Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Looker & TaskRabbit

Evan Tahler
January 22, 2015

Looker & TaskRabbit

Looker & TaskRabbit
Presentation given at the first "Look & Tell" event in January 2014

Evan Tahler

January 22, 2015
Tweet

More Decks by Evan Tahler

Other Decks in Technology

Transcript

  1. View Slide

  2. Lean BI
    everyone can be a data analyst!
    Looker & TaskRabbit
    Evan Tahler
    Director of Technology
    [email protected]

    View Slide

  3. What is TaskRabbit?
    TaskRabbit is a marketplace where
    people and businesses can go to
    outsource tasks and longer-term jobs to
    fully vetted people in their neighborhood.
    Our Mission: Empower people to do what they love.
    Our Vision: Revolutionize the world’s labor force.

    View Slide

  4. Who are the TaskRabbits?
    There are currently over 20,000 TaskRabbits and growing rapidly.
    All TaskRabbits must complete a 4-step vetting process, including a criminal background check.

    View Slide

  5. What is BI at TaskRabbit?
    In June 2013
    ● 2 data analysts
    ● a handful of engineers who understand the schema(s)
    ● 0 data science expertise
    In Jan 2014
    ● 1 full-time Data Scientist
    ● every engineer familiar with major tables
    ● a data subject-matter expert in every business area
    ● no need for full-time analysts

    View Slide

  6. Where did the Analysts go?
    ● We noticed that our analysts were spending
    the majority of their time creating reports for
    other teams
    ○ ...and that others were spending a lot of time waiting
    for reports
    ● Analysts are valuable, but at our stage we
    needed to focus on tools to enable product
    development.
    ● This is only possible when you can empower
    everyone to become comfortable with data

    View Slide

  7. How can you tell when your team is
    comfortable with data?

    View Slide

  8. Everyone is an Analyst?
    There should be 0 friction between you and
    the data you need to do your job
    ...of course there are times you might need
    help or custom development, but for 90% of
    your day, you can find what you need in
    Looker

    View Slide

  9. Every department now has a subject-matter expert in the
    data for their group, and is trained in SQL and LookML

    View Slide

  10. Looker Spin-Up Timeline
    ● (2-weeks) Looker Analyst On-Site for to learn our
    schema and help build out our key models within looker
    ● (1 day) Looker on-site to train “power users” on LookML
    ● (2 weeks) TR “power users” build out remaining looker
    models
    ● (2 days) Looker on-site to train business users in
    reporting and exploration
    ● (6 weeks) company becomes comfortable using Looker
    for day-to-day activities
    ○ Beware the stages of acceptance

    View Slide

  11. Data Pipeline
    Production
    mySQL
    Analytics
    Replica
    HTTP + Ruby API
    Production
    Elastic
    Search
    Forklift
    (ETL)
    LOOKER
    WEB
    SSH
    Tunnel
    Data
    Warehouse
    Vendors
    (Mixpanel)
    Replication
    ElasticDump
    Ruby + MySQL
    ● ForkLift (github)
    ● ElasticDump (github)
    ● WaistBand (github)

    View Slide

  12. What do we do with Looker?
    Easy view into device adoption & release correlation

    View Slide

  13. What do we do with Looker?
    Single source of truth for financial reporting (from 3 tools to 1)

    View Slide

  14. What do we do with Looker?
    KPI Reporting (from 90 min to 15 min)

    View Slide

  15. What does it mean to become more
    data driven?
    Individual teams & initiatives now have the
    ability to course-correct faster
    Experimentation becomes less risky

    View Slide

  16. What does it mean to become more
    data driven?
    With all team members having the ability to directly “play”
    with the data they care about, trends can be explored.
    Examples:
    ● Moving tasks posted at the end of the month
    ○ Change the distribution of marketing efforts to
    correlate and become more effective
    ● Posters accept one of the first 10 bids
    ○ Change the UI to direct rabbits to bid on tasks with
    less bids, leading to more tasks getting bids

    View Slide

  17. Lean BI
    everyone can be a data analyst!
    Looker & TaskRabbit
    Evan Tahler
    Director of Technology
    [email protected]

    View Slide

  18. View Slide

  19. View Slide