Week in my Life as a Data Scientist in San Francisco

Watch on YouTube

Now Playing

Transcript

190 segments

0:00

Heat. Heat. [music]

0:10

[music]

0:15

[music]

0:31

>> [music]

0:38

[music]

0:42

>> Heat. Heat.

0:49

>> [music]

0:58

>> As a [music] product data scientist, I'm

1:00

wearing a lot of hats. Running

1:02

experiments, strategic analyses, [music]

1:04

tracking feature launches, defining

1:06

metrics. It depends on the question the

1:08

business has. That's why I set some time

1:10

aside on Monday to plan [music] out the

1:11

week and make sure I'm on top of things.

1:16

Yeah. So today I'm going to [music] I

1:18

typically have quite a few meetings as a

1:21

big part of my role is collaborating

1:23

with product managers, designers,

1:24

[music] and engineers to decide which

1:26

features to build and how to measure

1:28

their success. I'm essentially the data

1:30

representative on the team, owning all

1:31

the analysis and communicating insights

1:34

back to stakeholders.

1:37

>> [music]

1:38

>> After the meeting, I headed downstairs

1:40

for lunch, which was a Thai grilled pork

1:42

chop. And I'm eating with some fellow

1:44

co-workers on a sister team.

1:53

After lunch, [music] I spent some time

1:55

prepping for an upcoming experiment,

1:57

which is essentially where we test two

1:59

versions of a feature and collect data

2:01

to see which performs better. Before

2:03

launch, I need to make sure the data

2:04

logging is set up correctly, calculate

2:07

how long we need to run the test to get

2:08

reliable results, and decide which

2:10

metrics are most important to track.

2:21

>> Wow. [music] Literally,

2:27

you guys are

2:30

so fun. [music]

2:43

>> [music]

2:49

[music]

2:54

[music]

2:54

>> Today I have my weekly one-on-one with

2:56

my PM where we align on priorities for

2:59

the week and discuss any decisions I can

3:01

help unblock with data. I usually also

3:04

share updates on the tests we're running

3:06

and use the time to brainstorm theories

3:08

on why certain metrics are moving.

3:14

[music]

3:15

>> Hi.

3:18

>> For lunch, I caught up with some friends

3:20

from my new grad cohort, which I still

3:22

hang out with till this day.

3:26

Normally, I get a little caffeinated

3:27

pickme up in the afternoon. Jasmine tea

3:30

for a chill work session or black tea if

3:32

I'm more desperate.

3:36

After lunch, it's back to work. I'm

3:38

usually running several experiments

3:39

[music] at once. For one of the tests

3:41

that just ended, we saw multiple metrics

3:43

move, including some we weren't

3:45

expecting. I'm doing some additional

3:47

analysis to understand why and to build

3:49

a strong story for whether the feature

3:51

should be launched.

3:55

[music]

4:01

>> [music]

4:02

>> I found them. These things are so good.

4:05

I'm getting two. Okay, I'm back home.

4:08

Groceries have been shopped [music] and

4:10

now I really need to eat something right

4:12

now. Like, I need something in my

4:14

stomach yesterday. You know what time it

4:16

is? [music]

4:27

>> [music]

4:32

[music]

4:37

[music]

4:37

>> Looks so delicious.

4:52

Today is my favorite day of the work

4:54

week because it's no meeting Wednesday,

4:56

which means [music] I can get some

4:58

precious focus time in. Today, I'm

5:00

scoping a strategic analysis for a

5:03

bigger initiative. These projects

5:05

usually start out pretty vague. For

5:07

instance, we might know we want to

5:08

improve one part of the product, but how

5:11

exactly? [music] My analysis will help

5:13

visualize where customers are dropping

5:15

off and where the biggest opportunities

5:17

are. I find this part of my role most

5:19

fulfilling because it's where I get to

5:21

use data to directly shape product

5:22

[music] decisions. A unique blend of

5:25

analytics, business intuition, and

5:27

storytelling.

5:30

For lunch, I'm making some spicy cold

5:31

noodles. One of my go-to work from home

5:34

meals that I of course learned from Tik

5:36

Tok. Basically, the criteria for a work

5:38

from home meal is that it needs to be

5:40

ready in like 15 minutes tops, or else I

5:42

will door dash. I'm just trying to feed

5:44

myself so I can last the rest of the

5:45

workday at this [music] point.

5:55

[music]

5:56

So, I actually heard that IKEA of all

5:59

places makes a good work spot. So, after

6:02

lunch, I headed over to check it out for

6:04

[music] myself. And lo and behold, it

6:06

actually was a good workspot and I got a

6:10

lot of work done. For the rest of the

6:12

day, I continued scoping the analysis,

6:14

gathering the most pressing questions

6:16

about a feature with high drop off

6:17

rates. I also started digging into the

6:19

data, [music] pulling highle usage

6:21

stats, and noting the key tables for my

6:23

analysis.

6:30

[music]

6:36

And then also bring up our friend.

6:41

[music]

6:51

[music]

6:59

I started the morning by checking in on

7:01

a test that just [music] launched. When

7:02

we run experiments, it's important to

7:04

monitor for any enrollment imbalance

7:06

[music]

7:06

between the treatment and control groups

7:09

and make sure everything is being tested

7:10

[music] correctly.

7:15

So, [music] this test I think still has

7:18

2 weeks left. After that, I had a few

7:20

meetings around sequencing upcoming

7:22

experiments where I weighed in on

7:24

whether we can run two experiments in

7:26

parallel without them conflicting with

7:28

each other.

7:32

Hi.

7:32

>> Hi. What's up? [music]

7:34

>> For lunch, I'm catching up with a fellow

7:36

data scientist who happened to also be

7:38

in my intern cohort.

7:41

[music]

7:44

I ended the day with an ad hoc request

7:47

to pull a quick stat on average usage

7:49

for one of our key features. [music]

7:56

>> [music]

8:02

[music]

8:04

>> Giants. Giants. You like baseball?

8:07

>> Yes. Especially of a certain player. He

8:10

likes the Korean guy. Jungu and Lee.

8:13

Yeah.

8:17

[music and cheering]

8:20

Oh my god.

8:24

Oh my god.

8:30

[music]

8:33

>> Four letters. TGIF. I have a pretty big

8:36

meeting this morning about an analysis I

8:39

did for a test that just wrapped. So, I

8:42

did the write up. I sent it to a group

8:44

of other data scientists to review. And

8:46

now we're going to discuss the results.

8:48

and I have to answer any questions. Wish

8:50

me luck.

8:55

Hey, happy Friday. Yeah, I can give a

8:58

general overview of our hypothesis for

9:01

this test and how our results moved

9:03

accordingly.

9:08

[music]

9:18

>> [music]

9:20

>> I ended the week with some work for a

9:22

data migration, updating some of the

9:24

queries that power [music]

9:25

experimentation metrics, making sure

9:27

they reference the correct tables and

9:29

logic is correct.

9:33

[music]

9:41

>> [music]

9:51

>> Good [music] night.

9:57

[music]

10:01

Heat. Heat.

10:04

[music]

10:14

>> [music]

Interactive Summary

Ask follow-up questions or revisit key timestamps.

This video follows a product data scientist through their work week, illustrating the diverse responsibilities of the role. Key activities include collaborating with product managers and engineers to define metrics, running and analyzing A/B experiments to evaluate feature performance, and scoping strategic analyses to guide product development. The video also highlights the importance of data storytelling and business intuition in shaping product decisions, alongside the balance between intense focus sessions and meetings.