In Dialog with Barry McCardel, CEO, Hex – Matt Turck

Within the ever vibrant world of the “Fashionable Knowledge Stack” (an ecosystem of principally younger tech startups that signify the rising era of information software program distributors, and combine effectively with each other), Hex has been getting growing visibility and momentum. At its core, Hex is a collaborative information platform the place groups can discover, analyze, and share. It goals to deliver collectively the perfect of notebooks, BI & docs right into a seamless, collaborative UI.

The corporate was based in 2019 and also you raised a complete of $73.5 million in enterprise capital thus far, together with most lately a $52 million Sequence B.

CEO Barry McCardel joined us at Knowledge Pushed NYC for a deep dive in to the product, the corporate, the info area and his journey from doing “unholy issues in Excel” as a younger marketing consultant to constructing an amazing startup.

Beneath is the video and full transcript.

(As at all times, Knowledge Pushed NYC is a workforce effort – many due to my FirstMark colleagues Jack Cohen, Karissa Domondon Diego Guttierez)


TRANSCRIPT [edited for clarity and brevity]:

[Matt Turck] Why does the world want a collaborative workspace for information groups? What’s the massive downside that you just’re engaged on fixing?

I’ve been working in information successfully my entire profession. I began as an undergrad, actually, doing a bunch of stuff, simply writing R scripts and all this analysis stuff earlier than information science was even a factor, actually. After which was doing unholy issues in Excel as a marketing consultant. Then I used to be at Palantir for 5 years, the place I received publicity all throughout a bunch of various technical issues. Then I labored actually intently with our information workforce at my final firm. And all alongside the way in which I mainly noticed the identical set of issues. In essence, Hex is supposed to unravel these.

So the very first thing we actually set out on and the factor that was a very acute downside that we needed to unravel was across the capacity to share work. It’s this quite common factor and we noticed it up shut at our final firm we had been at which is you could have information analysts and information scientists and simply folks working with information all around the enterprise doing actually fascinating, superior issues. They’re getting in. They’re asking and answering questions. They’re driving perception.

Then the precise capacity to share and publish that all through the group is terrible. It’s actually a catastrophe. You’ve folks screenshotting charts out of Jupyter Pocket book and pasting them in Google Docs. You’ve folks exporting a CSV from a BI device to allow them to construct the fitting match on this different factor after which put that in a deck. Then you could have folks hacking collectively scripts to attempt to construct a pipeline to place the forecast within the warehouse so you possibly can have a look at it within the BI device.

It was this big mess. So we began actually specializing in that downside. The preliminary factor we had been centered on was, how are you going to assist information scientists who’re engaged on one thing like a Jupyter Pocket book, take that and share it with others in a means that’s interactive and helpful and usable? As we began stepping into that, we realized the ache was actually a lot deeper than that. It was really like folks had been simply pissed off with the entire stack. You had people leaping round between instruments relying on whether or not they’re utilizing SQL or Python or no code. You’ve received groups actually unable to collaborate. The entire versioning and real-time collaboration for all that is only a mess. It’s very regressive in comparison with instruments in different areas, like Figma or Google Docs.

Then there’s simply an quantity of overhead and ache to getting these instruments up and operating anyway that’s really actually laborious. There’s a really traditional expertise the place you’ll see a brand new information scientist will be a part of and the primary two weeks are actually nearly getting all the fitting packages put in regionally of their Jupyter atmosphere after which ensuring that’s synced up. You wind up with this overhead that’s each very irritating for the people who find themselves doing these workflows but in addition prevents lots of people from accessing.

So again to your query, Hex is absolutely meant to be three massive issues. It’s an incredible, collaborative atmosphere for with the ability to do evaluation and information science. It’s received a pocket book UI that’s simply completely magical. I’ll present that to you in a bit. It’s very, very simple to take your work and share it and publish it as an interactive information app that anybody can use. Then that work is then stored and arranged in what we name a Data Library which makes it very simple for anybody else within the group to find and profit from the work that the info workforce has been doing. So mission smart, that’s actually what we’re about and we constructed a product that actually addresses that finish to finish.

Nice. And it’s, by definition, meant to be very inclusive, proper? So it’s information scientists. It’s information analysts. It’s enterprise folks as effectively. You’ve an expression that I learn someplace, which I actually preferred, which was the “analytically technical.”

Analytically technical, yeah. It’s fascinating as a result of you consider a few of the massive adjustments which have occurred in the previous few years. You see this explosion in people who find themselves information literate. They’re even, I might name them, considerably technical. And there’s extra individuals who know Python, positive. There’s much more individuals who know SQL. And lots of people have both discovered SQL on the job or are available out of undergrad with that skillset. There’s additionally this a lot greater inhabitants of those who I might argue which can be technical in their very own means; which in the event you’re an Excel Energy consumer and also you’re writing deeply nested features or VBA and even just a few pivot tables or IFs, you might be mainly writing code. I might argue you might be writing code. You might be technical ultimately.

