Leave a comment
Get the GH Bookmarklet

AMAs

Dr. Jessica Kirkpatrick is a product data scientist at Slack. Previously she has worked at Hired, Chegg, and Microsoft. 

Before making her transition to the private sector in 2012, Jessica earned a Ph.D. in Astrophysics from UC Berkeley, where she studied some of the most distant and brightest known objects in the universe, quasars, and worked with large and complex data sets. Today, instead of spending her days finding patterns in the structure of the universe, she spends them finding patterns in the behaviors of people in order to make technology work better for us all.

In addition to her work in private industry, Jessica maintains an active role in giving back to the astrophysics community. She is a blogger for Women-in-Astronomy, Astrobetter, Women 2.0, and Lady Paragons. She is also on the Board of Trustees for the American Astronomical Society and served on their Committee for the Status of Women in Astronomy for 4 years. 

You can follow her on Twitter: @berkeleyjess.

She will live on Nov 28 starting at 930 AM PT for one and a half hours during which she will answer as many questions as possible.

  • JK

    Jessica Kirkpatrick

    9 months ago #

    Excited for this conversation!

  • SE

    Sean Ellis

    8 months ago #

    Thank you so much for doing this AMA. What would you say should be the main role of data science in helping to accelerate growth?

    • JK

      Jessica Kirkpatrick

      8 months ago #

      The main ways that growth and data science interplay comes down to being able to use data to understand what’s working and not working in your product, and overall optimize the experience so you’re focusing on the right things to improve. This could entail identifying areas in your sign-up flows that are losing people or acquisition methods that are the most effective; determine the markets where your product is really taking off so that you can focus on the most fruitful areas. This will give you deeper insight into what’s going on with your customers and where to focus your business efforts to be the most effective.

      4 Share
    • JK

      Jessica Kirkpatrick

      8 months ago #

      A few more thoughts about data and growth from some other members of my team at Slack who have worked more closely with the Growth Team:

      • At Slack, we know that we’re not the customer. We use our product differently -we’re power users and we’re bought in. That contrast is more extreme in growth, where you’re dealing brand new customers - these are the people where your intuitions and expectations are most off. If you treat data as the voice of your customer, growth is the most valuable place to use data to develop empathy.

      • If growth is about acquiring new customers, retaining them (and if there’s a paid portion to your company, converting them), data can help with all of that. It’s more than just identifying the stumbling points in the signup flow (how valuable is the incremental user who was lost because the copy was less friendly). Instead, knowing what leads to initial customer success and experimenting to determine causation from correlation to better understand your product and prioritize development is where data can provide the most value.

      • There’s also a school of thought that if you have good PMs and vision, there aren’t *that* many huge wins on the table, and so growth is a game of racking up 50 small (~1%) wins, not a few large (~50%) wins. Data helps measure wins to allows you to double down on success. It helps scope reach and potential impact so you invest in the right products.

      • AZ

        Anthony Zepeda

        8 months ago #

        The second bullet is huge. I'm running through data right now to see the differences between in number in each stage of the buyers journey. I'm hoping it will identify stumbling point or if we need more leads for the top of the funnel.

  • MK

    Meredith Kelly

    8 months ago #

    How do you collaborate with other teams at Slack (i.e. engineering, growth, etc.)?

    • JK

      Jessica Kirkpatrick

      8 months ago #

      I work on the product data team and so I am mostly collaborating with the engineering teams who are developing the Slack product. I help them understand the impact of the work they are doing by helping product teams define goals and metrics and building reporting for them to monitor against those metrics. I help product managers determine if we should run an A/B test for a particular change, and then work with them to properly implement that test and analyze the results. I also partner with our user research and surveys team to look at behavioral data to determine if we can learn more about the trends they are seeing in their research or highlight new areas of research we might want to pursue.

      • AA

        Anuj Adhiya

        8 months ago #

        re: " I help product managers determine if we should run an A/B test for a particular change"
        Can you elaborate on how this process works?

  • MC

    Max Caldwell

    9 months ago #

    What are your favorite tools and techniques for exploratory data analysis at Slack?

    • JK

      Jessica Kirkpatrick

      8 months ago #

      At Slack, I mostly use Hive and Presto to directly query our database. We build dimension, roll-up, and aggregation tables in Airflow. We have an internally-built tool that we use for reporting and data-visualization that is similar to Mode Analytics. I do additional modeling, analysis, and visualizations in a Jupyter Notebook Analytics or Excel spreadsheet depending on the complexity of the analysis and who will be consuming the work.

      4 Share
  • CF

    Claire Fox

    8 months ago #

    What are your top tips for getting hired on the data team at Slack?

