about the business of cloud databases, data science and AI.
Sponsor: Linux Foundation
Sponsor: Linux Foundation
Artificial intelligence
I get the privacy concerns, but if this is going to be an issue with Echo products, it theoretically should be an issue with every product we use—especially those using voice services. It is how models and products improve.
fortune.com  •  Share
Once again, this is not a zero-sum game where China “wins” because its people and companies adopt certain aspects of AI before U.S. people and companies do. Here’s my previous take on this topic.
This is really about the “unreasonable effectiveness of data, deep learning edition” paper from Monday, which is so narrow in scope that it can’t really answer this question. That being said, the answer is YES for many other reasons.
I mean, you’re not training against the ImageNet library on your phone, but this could be really useful for targeted applications in areas without internet connectivity.
The three winners here are studying brain disorders, helping companies improve personalization features, and analyzing medical records. The first two are getting $500,000 investments, while the latter gets $1 million in Google Cloud credits.
The company has some cool applications for deep learning to optimize how its shoppers shop and drivers deliver. However, it’s now going to have to deal with Amazon, which will have a lot more data about everything happening in Whole Foods and already has mastered delivery.
If you haven’t been following, this is a particularly exciting area of AI research, and this post is a good intro. The general idea is that while deep learning helped slash feature engineering, etc., evolutionary algorithms could help automate the rest of the model-tuning process.
Sponsor: Cloudera
Sponsor: Cloudera
Cloud and infrastructure
It’s location No. 2 of 5 currently planned for Europe, with Belgium hosting the other live region. At what point will cloud providers need data centers in every country?
According to this report, the company that commercialized the Riak data store is all but dead, with very few employees remaining. It’s sad, but Riak is a respected open source technology and it sounds like the community will keep that alive and thriving.
These types of companies have been around for a while, but Spotinst does have some cool tech around helping customers maximize Spot Instances (and the Azure equivalent) and a third-party “serverless” platform.
Backing up Salesforce, Slack, etc., seems logical enough, but it also seems like something that must already have some solid solutions. If I’m wrong, please let me know.
I’m really not certain if other Kubernetes distributions currently offer what you’d call “enterprise-grade” multi-tenancy (I’m fairly certain DC/OS does), but it’s more or less a prerequisite for a containers-as-a-service and true resource pooling.
Government Computing News has a couple of noteworthy articles about how the federal government is thinking about cloud computing. The headline links to a story about how IARPA is seeking a way to store classified data on public clouds. This article is more a rundown of quotes (but good ones) from a recent government cybersecurity event.
gcn.com  •  Share
Sponsor: DigitalOcean
Sponsor: DigitalOcean
All things data
The technology will be folded into Teradata’s new platforms for helping customers deploy Teradata on infrastructure other than those huge boxes, but it’s unclear if StackIQ will still exist. I’m not sure this is a savior for Teradata, but it fending off cloud+commodity competition is a critical strategy.
Its product seems more or less like a recommendation engine, but as we’ve seen time and time again recently, that is still an unsolved problem. At least doing it right is still hard.
This is a pretty informative blog post from Cloudera about the EU’s General Data Protection Regulation set to go live in 2018. The law covers everything from when you can analyze data to how to explain algorithmic decisions to consumers.
I linked to the Bullet blog post when it was open sourced, but here’s a take from analyst Tony Baer. This line pretty much predicts the project’s fate (especially considering who built it): “There’s no vendor support and it’s not part of any tool, so you’re on your own with regard to managing and integrating it.”
It’s a project out of Stanford, but it portends what I think will be a pretty big deal going forward: ChatOps. The big challenge, of course, is making sure these tools fit into existing workflows and lexicons.
Sponsor: CircleCI
Sponsor: CircleCI
Did you enjoy this issue?
ARCHITECHT
The most interesting news, analysis, blog posts and research in cloud computing, artificial intelligence and software engineering. Delivered daily to your inbox. Curated by Derrick Harris. Check out the Architecht site at https://architecht.io
Carefully curated by ARCHITECHT with Revue. If you were forwarded this newsletter and you like it, you can subscribe here. If you don't want these updates anymore, please unsubscribe here.