And I believe conventional information science and analytics instruments have really been a excessive tower. They’re troublesome for these folks to entry. And so one of many issues that’s actually fascinating for us, what we see in our clients, is now we have numerous customers. In truth, most of our clients, many of the customers are principally writing SQL. And that’s very totally different than what you may consider if you consider a pocket book atmosphere, which is historically very related to Python and, quote-unquote, “information science.” However Hex makes it very simple to travel between SQL and Python. You may collaborate between these. And so it’s very inclusive.

It’s very cool for us to see that our clients will begin with a really small variety of information scientists, a pair people who find themselves migrating their workflows over from Jupyter however then will explode to the place you see all types of individuals utilizing Hex to ask and reply questions. That’s one thing we’re very enthusiastic about. I really feel like we’re simply nonetheless on the tip of the iceberg. And we consider it as constructing a platform that has a low ground and a excessive ceiling. We wish to have a platform that anybody can are available and ask and reply questions. Nevertheless it doesn’t arbitrarily prime out.

And I believe that’s a giant distinction between the final era of instruments, which is like, “Okay. It is a no-code factor. It’s received a low ground and a low ceiling.” However the second you wish to do one thing extra complicated, you’ve topped out. And now you could have a UX SqlRunner. Medium ground, medium ceiling. After which, “Okay. Now I’m over in my Jupyter Pocket book,” excessive ground, excessive ceiling. I problem why this must be three fragmented issues. And I believe we’ve finished an amazing job thus far with the ability to deliver a few of these extra collectively.

So to take a few of us via a bit and drill into the following degree, so the core is a pocket book. We talked about exhibiting the product. So I’m excited for a product demo. However simply at a excessive degree earlier than we leap into the demo and possibly to make it inclusive for everybody, so only a 10-second definition of what a pocket book really is.

Yeah positive so notebooks have been round for a very long time. As legend has it, they had been first pioneered at Mathematica. And the commonest one now’s a venture known as Jupyter. It was once known as IPython.

That was within the ’80s, proper?

Yeah. Properly, I imply, Mathematica is an actual OG. IPython’s a bit bit newer. After which it was rebranded as Jupyter, I believe, in 2015, one thing like that. However anyway, the pocket book format is mainly you’ll have cells which have code historically. After which these cells present the output of that code. And people cells may be evaluated individually. That is totally different from a script. A script is one file. And the script is often evaluated, the entire thing, prime to backside.

And this breaking it up into cells makes it actually nice for iterative and exploratory evaluation. So you possibly can say, “I simply wish to run this little chunk. And, oh. I wish to do the aggregation a bit bit totally different. I wish to do that.” And that is all an expression of a factor known as literate programming. I cannot go within the deep finish on this. However mainly, it’s this concept that you would be able to see your logic after which the outputs in a single place. It’s a really, very talked-about format. I imply, tens of millions of individuals use notebooks. However we expect that it’s really a format that much more folks ought to be utilizing. We’re very completely happy to see that with our consumer base and clients.

Yeah and simply even at the next degree, a pocket book is a spot the place information scientists and information analysts work collectively. And it’s a mix of code and clarification. So it’s like a piece area.

That’s proper. And it’s actually the factor. Properly, in the event you speak to numerous the info scientists, particularly, it’s the factor they use all day. It’s the factor the place they’re going and writing code. They usually’re iterating on one thing. Now, notebooks additionally historically have numerous points. There’s a well-known speak known as I Don’t Like Notebooks that this man Joel Grus gave at JupyterCon. It was very simply exhibiting up within the flawed place to present that speak. However he was proper. There’s all these points.

It was like 4 years in the past or one thing like this.

Yeah. It was 2018, I believe. Nevertheless it was all these points. A part of what we’re doing at Hex is, “Properly, notebooks are nice. They’ve some points.” I believe there’s a camp of individuals which can be like, “Due to these points, everybody ought to be doing one thing like writing scripts or no matter.” I believe we’re looking for that synthesis of, “Properly, what if we simply repair these points with notebooks and made them superior and made them accessible to 100 instances the folks? I believe this really might go someplace.” And that, in a very simplistic means, is what we’ve been as much as on numerous issues.

A part of what you had been describing was one of many key points – simply to verify I paraphrase and I make sure that I understood accurately, one massive subject of notebooks is that you would be able to have totally different definitions of a variable in a pocket book.