    • JK

      Jessica Kirkpatrick

      8 months ago #

      There are many different data teams at Slack. Each team is looking for slightly different skillsets. We have data analysts, data engineers, analytics infrastructure engineers, analytics tool engineers, and search learning and intelligence (SLI) engineers, all of which are considered part of data teams at Slack. All of these teams are looking for slightly different skillsets from machine learning, to statistics, to python, to software engineering. In terms of getting hired, the tips I would give are not particularly Slack specific.

      First, read over the job descriptions and highlight in your cover letter and resume how your past education and experience is relevant to the responsibilities and requirements of the role. Do some research on Slack the company by reading our blogs to understand the projects we are tackling and the values of our organization. If you haven’t used Slack before, create a workspace and familiarize yourself with the product.

      In the interview be able to articulate what specifically makes you want to work at Slack. Be flexible in interviews and incorporate feedback or tips that the interviewers are giving you, we are trying to help you, not trip you up. Try and have fun during the process, we aim to give you questions that are relevant to the role you would be performing, and so hopefully the process is at least somewhat enjoyable as it will be similar to your job (if hired) at Slack.

      2 Share
  • CA

    Camille Acey

    8 months ago #

    How should a company prepare to bring on a data person/build out a data team?

    • JK

      Jessica Kirkpatrick

      8 months ago #

      First, you should define your data needs. Since data science is multidisciplinary, the specific skill set and responsibilities of individuals filling these roles will vary by the size of the organization and their data strategy and plans. For example, an organization with a large investment in data and analytics might build a data science team consisting of several discrete roles/subteams, such as data engineering, data analytics, data architects, algorithms/modeling/machine-learning, and data tools developers. At a smaller organization, a data scientist might take on a more generalized role. A good first step is to assess your data needs and define a data science role that makes sense within the context of your organization.

      Since data science is still a developing field, it can be challenging to find all the required skills to meet your data needs in one person. You can more easily find people with either analytical/problem-solving skills or data engineering/development skills than a candidate with both. Consider hiring two people to fill those needs or a more analytical data scientist who will have a backend/data engineer who can do some of engineering/infrastructure work for them.

      Too many companies prioritize overly specific technical skills, such as knowledge of a particular platform or programming language. At Slack, I look for individuals with experience in data analysis and statistics, but I’m less worried about specific technical knowledge. Technical skills like Python and SQL are easier to teach and can be learned more quickly than critical thinking and problem-solving skills.

      A data scientist must be able to break down a complex mathematical analysis using language that someone from a different background could easily understand. So, in addition to analytical, coding, and problem-solving skills, I also focus on "softer" skills like communication, project management, and prioritization, which can sometimes be harder to come by in a candidate.

      It can be challenging to find a data scientist with many years of industry experience, especially if you are at a small startup without much brand recognition. I recommend considering hiring people who come from academic/scientific backgrounds, which is actually my background. There are programs like Insight Data Science Fellowship which help people transition from academic science to data science, and I have hired quite a few people out of that program.

      2 Share
  • OQ

    OJ Quevedo

    8 months ago #

    How do you know which set/s of data to prioritise?

    • JK

      Jessica Kirkpatrick

      8 months ago #

      Determining which projects to work on, prioritize, and how to most effectively do an analysis is the real challenge of my job. I use Jira to help me manage requests and prioritize my work. This also allows transparency into what I am working on so that requests from different parts of the company can be balanced against my current roadmap and priorities are easier to determine.

      I first attempt to understand the reasons behind why a person wants a particular analysis done, data pipeline built, or question answered. I make sure that the work is actionable (going to be used to make a decision) versus just something that someone is curious about but really won’t influence what they do next.

      If I decide to prioritize a particular piece of work, I work on first producing the simplest, fastest, model/analysis I can (Minimum Viable Product) to get a ballpark idea if the outcome will be interesting or not. Often you can get to a 90% correct solution with 20% of the work it would take you to get to a 95% correct solution. Then, depending on how precise the analysis has to be, I will improve and iterate on my MVP.

      Being able to determine which projects will be the most impactful, and the fastest way to get to a solution is something that tends to set apart more senior data scientists from more junior people. It is an intuition that you build with experience and from working on several different types of projects.

      3 Share
      • AA

        Anuj Adhiya

        8 months ago #

        re: Often you can get to a 90% correct solution with 20% of the work it would take you to get to a 95% correct solution
        Can you illustrate this further with a real-world example from your experience?

