Some Notes


  • #data #a16z #network-effects
  • Basic keystones necessary when thinking about "data moats":
    • Bootstrap a Minimum Viable Corpus to start training your models without waiting for a full training set
    • Think about Data Acquisition Cost. As the CAC will decrease the DAC will rise.
    • Realize that the value of new data decreases as you gather more data (Incremental Data Value) pushing your variable costs higher
    • Data Freshness is more and more relevant for the newest models, and it costs more to collect and keep up to date (variable costs again)
    • Know the Distribution of the Data and you can try to prevent biases
    • Secure data sources: it's easier to copy than to find new data sources. But to bootstrap use other people's open data!

Deep Tech

European Startups - Dealroom report - 2020
  • What about Italy?