Yeah. We name this a state subject. So I might get away some points with notebooks. I might say the very first thing is accessibility. I used to be getting at this earlier, however most individuals working with information in most locations have by no means used a pocket book as a result of the first step is studying computer systems. You need to determine the right way to arrange a neighborhood Python atmosphere and set up Jupyter. And most of the people aren’t going to do this. Factor two is state. That’s what you’re getting at. And the quick model of that is notebooks historically run in what’s known as a kernel. It’s mainly reminiscence area the place you run one thing like, “X equals 1.” Now in reminiscence, X equals 1.

However as a result of you possibly can run cells out of order, it’s really you may get in these bizarre state points the place you possibly can’t really know what state issues are in. You’ve one cell that’s X equals 1. And one other cell is X equals 20. In the event you ran X equals 1 earlier than X equals 20, effectively, now it’s one and vice versa. So it will get actually sophisticated. For many who aren’t acquainted, now we have an entire weblog put up about it. However the quick model is it is a ache within the ass for individuals who have been utilizing notebooks a very long time, like me.

Nevertheless it’s actually painful for people who find themselves new to it, who’re like, “What’s happening?” You lose lots of people. And we consider this as one of many many issues which can be in that low ground, excessive ceiling of, how will we make notebooks superior, higher for these energy customers? But in addition, how do you make it extra accessible and usable and welcoming for this greater inhabitants of those who we expect deserve nice instruments?

There’s different points with notebooks we’re engaged on too. However that state subject… we launched a function final October. We known as it Hex 2.0. Nevertheless it was this reactive compute engine we had. And I’ll present it off in a minute. Nevertheless it’s successfully saying, “What if notebooks labored a bit bit extra like a spreadsheet the place cells have this sense of provenance between them?” While you replace one factor, it routinely updates downstream cells. And the state is in a significantly better state. State is in a greater state.

And that is nice. That is higher for these energy customers who’re like, “Man, that is the way in which I at all times wished this labored.” And it’s nice for novice customers, who numerous them have by no means used a pocket book earlier than. They’re not even conscious there’s a state subject. They only know they don’t have that in Hex. So it’s all good. And that was the purpose of that function for us.

As only one final query on notebooks earlier than we leap into the demo – a part of the worth proposition as effectively is that you are able to do information science within the Python world. However you may also do SQL and databases. And I believe you are able to do that in Jupyter as effectively by putting in packages. Nevertheless it all comes out of the field. Or is that not appropriate?

Properly, you possibly can. I imply, this was the factor after we began out once I was like… if you’re beginning an organization, you get an concept, you’re pitching folks the concept. And it’s not unusual for folks like, “Properly, that’s already doable.” I’d be like, “Properly, what if it labored like this?” Persons are like, “Properly, Barry, that’s already doable.” “Oh. Actually? Have I missed one thing?” “Properly, in the event you set up these three packages and then you definately’re keen to, when you have the atmosphere variables all arrange accurately and then you definately roll your personal reference to SQLAlchemy after which write your [inaudible], yeah, you would completely write SQL in notebooks.” That’s an terrible expertise. And never solely do I hate doing it as somebody who’s technically able to doing it however what about all these people who find themselves not going to battle via all of that ache or don’t have the flexibility to do this?

And I believe it’s the identical with the sharing factor. I used to be like, “Properly, what if it was very easy to publish your pocket book in a means that anybody might use?” And it’s like, “Properly, that’s doable.” There’s these three open supply packages that in the event you set up them in your JupyterHub occasion and everybody’s utilizing the fitting model of JupyterLab and so they’re all updated. And, oh. Properly, these extensions are incompatible. However ignore that. And in the event you do that all proper after which Mercury is aligned with Jupiter the fitting means, then you are able to do it. And by the way in which, you’re going to want a full-time individual to handle all of this.

That is the kind of shit that’s solely accessible to those actually technical customers and turns lots of people off from these workflows. And we don’t suppose it must be this fashion. So whether or not it’s the SQL stuff or reactivity or lovely no-code charts, we’re simply making it actually freaking simple to share your work with anybody. We predict that there’s a approach to make this extra accessible with out dumbing it down. Our energy customers love these things too. That’s the place, I believe, there’s this false dichotomy generally of, are you constructing for low-end customers or constructing for energy customers? We predict there’s numerous sensible folks. We predict there’s lots of people that, given the fitting instruments, will have interaction with these information workflows. And we’re all about constructing for that inhabitants.

Superior. Adore it. All proper. Let’s leap into the demo.


So switching tacks a bit bit, you guys appear to have finished a very nice job partnering with numerous corporations within the ecosystem, together with numerous corporations we’ve had at this occasion through the years, together with, in your spherical lately, I noticed that each Databricks and Snowflake invested within the firm. However earlier than that, you had bulletins with metric retailer corporations and different corporations, like dbt.

dbt is a giant accomplice. Yeah.

Yeah. So is {that a} go-to-market technique? Is {that a} product? How do you consider it?

