Denver, CO

47.81°F

12:00pm
/ Cambridge, MA

50.08°F

12:00pm

Konvoy Ventures is a thesis driven venture capital firm focused on the video gaming industry. We invest in infrastructure technology, tools, and platforms.

Our Investment in Stagezero

ML Data Labeling with Games

Games are about fun, learning, and exploration; but could they be about value creation as well? One thing we’ve been spending time on at Konvoy is the idea of people creating value when they play games. Over the past year we’ve started to see elements of this emerge within NFT gaming economies like Axie Infinity (portfolio company) and others. We also think external value can be created in games through appropriate interfaces.

Almost all games have some core task element. You have a goal, and you must complete a task to achieve that goal. Gamers understand it, and love it. It’s why a lot of people play games; for the enjoyment of the challenge, of the journey and the satisfaction of accomplishment. How can we leverage that desire to do something that is not just valuable for players (what is achieved today), but also accomplish some form of positive externality?

With an appropriate API layer to games, game economies and players can interact more seamlessly with the real world and other areas of commerce. The first API layer we’ve seen of this nature is in-game mobile advertising. Along with in-app purchases (IAPs), it’s the lifeblood of today’s mobile gaming industry. Ad consumption by players is wonderful, but there is no value creation that happens within the game.

Earlier this year we met the Helsinki-based, Stagezero team and they introduced us to their concept of ‘microtasks’, small tasks that can be delivered into mobile games with their API and Android/iOS SDKs. As Descartes (and many others have subsequently said), “Divide each difficulty into as many parts as is feasible and necessary to resolve it.” Or in computer science terms, decomposition.

Stagezero delivers small, repeatable bite-sized pieces of work to players within mobile games. Instead of gamers being forced to watch ads they are given agency to complete a task within the game - a more natural transition than breaking up gameplay with ads. These microtasks usually come in the form of identifying and tagging a picture or providing some small translation. This data labeling is a key ingredient in building machine learning models, albeit an arduous and time intensive one. From this idea, Stagezero has created a two-sided marketplace; on the one side data science teams can quickly get data accurately tagged from a diverse base of labelers, and gaming companies can leverage an alternative revenue model that is more engaging and profitable than traditional mobile advertising. Microtasks today do take the user out of the game experience, but there is a future where they are incorporated into the game design, art style, and even lore to make it part of the holistic experience. Where gamers can do work without even realizing it. 

The team has even seen some interesting use cases as a payments platform. One of their partners leveraged Stagezero to offer a way for low-income users to complete microtasks and build up credit to pay for the service. Stagezero is working on opening up the distribution of work; I can’t imagine what incredible opportunities for businesses and people will be possible when these frictions and barriers are torn down.

The team is innovating on some already proven concepts in the real world today. One inspiration this may remind you of is Captcha and ReCaptcha, where you can prove that you’re a human to online services by selecting which images out of a number of options are bicycles, or cars, or traffic lights. Others like Mechanical Turk from Amazon offer a basic marketplace of task workers for businesses to leverage. These are successful, but they struggle to scale because tasks are not well defined enough to automate distribution and they don’t have a large enough scale of users. By restricting the type of tasks that Stagezero supports and delivering to potentially billions of users (the number of mobile game players), Stagezero can offer a fast, efficient, and scalable solution.

Through this scale of mobile gamers, Stagezero has the unique ability to address the issue of diversity in ML model creation. ML models today often rely on uniform data or data that has not been tagged by diverse labelers. Data science teams struggle to find appropriately diverse labelers independently or even through data labeling platforms because most platforms employ labelers in fixed regions. Stagezero can help solve this with the scale, breadth, and the diverse user base that mobile games have.

We recently led Stagezero’s Seed financing; we’re excited to partner with Thomas, Nicklas, Patrick and the rest of the team as well as existing and new investors (Ludus, Hyperamp [the investment arm of HUMAN Protocol], Into Ventures, Nordic Game Ventures, Alexis Bonte). We strongly believe in the opportunity in front of them to change the way mobile games monetize, help data scientists build more robust and diverse ML models, and open up distribution of work to more workers.