The primary difficulties are:
- Privacy
- Security
- Intellectual Property
An organization cannot risk streaming actual client data which exposes original data sources.
The cost to implement a viable, safe monetization strategy which mitigates these challenges is resource intensive, and creates a situation where data-oriented business models are the main actors and stakeholders of the actual data market.
Furthermore, each company that operates trained algorithms sit on a goldmine that can be monetized externally from their market.
In the actual paradigm, a firm that wishes to buy or build a structured dataset must find data providers, analyze data quality, and implement data structuring in a unique way, making cross-service creation rigid and linear.
This entire process creates a dominant group of stakeholders such as: Google, Facebook, Microsoft, Yahoo, Uber, etc., which slows the actual development of machine learning based services.