Without experimentation, if we release a new feature and see the following graph, we can infer that with the introduction of a new feature the sales have increased.
But if we could have done experiments, we may found that it is not true. In the absence of this new feature, the sales might have increased drastically, the reason for this can be anything.
Controlled experiments are building block for decision making
How useful is this feature
Is it moving towards my and company goal
How team is performing
Characteristics about experimentation at Airbnb – different in different companies
Data related tools and Infra
They use tableau and superset extensively
How to make sense of this data
It can analyze billions of rows in no time.
Airbnb has a data portal for all the data related work different teams/individual have done. They use the similar tech to google page rank, its kind of search engine.
Data is at the core of everything.
ML is everywhere in Airbnb.
ML mapped to the every step of the process flow