This week Shirin Deshpande, Strategy Analytics Lead @ Poshmark gives us 4 tips to become a data pro.
February 7, 2019
Can you tell us a little about yourself and what you’re working on?
I lead Product Analytics for Poshmark. I have a team of three and we work together on strategizing various products and also supporting decision making for the entire business. Poshmark is a very innovative company, with a bi-weekly release schedule; every two weeks we launch at least one or two new products or features. My team decides whether these products or features should be launched or not, and decides how they are attributed over time.
Why is this important?
As an online company, it’s very important to continuously innovate and think about what’s next. One great product or feature doesn’t ensure high engagement or revenue in the long term, because times and trends change so quickly. To keep up with this pace of change, it’s very important to involve new kinds of users all the time, as well as help current users to be more efficient on the platform. So, for those purposes, it’s very important to launch new products.
How do you know which features will be successful or not?
Just like in the startup world, there are several which become successful and several that don’t. In the same way, we have thousands of ideas, but perhaps only hundreds become a viable product, and only ten of these go on to become a successful product. We must evaluate products properly and give them enough time to mature, or be removed altogether so that the other ideas can thrive. For that purpose, my team is the key entity to decide whether certain ideas live or die. If we don’t evaluate their entire impact properly, good ideas might disappear and bad ones might still persist on the platform.
In a general way, how do you actually decide which features to include and which not to include? Do you have a very defined process to do that?
We would like to have a defined process to do that and we’ve tried to a point, so that’s where specialization comes into play. We need to have a very holistic business perspective. For example, does a certain feature lie within the vision of the company or not? We are a social platform, so we want to ensure that every feature doesn’t create any negative value within our community. We have standard metrics that we tweak as we go along.
Can you give us an example of a feature that did work? Or maybe one that did not work?
If something completely doesn’t work, it gets cut away at an experimental stage, where we test it out for a few weeks and we see.
Something that does work is our ‘testing rooms’, where we have several types of people style each other. We launched a very simple test to see how many people would like to be styled, and eventually we saw that the answer was ‘a lot’. It wasn’t just commercially successful, but also in terms of customer engagement too. And sellers were happy because they were able to sell based on their capability to style people.
So with features like the virtual fitting room, you kind of start small with a limited sample size and then if it works go out bigger?
That’s exactly right. We do a lot of baby steps.
Poshmark’s is well known for it’s data science capabilities. Overall, what’s the recipe for success here?
You have to focus on four things. One is accuracy of data. You don’t have any reference points with your initial dataset, so it’s very important to ensure it’s accuracy from day one.
The second thing is to ensure scalability. You don’t know how fast a particular product or feature will scale, so you need to consider whether it will actually be sustainable in five years time, for example.
The third thing is to stay focused on the initial goal. With growth comes inefficiency; its easy to start moving in too many directions. To stay focused requires a conscious effort. Stick to your goals, don’t duplicate efforts and allocate resources in a way that’s efficient, for example, by assigning specializations to everyone.
The last thing which I believe is the glue between the company and the team is the culture. Culture is a very difficult thing to build. Poshmark has amazing core culture values and we are very focused on people. We have very high benchmarks on how culturally fitting the person is and that has helped us a lot because it helps us work better together.
If you get all these four elements right, you’ll be a data science pro!
This week Shirin Deshpande, Strategy Analytics Lead @ Poshmark gives us 4 tips to become a data pro.
February 7, 2019
Can you tell us a little about yourself and what you’re working on?
I lead Product Analytics for Poshmark. I have a team of three and we work together on strategizing various products and also supporting decision making for the entire business. Poshmark is a very innovative company, with a bi-weekly release schedule; every two weeks we launch at least one or two new products or features. My team decides whether these products or features should be launched or not, and decides how they are attributed over time.
Why is this important?
As an online company, it’s very important to continuously innovate and think about what’s next. One great product or feature doesn’t ensure high engagement or revenue in the long term, because times and trends change so quickly. To keep up with this pace of change, it’s very important to involve new kinds of users all the time, as well as help current users to be more efficient on the platform. So, for those purposes, it’s very important to launch new products.
How do you know which features will be successful or not?
Just like in the startup world, there are several which become successful and several that don’t. In the same way, we have thousands of ideas, but perhaps only hundreds become a viable product, and only ten of these go on to become a successful product. We must evaluate products properly and give them enough time to mature, or be removed altogether so that the other ideas can thrive. For that purpose, my team is the key entity to decide whether certain ideas live or die. If we don’t evaluate their entire impact properly, good ideas might disappear and bad ones might still persist on the platform.
In a general way, how do you actually decide which features to include and which not to include? Do you have a very defined process to do that?
We would like to have a defined process to do that and we’ve tried to a point, so that’s where specialization comes into play. We need to have a very holistic business perspective. For example, does a certain feature lie within the vision of the company or not? We are a social platform, so we want to ensure that every feature doesn’t create any negative value within our community. We have standard metrics that we tweak as we go along.
Can you give us an example of a feature that did work? Or maybe one that did not work?
If something completely doesn’t work, it gets cut away at an experimental stage, where we test it out for a few weeks and we see.
Something that does work is our ‘testing rooms’, where we have several types of people style each other. We launched a very simple test to see how many people would like to be styled, and eventually we saw that the answer was ‘a lot’. It wasn’t just commercially successful, but also in terms of customer engagement too. And sellers were happy because they were able to sell based on their capability to style people.
So with features like the virtual fitting room, you kind of start small with a limited sample size and then if it works go out bigger?
That’s exactly right. We do a lot of baby steps.
Poshmark’s is well known for it’s data science capabilities. Overall, what’s the recipe for success here?
You have to focus on four things. One is accuracy of data. You don’t have any reference points with your initial dataset, so it’s very important to ensure it’s accuracy from day one.
The second thing is to ensure scalability. You don’t know how fast a particular product or feature will scale, so you need to consider whether it will actually be sustainable in five years time, for example.
The third thing is to stay focused on the initial goal. With growth comes inefficiency; its easy to start moving in too many directions. To stay focused requires a conscious effort. Stick to your goals, don’t duplicate efforts and allocate resources in a way that’s efficient, for example, by assigning specializations to everyone.
The last thing which I believe is the glue between the company and the team is the culture. Culture is a very difficult thing to build. Poshmark has amazing core culture values and we are very focused on people. We have very high benchmarks on how culturally fitting the person is and that has helped us a lot because it helps us work better together.
If you get all these four elements right, you’ll be a data science pro!