Place in vector space
We mapped Slush24 attendees and created a universe of companies which you can explore. If you are attending Slush, you are probably somewhere there.
Starwatcher is out there, observing the startup universe. We find companies, map their traits in a multidimensional space, and add them to our growing knowledge graph. It’s exciting work, but sometimes the connections between companies aren’t obvious. In our dashboard, we distill everything into one score—a relevancy or similarity number. "Just give me the number and sort by relevancy," people often say. But that single score hides a lot of detail and context. Even a 2D view offers broader insights. But what if we could build an interactive 3D representation? All we’d need is a diverse set of companies with some common ground. And we found it—Slush 2024 attendees: a mix of startups from countless industries, all in one place.
How it works?
Just fly with Zoom in, zoom out or if you are on mobile - pinch;
Click on logos and see their profiles;
Hold “shift” and change the “centre of constellation”;
Search for company hostname and explore neighbourhood of a company.
Go check it and share - https://starwatcher.io/slush24
Data
The only piece of information we start with is a company’s website address. Sure, Slush's matchmaking tool contains detailed company profiles that people have spent time creating, but for our work, we don't rely on them.
We don’t trust humans to provide objective data at a scale necessary for comparing companies. Humans over sell, under sell, miss sell, are subjective, too emotional, too technical, excited, desperate, funny, serious, sarcastic. You know - very human. At Starwatcher we do our own analysis of companies (some AI pixie dust one may say). All we need to begin is a company’s hostname. The equivalent of modern day office for business entity. With this, our system forms an opinion about the company in 20 seconds or less, mapping it into our company universe—a knowledge graph of the business world.
The map
The obvious question is: what are the dimensions of this 3D space? The honest answer is, “I don’t know.” Those dimensions don't make any human relatable sense. The map is created by an algorithm—right now we’re experimenting with tools like UMAP and t-SNE. The picture changes every time, even with the same data.
These algorithms work by taking multidimensional data and compressing it down to something we can visualise in 2D or 3D space. Their job is to make sure similar companies stay close together while also showing the bigger picture of how different groups relate. The ‘map’ they create can change a bit each time because they rely on some randomness and approximations. Still, patterns like clusters of cybersecurity or health tech companies will emerge.
So what are some interesting parts of constellation Slush24?
By far the biggest and most noticeable cluster of companies is “Healthtech hook”. It has this distinguished hook shape.
There is also a peak of games and tail of cybersecurity
Some companies are alone in the space.
There are also two wings with EV charging and material science companies. But they are hard to capture in 2D picture. Just go and explore it!
Whats next?
This is a snapshot and doesn't have latest set of companies. Will update and create new representation probably in upcoming days.
This is startup universe but we do have information also about investors and we might add them to the mix. That is different kind of dimension and perspective on the same data.
Constellation Slush is just a small part of Starwatcher universe. We are integrating this visualisation into platform so you can explore any set of companies.
For the hackers. You may find the source file. It is boring and contains hostname of a company, logo url and 3 coordinates. We have an idea to create some tools around this format. So if you are doing something interesting, let us know. We might collaborate.
Are we going to be in Helsinki for Slush? Kind of. I'm going to be in Helsinki.
Are we raising? Maybe. Shoot me a message at observatory@starwatcher.io
More about our views you can read in Whom should I meet and in our blog series about AI stack - Part I - Quick history of information and Part II - Compression of knowledge.
Follow us - Discord, Twitter, Linkedin
Inspiration
We have been tinkering with an idea about space visualisation of data for quite some time. Big inspiration is Spotify galaxy