As someone who writes software for data scientists (https://github.com/snowplow/snowplow) I definitely agree with his analysis. But I would go further: without _domain knowledge_, a data scientist is really just a ETL guy who cleans up big data for the real analysts to make sense of. Applying the whole toolkit to a specific domain (and SaaS B2B looks totally different from supermarket loyalty schemes and from F2P mobile games) is key.