Thoughtful Friday #3: A Deep Dive into Hex Technologies (Or Any Other Data Company)
The series A announcement of the company Hex Technologies just prompted me to take a closer look at them.
I already took a quick look inside my newsletter #43, but I only focused on one interesting thing there, the “data vision” which they seem to have and play out really well.
This week I want to take a much deeper dive and look through the seven different lenses I’ve been using throughout the year to look at data companies….
So let’s take a real deep dive!
Svens Actionable Thoughts
- Hex Technologies possesses a still evolving but existing vision for the whole data space. Since the space is in turmoil, this is crucial to any company! It’s even more important because they are building an internal platform, which relies on a strong story because it has to be sold not just to one user, but to a set of multiple users all at the same time.
- Hex Technologies is a company aiming to combine different parts of the data pipe to provide value to the end-user. Combining more than one stage enables data companies to remove bottlenecks in terms of the value they can provide to their end-users.
- Hex Technologies indulges neither in network effects nor in open-source. Both strong points which data companies should ask themselves whether they cannot change course to include either into their business core.
- The value the company will provide to users in the future will grow with the exponential data growth. This is because they tap into two major sources of value, the decentralization of data & data usage as well as the data snowflake problem. Currently, they are missing out on the other two forces.
- The company doesn’t have a public pricing model yet, so we’re playing around with one; Read below how a data company could derive one which is aligned with the end-user value as well as optimizes for a heterogeneous set of needs across different users by taking a buyer-side perspective into account as well.
Hex Technologies Facts
- founded in late 2019 by Barry McCardel, Caitlin Colgrove & Glen Takahashi all with a serious background at Palantir and year-long experience inside the data space.
- Seed 5,5M$ round announced 03/2021 “to help data scientists share data across the company” - employee count 9… early customers Glossier, imgur, Pave. Going from limited beta to wider availability.
- Series A 16M$ announced 10/2021 “Help Data Teams Do More, Together” (notice the progress here from “data scientist” to “data team?” - I’m pretty sure the next one will read “Help companies to get further, together, with data!”.)
- Additional customers: dbtLabs, Angellist, Whatnot.
- Announced integration with dbt (Dbt user base currently: roughly 10k companies).
Ok, enough of the facts, let’s dive into a multifaceted analysis using my favorite set of seven frameworks.
0. Data Vision
For a short take on the data vision, read my newsletter #43. The short version is, they have a decently large vision for the space which bets on breaking down the current technological separation of, say, data disciplines into something that is cut alongside business lines.
They place their tool right at this cut to allow sharing between “data teams” which span
- data scientists
- data engineers
- analytics engineers
The platform framework is a simple question: Are they developing an internal, or external platform? For a little bit more details, I got to again refer you to one of my Thoughtful Fridays (#2).
Hex is building an internal platform.
This internal platform requires network effects to work. Meaning the more people inside the notebook collab app, the more the marginal value of the next one is added.
For internal platforms to work, it’s not enough to produce a good product. The story is even more important than for other products. They already bring a fair amount of that.
But if we compare to companies like montecarlo data, which are producing books to evangelize for “data observability” or Dbt Labs which found hole eh disciplines (like “analytics engineering”), there still is a lot more to add to the story.
2. Network Effects
Since Hex is building an internal platform, and currently have no other mechanism to leverage networks effects, they got none.
As far as I can see with a short look, however, it doesn’t seem like any competitor currently has them, so that’s “fine for now”.
Let’s take a look at the value an end-user has from using Hex.
3. Data Value Pipeline
The only value of data, in my opinion, is to let someone, in this case, the end-user, make better/faster decisions.
Now let’s talk about the value from a simple present tense perspective. My data value pipeline looks like this:
- 1a. I can push data to someplace,
- 1b. I can access it from somewhere,
- 2. I can transform data into something valuable
- 3. I can use this valuable thing in an applied form
All four stages add value to the final value of my improved decision.
So where does Hex set in? Not really at 1a or 1b, although I can load some data into hex, and can use it to access my data, there’s no improvement over my alternatives here.
But it’s the combination of 2 & 3 that Hex is right in.
The good news: Combinations mean greater & faster value creation because we’re not locked up by another step in the pipeline. Of course, 3 is limited, I cannot (yet) provide the data as an API, just a “mini Application” - although I would think an API is actually just another way of a “mini Application”.
4. DAKS - Future Value
The next question is about what kinds of future values the company is tapping in. If they are tapping into any of the four forces, the more the better, they are tapping into exponentially growing value.
If you want to get a comprehensive understanding of what I think will be those four forces, listen to my interview on the Data Stack Show or read my article on the future of Good Data.
Amount of data: If we take a step back, Hex does not help us more if we double the amount of data we have. We would also need to double the number of people on the platform to increase the decision output.
