Erika Pullum (she/her)

We're looking for someone to join our lovely team at Data Orchard as an Analytics Engineer! This is a really diverse and exciting role working on a wide range of #data projects for our wonderful nonprofit clients and our own data products. Please do apply if this sounds up your street 👩‍💻 #dataSky

This is fun and I haven't seen it before! datavizcatalogue.com/index.html

#dataBS

The biggest challenge at a startup that's growing quickly is how often your job changes. What matters and where your focus needs to be is always changing.

I wish there was a way for the dbt Cloud Explore DAG to filter to content that exists *between* two nodes. Anyone know of a fun work-around for this with other tools?

#dataBS

What's on my mind this week (as Head of Data at a growth stage startup)

* Hiring 🔜
* Upcoming vendor contract negotiations
* Priorities pulse check with our stakeholders
* Couple lil IC tasks to standardize a metric a bit better

#dataBS

Naming is hard.

#dataBS

Perpetually flip flopping on self serve analytics

Some days it feels like it could be the holy grail - enabling downstream users to do their own exploration and analysis on governed metrics. Fewer ad hoc requests! Reporting automation!

Other days: So what? Does that actually move the needle?

The questions to decide are:
* Where is the inflection point where ROI and effort are aligned -- aka how deep to invest
* For which functions should it be prioritized
* What does the support model look like
* What is the process for getting more help when needed

As a leader, you can't ignore it and it's impossible to universally achieve. When making consequential decisions stakeholders want consensus and a partner.

When assets exist you still have to teach new people what they're called and where to find them. Metrics have to keep up with the biz.

Don’t think of it as whether the data has error or not. Most of the time it will.

Think of it as whether the error in the data will cause you to make a different decision than if the data was perfectly clean. Most of the time it won’t.

One of the biggest tips I have for anyone doing data analysis, especially data from people, is to spend some time drilling down to the most granular data and just looking at individual records. You will find the craziest shit you never imagined and your analysis will be better for it #databs

My #dataBS new year's resolution is to stop feeling guilty when I tell someone to go self serve their QQ. What's yours?

Dear Santa, All I want for Christmas is for any of the abandoned GitHub repos that half-implemented the Hemingway writing app as a VS Code extension to finish the job. Please put your best TypeScript dev elf on this 🙏

Watching Hex's internal hack week demos is like Christmas early 😍

Has anyone cracked the code on the most efficient way to align on metric definitions with your customers - key word being efficient?

I'm going to cite this as evidence of how great our team is next time we get to hire, watch me.

turning your pain into art I see 😆

a poem by poetry:

poetry run dbt compile
command not found: dbt

poetry show | grep dbt
dbt 1.0.0.38.22

pip freeze | grep dbt
dbt==1.0.0.38.22

which dbt
dbt not found

(look, no one said it was a happy poem)
#dataBS

Which is a bummer because unless you carefully guard the inputs the recursion can run away forever!

Fascinating! Confirmed this works in DuckDB, but would need some concepts ported to Snowflake to work in our warehouse. Nice work on this! #dataBS

Wait I found a solution in pure SQL using list transformations (works in DuckDB): gist.github.com/aranke/74206...

It nerd sniped our whole team in the best way! We used a recursive solution!

The Boolean is the correct answer! The challenge is to create it. I haven’t found a way to do it in SQL with lag functions but I’d love to see one!

And full confession I did not solve it

to make the base
select '2024-06-10' as d

union all

select '2024-08-20' as d

union all

select '2024-08-22' as d

union all

select '2024-09-17' as d

union all

select '2024-09-19' as d

union all

select '2024-11-01' as d

union all

select '2024-12-11' as d

union all

select '2024-12-21' as d

A table of dates. 2024-06-10, 2024-09-17, and 2024-12-21 are labeled true. 2024-08-20, 2024-08-22, 2024-09-19, 2024-11-01, and 2024-12-11 are labeled false for is_after_cooldown

Super #SQL brainteaser for your Tuesday that kept a few folks on our team thinking pretty hard until someone made an elegant solution.

The desired behavior per unit:
* First event qualifies
* Subsequent events qualify only if it's been more than 90 days since the last

Sample dates below #dataBS

A response from github user caitlinmoorman that says I.love.it.

aaaand it worked

A PR title for #1607 that reads `Erika/delight caitlin`

Continuing my theme of banger PRs -- this one improved documentation for a field confusing to a lot of folks

#dataBS

Just had a great chat about being the first data person at a startup.

Pros
* Building at high velocity is fun
* Work with smart people

Challenges
* Always more than you can do
* Hard to know how good to build when

Folks that have done it, what's on your pro/challenge list?

#datasky #dataBS