  • HK

    Haad Khan

    8 months ago #

    1). How do you apply new research that your team may not be familiar with. How do you come up with ways to sell those approaches?
    2). Sometimes there are aspects about data, models that you find really interesting but its tough to explain to team members and managers why that is interesting. What do you do in that case.
    3). What are your tips on learning something quickly?

    • JK

      Jessica Kirkpatrick

      8 months ago #

      A1) If I learn about new ways to approach problems that I think might be useful for my team, I will try them out on a small/quick application to prove the value and determine if it's worth implementing more widely. I haven't really encountered resistance to applying new methods and techniques with this approach.

      A2) It’s not enough for data to be interesting. It has to have a business impact. Often when I’m trying to convince other people on my team or in my organization that an analysis is interesting, I try to position it in a way that shows how this will improve the product, the experience for our customers at Slack, or the bottom line of the business. If that isn’t the case, then I probably shouldn’t dedicate much time to working on that particular problem.

      A3) I have a learning disability which impacts the way I process certain information. Because of that, I’ve thought a lot about how to learn effectively.

      For me, it’s important to approach learning in a multi-faceted way. I learn best by working through problems and discussing them with people. For example, at Slack, I had to learn a new ETL system and I did this by pair-programming with one of my team members. I immediately applied this learning to a project where I could practice what I just learned. Slack as a tool is really helpful for me because it allows you to search for other conversations within your organization about a certain subject and identify whom might be able to help you solve a problem. When I need to know a certain piece of information, I first search via Slack to figure out whom I could reach out to within our organization to get more context and an understanding.

      3 Share
  • AD

    Alex D

    8 months ago #

    This is exciting!

    * Top 3 data science blogs you personally read (or think everyone should read)?
    * Tell us about your favourite project you've worked on since your transition to the private sector.
    * Imagine a person knows nothing about data science. Where does she start?

    Thank you :)

    • JK

      Jessica Kirkpatrick

      8 months ago #

      A1) The blog that first got me interested in data science was the OkCupid Data Blog, OkTrends. I thought it was fascinating to be able to gain insights into human behavior by looking at large-scale trends in the way people interacted through digital products. Christian Rudder has since turned this blog into a book called Dataclysm, which I highly recommend. If you’re interested in learning how our team specifically works with data, I’d encourage you to visit Slack's Engineering blog and read Data Wrangling at Slack.

      The blogs I read today help me to understand the different approaches people in my field are taking to the business problems that I’m trying to tackle at Slack and the technical challenges that come with dealing with large datasets. AirBnb has a really great data blog that I enjoy reading.

      A2) At my last job at Hired I published several papers about work I did on wage inequality, which is a topic I’m personally passionate about and one of my favorite projects I worked on at Hired. At Slack, some of things that my team is working on now include Shared Channels and Localization, where we’ve launched Slack in French, German, Spanish, and most recently, Japanese. We’re constantly evaluating how customers are responding to these new features and continuing to think of new ways to iterate and improve them for our users.

      A3) There are a wealth of free online courses where people can get an introduction to the concepts of data science, such as Khan Academy, Coursera, and Udacity. I also have written a lot about my experience as a data scientist and how to break-into the field on my blog. Here you’ll find the tips and advice I’ve shared from my own experience and learnings.

  • VB

    Victor Borda

    8 months ago #

    Very exciting!

    It would be great to hear about the lifecycle, pipeline, and environment you use to release and maintain data models and algorithms.
    1. How do you approach comparing newly tuned models to existing models?
    2. How do you handle versioning data models?
    3. What kind of process do you use for developing new models? (As in, say, if you have a new model idea, spend 1 week on it then present results to team before proceeding - in other words, checkpointing and team input approaches).

  • ME

    Martin E

    8 months ago #

    Jess,

    We've been tweeps for several years now. I've always been inspired by your #AstroSH stories. As a man working in STEM, your stories have opened my eyes and you've inspired me to get more educated on this topic. It humbled me when I realized that I had so much to learn from you and others in terms of what goes on in (often right in front of our eyes) in academia and workplaces. Considering the #MeToo campaign right now, would you care to share a quick story or two of your experiences to increase awareness to those of us who've been ignorant of this important issue?

    PS- You're honestly one of my heroes and I sincerely hope my daughter sees you as a role model for herself in the same way I do for myself.

    • JK

      Jessica Kirkpatrick

      8 months ago #

      Hi Martin! Thanks for always being a great ally.