Properly, it’s each. I believe the partnerships with Snowflake and Databricks are very fascinating in that… I didn’t discuss this earlier however we’re actually constructing a product to embrace what we consider because the cloud information period, which is you could have information that’s an enormous scale, saved in cloud information warehouses. And people cloud information warehouses aren’t simply there for storage. Databricks and Snowflake and different corporations are additionally constructing very highly effective compute primitives whether or not it’s simply with the ability to push a question down totally different warehouse sizes and even with the ability to push Python code down. We predict they’re doing an amazing job with that. We predict that they’re going to proceed to do an amazing job with that. And we wish to accomplice actually shut with them on that.

So the partnership makes a ton of sense as a result of when individuals are utilizing Hex, they’re going to be asking and answering questions on extra information. They’re going to be pushing extra workloads all the way down to these information warehouses, which is nice for them. And people information warehouses additionally present a very nice scale and information story for us. We really must do much less on our finish to construct out an entire compute infrastructure and ecosystem ourselves in the event that they’re doing an amazing job of that. So we expect that partnership makes a ton of sense. We see our clients actually pulling us on, how are we integrating very intently with these applied sciences that they’re already investing in?

After which dbt? dbt is the concept that you’re constructing some transformation within the pocket book?

You actually might. There’s a pair fascinating angles. We really simply revealed a weblog put up, one in all our first analytics engineers, she makes use of dbt and Hex all day. She’s received a really cool workflow the place she’ll develop numerous stuff in Hex, deliver it over to dbt. She’s revealed a weblog put up on our website about how she makes use of them collectively, which could be very cool. However going a bit bit deeper than that, I believe, if you have a look at what dbt is doing or what corporations like Rework are doing on the metrics layer, they’re actually virtually unbundling BI on this actually fascinating means the place they’re saying, “Hey, it’s not nearly remodeling information as normalized tables in your warehouse. It’s about the way you’re then really turning them into metrics and measures and semantics which can be accessible to BI and analytics layer.” And so we’re very enthusiastic about what’s taking place there.

We predict it’s very a lot in its infancy. However because it matures, we expect there’s a very cool alternative to deliver that extra in Hex the place you would have folks as a substitute of getting to put in writing a ton of SQL in Excel, possibly they’re capable of write one thing way more concise or possibly one thing extra UI pushed, the place they will simply choose a metric they need, get a knowledge body again after which begin working in opposition to that. So now we have a ton of shared enterprise with dbt right this moment. However I believe with the place they’re going and the place we’re going, there’s much more that we’re going to be doing collectively and others in that area.

Nice. So possibly to shut, I’d like to spend two or three minutes on go-to-market gross sales. Who do you promote to? Who’s an amazing buyer? Perhaps, who’re some current clients? That facet of the enterprise.

Yeah. So we’re utilized by over… I believe the final rely was over 150 groups globally now, information groups paying us for Hex, which we’re extraordinarily pleased with. We help actually massive public corporations, like Persion Prescribed drugs, for instance. They use Hex enterprise large to help their analysis efforts.

We’re additionally utilized by small startups. I believe the one constant factor throughout our buyer base and the place we’re actually resonating, they’re making investments in information infrastructure and information. And we simply talked about Snowflake and Databricks and dbt.

If corporations are adopting these applied sciences, they’re typically then coming to Hex for, “Nice. I’ve received all this information now in my warehouse. I’m remodeling it with dbt. Now I really need to have the ability to ask and reply questions of it. I would like to have the ability to do extra with it than I’m capable of in legacy instruments or Jupyter Notebooks or SQL Scratchpads.”

Hex is a brilliant good complement to corporations which have invested in that stack. And what we’re seeing is any firm that’s hiring of us in roles like information science, analytics, analytics engineering actually wants and desires and will get a ton of worth out of Hex. So, yeah. From a buyer and goal perspective, that’s the place we’re at proper now.

Very cool. Congratulations on all of this and this journey. The corporate remains to be fairly younger. I imply, it looks like you guys are executing extremely effectively and really quick.

Yeah. We began in late, late 2019. So actually, not been at it too lengthy. I used to be very, very lucky to start out at an organization with two of us that I had labored with at Palantir, Caitlin and Glen. And we’ve had numerous enjoyable and a bit little bit of luck the final couple years constructing this out. So we’re trying to proceed that streak for a bit bit longer and preserve happening. We’re having a great time.

Very cool. Properly, thanks a lot, Barry, for coming to this occasion, telling us your story. Better of luck for the longer term. I hope you come again in a few years.

Yeah. Mockingly, I’m really in New York this week, not that removed from you. However we’re nonetheless doing this digital. So possibly subsequent time I come again, we’ll have the ability to do it in individual.

Do it in individual. Sure. I can’t look forward to that. Okay. Cool. Thanks a lot, Barry.

Leave a Reply