Building a Data Practice from Scratch

Investing in a lightweight system now will pay dividends down the road.

My first choice for this tool is Trello, but use whatever integrates well into your organization.

Create communication channels for dataStrive to make all your communication transparent.

Transparency, documentation, and reproducibility go hand in hand.

If you’re on Slack, create a channel for data.

People will email or Slack you directly, but, when possible, move those conversations to open channels on Slack or on your Trello board so there’s documentation of what was discussed.

A valuable by-product of this is others can learn about what you do.

Slack and Trello can be valuable channels for data evangelism.

If you get reports or alerts emailed to you, move them to group mailing lists.

I recently created a data pipeline, and rather than having error alerts emailed only to me, I set up a mailing list dedicated to data pipeline alerts.

This creates a paper trail that others can search, and it enables people to subscribe to alerts on their own.

Start a code repositorySet up a GitHub repo and use it liberally.

Save your queries.

Save your code.

Write comments in your queries and code — for your future self and your future colleagues.

Add your commits to your Trello cards.

You can even integrate GitHub into Trello to handle permalinking for you.

Don’t worry too much about what goes where.

You’ll end up tinkering with the organization of your repo over time.

The key thing is to get in the habit of committing everything and linking those commits to wherever you document your work.

Talk to everyoneSchedule time with at least one person from every team, if not the whole organization.

Ask them how they think about success and whether they track any metrics.

Identify the data sources they use.

Find out if they’re running any reports, and consider creating the reports for them.

While doing this, figure out who is data savvy in the organization — they will be great resources and make your life easier.

I have found that, outside of Engineering, the most data savvy folks at startups tend to be in Finance and Operations.

Of course, you’ll want to get close with the Engineering team.

They will be your partners in determining how data is generated, processed, stored, and accessed.

Embrace yesOne of the core values at Sawyer is “Embrace yes — always.

” In my first 30 days, I simply said yes to every request that came along.

I wanted to get a feel for a variety of projects, work with different teams, and use that experience to build my understanding of the business and its data ecosystem.

I also wanted to gain the trust of my coworkers, proving myself to be a willing collaborator before showing my skeptical side.

After 30 days, I started the hard work of clarifying and prioritizing, asking questions like, “Why do you need this?.What will you do with it?.How will it impact your decisions?” I pushed back on some requests.

I looked for faulty assumptions, sources of bias, errors in logic.

Was this the right approach?.Frankly, I don’t know.

I could see it backfiring, especially if you can’t switch gears to questioning mode.

My plate filled up quickly by the end of the 30 days, and it’s hard feeling behind in a job you’ve just started.

Ultimately, whether this is successful is dependent on you and your organization.

Build your networkYou may be the only data person at your startup, but don’t feel you have to go it alone.

Work on building your network of other data practitioners so you have a place to go with questions, to get feedback on your ideas, and to keep up with the field.

There are great communities out there to get you started.

One I recommend is the Locally Optimistic Slack group.

If you find yourself in this position, congratulations!.It’s an exciting time to be at a company.

The work you do now can have tremendous impact.

Don’t let that stress you out though.

Take an iterative approach, trying things, reflecting, and changing course when necessary.

Remember that documentation, transparency, and reproducibility are key principles, and they will pay dividends down the road as you lay the foundations of a successful data practice.

.. More details

Leave a Reply