      As I've shared before, I’ve had plenty of experiences with sexual harassment during my time in academia and also once I transitioned to the tech industry. One example was the time when I was conducting a phone interview of a candidate, and a coworker thought it would be funny to distract me by pretending to make obscene gestures in front of the glass-walled conference room where I was sitting conducting the interview. When I didn’t react to his gestures, he pulled over another coworker, pressed him up against the glass and they pretended to be having sex. These type of sexualized “jokes” were commonplace at one of my previous companies. It frustrated me because not only is it unprofessional to engage in such behavior, but it made it hard for me to conduct the interview and I worried that the candidate was not receiving a fair experience because I was upset and distracted by my coworkers’ actions.

      One of the reasons I decided to join Slack is because its reputation preceded it as being an equitable and respectful place to work. Slack is one of the more diverse tech companies with 14.5% of the engineering team being black or Latinx and 43% of managers identifying as female. In fact, Slack is the first tech company I have worked at where I have a female manager. Slack has employee resource groups for people of color, individuals identifying as LGBTQ+, veterans, women, and people with disabilities. As a queer woman with a disability, I have really appreciated connecting with coworkers who share my various identities, and it has helped me feel comfortable to be my authentic self at work.

  • AL

    Arsene Lavaux

    8 months ago #

    Bonjour Jessica,

    Merci for doing this AMA.

    Which signal processing theories used in Astrophysics to study plasmas or else most relevantly apply to scaling a B2B digital experience in your opinion?

    Merci encore!

  • FO

    Felix Maximiliano Obes

    8 months ago #

    What are the biggest challenges, present and future, for data scientists?

  • BI

    Benjelloun Ibrahim

    8 months ago #

    Hey Jessica,

    Thanks for sharing your experience with us.

    Wich book would you recommend me for someone who want to start from zero in data scientist field?

    Thanks in advance,

    Ibrahim

  • EF

    Ed Fry

    8 months ago #

    Thanks Jessica for doing this -

    I'm curious to hear how the data team has interfaced with other teams in your roles at Slack, Hired, and others. Do you work as a support function for different teams, or have your own separate objectives?

    Is there any one team you tend to spend the most time working with?

    Is there much collaboration between multiple teams on data science projects? For instance between marketing/growth, data, and operations?

  • AD

    Alesia Dubrovskaya

    8 months ago #

    Hi, Jessica!

    a. What's your favorite scientific book, that changed the way you think (easy to read)
    b. And your favorite digital tools (anything for digital marketing or Data)

    Thank u
    Greetings from Belarus😚

  • AA

    Anuj Adhiya

    8 months ago #

    So cool to finally have you on, Jessica!

    Should all business decisions be data-driven?
    If not, is there any correlation to stage of company or type of decision when this should not (or need not) be the case?

  • GH

    Glen Harper

    8 months ago #

    Thank you joining us today, Jessica.

    Is there a way we should be asking questions of an analyst to get actionable answers?

  • JP

    John Phamvan

    8 months ago #

    Hi Jessica

    a. What tools does the team at Slack use for experimentation & analytics right now?
    Have any new tools been added to the stack recently? if yes, why?

    b. Where does your data live, ie, what is the universal source of truth?

    c. Other than Slack (obviously :)), does your team use any other collaboration tools?

    Thanks!
    John

  • MD

    Mark Anthony de Jesus

    8 months ago #

    Hi Jessica,
    When analyzing an experiment, is there a way to tell that the test results are a reliable predictor of future performance for any specified time frame?
    If yes, how would you know that at the time of analysis and/or after the fact?

  • TN

    Tri Nguyen

    8 months ago #

    Hi Jessica
    Do you have any favorite analyses that are relatively easy to do but can pay big dividends?

  • DH

    Dani Hart

    8 months ago #

    Hey Jessica!
    Do you have any tips for people who are presenting data to their teams?
    Is it best to start with a conclusion and show the data that supports that conclusion or is it better to work your way up to a punchline? Or some other approach?

  • JF

    Javier Feldman

    8 months ago #

    Thanks for doing the AMA, Jessica.

    What advice would you have for a small team with limited bandwidth about prioritizing building new data infrastructure and reporting?
    How did you think about it as your team scaled, and any advice on where to focus first and/or how to ultimately get both done?
    Any thoughts on managing data priorities would be awesome.

  • GH

    Gabriel Hernandez

    8 months ago #

    What tool do recommend for data analytics, focused on user activity? To know where and what they are doing in the platform.

  • AP

    Aniket Pasalkar

    8 months ago #

    integration of tools with Slack is possible? if yes, What are those?

  • AA

    Angela Agurto

    8 months ago #

    hey there! really interesting, thanks a lot for sharing!

Join over 70,000 growth pros from companies like Uber, Pinterest & Twitter

Get Weekly Top Posts
High five! You’re in.
SHARE
50
50