ideas for teaching/mentoring

[epogrebnyak]

Intro:

1. excercise vs product
2. data engineering vs EDA vs model vs business use / actions
3. corporate info disclosure: sources + data model + gaps
4. code or 'point-and-click', reproducible code
5. raw or vendor datasets

Corporate disclosure:

• who provides what information to whom and why
• Russian corporate data
• role of data providers (Interfax SPARK, PRIME Bir, etc)
• good disclosuse at source (government service API) + own parsing + vendor GUI or API

Dig in boo dataset:

• data access excercises
• understanding data structure
• refreshing accounting and corporate finance
• constraints, data quality

Dig into problem-solving:

• what decisions or actions someone needs to take (the product)
• solution: "miracle" vs pipeline
• face the client + know your resources
• replicate known products or design your own (see "mircale")

Risks and pitfalls:

• data product success not guaranteed - are ready for this?
• "there is always a through-away" (F. Brooks)
• failing on right process is better than just failing
• need to manage risks (iterate, refine, update expectations, manage scope, face the client)

Notes:

• corporate datasets are among most popular Rosstat open data