Open Sesame
Is an Open-Source business model viable?
With the COVID-19 pandemic halting many businesses and even changing the organisational structure of many (especially nascent) firms, the use of open-source software has become a key tool in the architecture of many organisations. It is, however, a non-trivial challenge for many organisations that build said open-source platforms as their underlying technologies are being given away for free. So, the greatest challenge is to set up a cost-benefitting structure that allows the flourishing of such organisations.
Establishing a business model around open-source software typically involves a combination of a number of commercial strategies, tailored to fit the specific needs of the firm in question. Apart from an initial fundraising campaign, the use of professional services, such as in the installation and maintenance of structures as well as consulting projects for customised solutions, is thus a central part in the setting up of such business models. Commercial extensions, including installers for pre-packaged distributions, advertising for selling the merchandise and hosting (generally as a cloud-based service) is key as well.
Whilst there are various firm-specific as well as location-specific variants of the above, what all models have in common (albeit to different degrees), is the leveraging of the open-source community for early adoption, essentially crowdsourcing many marketing and presales activities.
The Case of KNIME
And, it is possible – to build an open-source software platform that becomes economically viable. Several firms worldwide have already managed to establish new business models to provide production-ready open-source software.
Consider the case of KNIME (Konstanz Information Miner): an Austrian firm specialising as an open-source data analytics and reporting platform that integrates various components for machine learning and data mining through its modular data pipelining concept. Instead of selling a proprietary version of an open-source software application, KNIME went about the process by creating two separate, but complementary pieces of software.
According to CEO and Co-Founder, Michael Berthold, “I’ve found this allows for a clear division between the open-source application and the commercial offering so that individuals have the typical open source innovation climate, while the commercial software helps the organization productionise their results in a scalable and risk-mitigated way. This approach allows individuals and organizations to stay at the forefront of an innovative field (here: data science) while at the same time productionising what has shown to work. It also provides a solid revenue model for the software provider. I’m happy to say KNIME, which employs 100 people, has been profitable since day one.”
The case of Feast
Given the renewed importance placed on machine learning owing to the pandemic, a major aspect that has been called into focus is raw data, the engine that feeds learning. As a part of the process of feature engineering, i.e. the practice of extracting useful features from raw data that can be used by ML and eventually scaled up and made more complex, the ready availability of a feature store, a tool that automatically manages and serves up features, becomes crucial.
Enter Feast (Feature Store): an open-source ML feature store born from collaboration between Google and Indonesian start-upGojek. First launched in 2019, it was created as an operational data system that acts as a bridge between data engineering and machine learning, by helping to automate some of the key challenges that arise in producing machine learning systems.
Postmates, an American company that offers the local delivery of restaurant-prepped meals and other goods, uses Feast for fraud detection. They benefit by “just being able to publish features to a certain topic and automatically have offline/online stores persist their data for model training and serving.” Agoda, an online travel agency and metasearch engine, also uses Feast as the online serving layer throughout their data centres. Their ML team also values having a standardized definition of features in production and being able to serve feature data to models at scale for online use cases.
The use cases of Feast are several more, and simply goes on to show that with a strong business model, even an open-source software model can be profitable in the long run.
