Please enable JavaScript.

Coggle requires JavaScript to display documents.

Building DS Team - Coggle Diagram

- - - - Solving Organizational Problem
        
        identify organizational problems as a target to try to hit.
        
        example: increase user growth by a certain percentage by a certain amount of time.
        
        but setting specific targets can be very hard to hit all the time.
        
        be careful to internally calibrate when you didn't hit the target.
      - Solving internal problems
        
        example: data security, cleaning up bad quality code, cleaning up documentations.
        
        targets should be more concrete, definable, easily to hit & solved in a finite period of time.
      - Metrics for success
        
        very concrete
        
        progressing in company direction
        
        be able to prioritize those metrics
        
        very specific metrics not vague metrics
        
        it's a key job of the DS manager
    - - Completion of projects
        
        identifying individuals' projects and once completed documenting as a success is not often that easy because:
        
        involves multiple people at multiple different times on multiple different levels
        
        hard to monitor individual success
        
        easier to monitor day to day activities, or
        
        week to week activities in meetings, participation, contribution
        
        do other people get what they need from them
      - Personal improvement
        
        are they learning new techniques?
        
        can they personally improve and hit goals for themselves?
    - - Take responsibility when there's failure
        
        particularly externally facing, as DS manager
        
        remove some of the heat from the team
        
        let team feel empowered to work the right way
      - Identify problems (examples)
        
        lack of communication with external people talking to the data science team, or
        
        within the data science team itself between the data engineers and the data scientists
        
        team is not working on the problem that needs to be solved
        
        team is working on a variety of different projects and they need to be consolidated
      - Propose concrete steps towards a solution
        
        and people responsible for taking those steps
        
        no-failure policy means no-innovation policy
        
        it scares team off from trying new things, being creative and adding value
        
        failure is not hitting the exact target metric, but not using the right process to achieve it
        
        It's about process and not necessarily about outcomes
    - - Over time this is frustrating and can grind team down
        
        important to make the team known that we did an excellent job
        
        they can brag about
        
        show them how that success is being realized in other areas
        
        how they improved metrics for the organization
        
        keeps team motivated
  - - - Regular
        Updates, Problems & Goals
      - Peer Review
        
        team members make presentations
        
        others empowered & able to comment
        
        make criticism without being rude or mean
      - setting team Priorities and Motivation
        
        remember them with the bigger picture & why the team is doing this
        
        keep everybody excited & fresh on what they're working on
      - mix
        
        Individual meetings to deal with personal issues & personal successes
        
        Group meeting where everybody can come together & communicate
    - - Regular
        
        meet with every member every week
        
        present what they're working on
      - Updates
        
        check what's accomplished
        
        resolve problems that they have
        
        get them unstuck and keep moving
        
        goals for the week
    - - Interactions
        
        Internal: team sits all by itself, talks to each other
        
        external: spend some period of time embedded in other organization group
      - Actively
        
        being Cc'd on messages
        
        involved in the slack channel
        
        going to meetings
      - Passively
        
        indirect reports & feedback from both external group and embedded person there
        
        make sure that communication is flowing with no breakdowns
      - too fast vs too slow communication
        
        too slow: they're not getting enough details back & forth, slowing the progress down
        
        too fast: having an external unit that's making extreme demands on data scientist
        
        know how to modulate those requests taking some of the heat
    - - open door policy: people is able to drop by at any time and ask questions, immediately getting unstuck
        
        allowing quick communication, email or chat, responding to questions:
        
        what can be resolved very quickly without a formalized meeting
        
        team requires large amounts of uninterrupted, concentrated effort
        
        minimize the amount of formal meetings that slow down work
    - - growth not only in number of people but also growth of individual people
        
        learn new Tools
        
        give access to courses, or through presentations in group meetings
        
        introduction to new interactions
        
        identify an external group or a person that makes a good connection with team, facilitate that interaction
        
        identify opportunities for advancement for team
        
        promoted inside organization or externally at conferences or at meet-ups
        
        take on new responsibilities that look good on their resume
  - - - Overview of their position
        
        Expectations
        
        projects
        
        who to interact with, at what time scale
        
        start interactions, or give contact information to start themselves
        
        Establishing Policies
        
        Internal: what's okay & what's not about feedback & review
        
        External: interactions flow through the management , or embedding people briefly in other teams and how to communicate with them
        
        these compressed information should be stored internally on a website, or shared/ distributed PDF document
        
        as a reference for whom to contact with, how to be acting
    - - payroll
        
        benefits
        
        documentation they need
    - - data science team
        
        other teams (external points of contact)
        
        everybody knows the new member
    - - Hardware
        
        computer associated with the activity
        
        database being used & framework
        
        software
        
        available for them to install, what are their purchasing limits
        
        What's their username
        
        How to pull out data
        
        tutorials & introductions
        
        Communication channels
        
        by email or through chat
        
        tool like "Slack"
    - - start them on a small, finite project
        
        Short duration, Achievable goal, Concrete & specific, Relevant to future projects
        
        get a lot of confidence
        
        be integrated into the team