Kind of data: It also does not really help us deal with the ever-increasing diversity of data. Here still the bottleneck is the users.
Decentralization of data: However, Hex does help us deal with both, the decentralization of data sources and even more with the decentralization of data usage. After all, that seems to be the point of the tool, to share within a decentralized context of usage.
Snowflake data problem: It also helps us deal more with the problem of data integration because it taps right into a lot of sources and allows us to transform data on the fly in the way that suits us.
So we check 2/4 on the side of the forces of things, setting Hex up to grow the value it can provide for users exponentially as these forces take off exponentially.
5. Datacisions Cycle
I know by now you must think we got the value Hex offers a customer nailed, but I like to take a third perspective on the value a bit more deeply. Let’s take the “datacisions cycle” as a final framework. It goes like this:
- Actions produce -> Data
- Data is turned into -> Information
- Information is turned into -> Insights
- Insights are turn into -> Decisions
- Decisions -> Actions
So Hex doesn’t help us collect data from actions, neither does it help us turn decisions into actions.
Note: Not yet! If the app would also allow for publication of APIs, then we could indeed build small recommendation engines etc. right out of hex. And follow Eugene Yan’s approach, we want to start without machine learning anyways for every machine learning problem.
Hex does focus on the process of turning data into information & insights, possibly even extending to the decision-making phase because it is a great tool for telling stories, complete analyses which help make decisions faster. So it is involved in multiple stages again, more towards the end of the cycle.
Nothing good or bad in here, it should just help understand the dynamics & landscape of integration, because the datacision cycle is complete in every company. So that means other tools & processes take up the rest of the space and the company has to think closely about the integrations that yield the most value for the user at the end.
Ok, now I think we nailed the value as far as we could.
6. 3-Tier Framework
No open-source, no 3-tier framework. For reference, I suggest my article “How to build the next mega open-source project”.
7. Commercialization Strategy
I find one thing about commercialization crucially important, and often lacking in data companies:
- alignment of the value creation (which happens on the customer side!)
- with what a customer actually pays for.
Why? Because commercialization provides incentives to produce products. If you monetize for plain storage size, your products will tend to produce more storage, not less.
It’s been one great insight in Snowflake to align here, and I believe it laser focuses a data company on creating value for the end-user. It also strengthens the competitive situation, as you might have observed with Snowflake.
The problem: We cannot evaluate the pricing, because they don’t have a public one yet!
@ale_lollo87 We're still working on public pricing info. In the meantime, you can shoot an email to email@example.com and our sales team (totally not me wearing a hat and mustache) will get back to you.
So let’s quickly outline what it could look like before they publish it.
A cost-based perspective could lead one to try to price for:
- computation time
- storage amount
- … any other metric
A company size based perspective could lead one to price for:
- number of users
- number of dashboards
- number of sources…
So let’s quickly pause and think about the value perspective here taken from above.
A team & company using hex derives the value from quickly getting from raw/modeled data to a decision. The alternative is no fun - involving long pipelines & development time. So it probably is the actual viewing/sharing/App using what we would want to price from a value perspective.
After all, actually pricing made decisions seems not possible. So we select a good proxy, “viewing of the decision basis”.
A proxy for that could be users, but with the app capability, I would actually prefer to go for actual views. An analysis that is not viewed at all, should cost exactly 0 because it adds exactly 0 of value for the customer. If we price more, the customer will switch to a product that does not price for that.
Next, we take a buyer-side perspective trying to understand different buyers, because as in most B2B contexts, it’s not the end-user that’s buying something like hex.
Hex’s tool is currently targeted at data teams, so we can assume either one team will use it, so a team lead will be the buyer, or 2-3 individuals to share their stuff (no sense in a collaborative workspace if you are alone!), or a set of teams, a department, a smaller unit inside a company.
Following this, a possible nice value-aligned pricing strategy, taking into account a buyer-side perspective could look like this:
If any of the cost drivers actually become significant, you could add XXX minutes/free varying with each stage, then with a “buy additional for XX$”. That way you combine the both of best worlds,
- of value pricing for the information product side of things
- of cost pricing to take care of the competition in the commoditized side of things (storage, compute,…)
I hope this deeper dive into the company Hex helps you to either take a different perspective when you’re thinking about data companies or it helps you to think deeply about where your data company is headed. Maybe this is a complete mess, in which case I’d still love to know your thoughts!
And of course, leave feedback if you have a strong opinion about the newsletter! So?
It is terrible | It’s pretty bad | average newsletter… | good content… | I love it, will forward!
Data; Business Intelligence; Machine Learning, Artificial Intelligence; Everything about what powers our future.
In order to unsubscribe, click here.
If you were forwarded this newsletter and you like it, you can subscribe here.
Powered by Revue