Ep. 110 | Scaling a Startup with Data, Barkha Saxena, Poshmark CDO

This week Barkha Saxena, Chief Data Officer at Poshmark joins Allison Hartsoe in the Accelerator to talk about scaling with data. Barkha shares how her team added the customer-centrics layers to boost both the intelligence behind the data as well as the value for the community. This sense of a “data-driven heart” has fueled six-years of Poshmark’s exponential growth. Hear how she did it on this week’s episode.

Please help us spread the word about building your business’ customer equity through effective customer analytics. Rate and review the podcast on Apple Podcast, Stitcher, Google Play, Alexa’s TuneIn, iHeartRadio or Spotify. And do tell us what you think by writing Allison at info@ambitiondata.com or ambitiondata.com. Thanks for listening! Tell a friend!

Podcast Links:

Quantifying The Value Of A Data Organization: A Tale Of Two Axes

Read Full Transcript

Allison Hartsoe: 00:00 This is the customer equity accelerator. If you are a marketing executive who wants to deliver bottom line impact by identifying and connecting with your revenue generating customers, then this is the show for you. I’m your host, Allison Hartsoe, CEO of ambition data. Every other week I bring you the leaders behind the customer centered revolution who share their expert advice. If you are ready to accelerate, then let’s go. Welcome everybody. Today’s show is about scaling a startup with data, and to help me discuss this topic is Barkha Saxena. Barkha is the chief data officer at Poshmark, which, if you haven’t heard, is a social commerce marketplace where people buy and sell new and used clothing, shoes, accessories. It’s actually quite extensive. I was really surprised at the spread. Barkha, welcome to the show.

Barkha Saxena: 00:55 Thank you, Allison, for having me. It’s great to be here.

Allison Hartsoe: 00:58 Tell us a little bit about how you got to be part of the Poshmark team and then if you could also tell us a little bit more about Poshmark. As I know, there are some very early adopters out there, but maybe not everybody understands the platform.

Barkha Saxena: 01:15 Absolutely. So I joined Poshmark almost six years ago. I had just quit my job before that, and I was trying to figure it out like what am I going to do next? And I pretty much wanted to stay in the data field, which is what I had done before coming to Poshmark. And I was looking for something in either the mobile or the commercial space, and it just turned out that I found Poshmark, which was mobile and commerce as the less social, all of it was together. So that taught me to just reach out to Maneesh and have an inch of conversation. But as I met with more and more people like Poshmark, I just fell in love with the team. I mean, we were still very early. Poshmark was just two years in and they were like 35 or so people in the company, so pretty small. But I really liked the team, and I love the concept of a time to build because if you think of it in 2014, mobile itself was still pretty new and trying to build a fashion commerce, but the social aspect was a very unique idea. So I was very excited that the role data could play in there. And then you get to do that with really fun people that just to know planos. So I joined Poshmark six years ago, and I’m still here and having as much fun.

Allison Hartsoe: 02:29 Could you take a sec just to say what the culture’s like. You’ve said that you really liked the team, but I’ve been thinking a lot about customer-centric cultures and it strikes me that Poshmark probably has this, how would you describe what it feels like inside the company?

Barkha Saxena: 02:42 Yes, so it’s a very people-centric company and the best way to understand that is we have only four cultural values for Poshmark and they are focus on people lead with love, together we go, embrace all weirdness and the apply these schools values internally when we work together. But we also look at the community via serving to the same aspect. So again, focused on people. We are in the service of our community. So we are building the space-time every decision they make. Let him talk to our community and how it will help them. Similarly, internally, every decision this starts with people-focused, and then we lead with love, and that just helps in so many decisions. Whenever in doubt, just lead with love, and things will become much more clearer. Together we grow again together; they grow from the employees’ perspective, the focus on each and everyone’s growth and working together so that we are going.

Barkha Saxena: 03:37 But then we also look at it from the community perspective. Poshmark is only going to grow if our community is growing better, and they are benefiting from the Poshmark code. And lastly, the embrace with all weirdness. It is such a powerful thing because what we are saying, and there is the embrace all diversity. Everyone with any convictions, license, title, anything is welcome at Poshmark, whether it’s an employee or it’s one of our community members. So if you think of it in all those four values, what is underlying is the people, and that makes Poshmark does so much fun. It’s a company that people are focused really smart people focus on working together in building a product which is binging of building a community and bringing people closer together. People who are passionate about fashion but doing it in a very selfless way, so everyone focuses on the other person’s goals, and that just allows the whole company to go together.

Barkha Saxena: 04:33 If you think of it, we are not a pretty big size company. Just late last year, we announced that we had distributed our sellers $2 billion and 1 billion of that happened in just the last one year. They are able to do that despite still being a much smaller company. They are like 450 people at the company, but because we are all working together with a very common goal that it’s about the benefit of the community and how we are serving the community and how we are growing Poshmark as opposed to focusing on the self-interest. That does squeezed this beautiful culture, and honestly, for me, it just feels like another home.

Allison Hartsoe: 05:08 That is fantastic. I haven’t heard any company express those cultural values in such a spectacular way. I can see why you love working there. Can you tell us a little bit more about your specific role and what your team does, and maybe you can relate that back to the same values.

Barkha Saxena: 05:25 Yup. So I’m keeping it off. Inside in Poshmark, my role is at the highest level, my function exists to create value for the community, but we look to create that value to the data. So my team works at cost, all the business functions from the growth team, marketing team, part of operations, finance, accounting as well as like I worked very closely with Manish Chandra our CEO and I would go there to look at all the data and how we can use it to keep building the product which will bring highest ROI to each of these functions. And in terms of the community. So to give you an example, my team is divided into like multiple smaller teams, which are very closely aligned with each of the business functions. So the head of my product data team works very closely with the product and engineering team and is very focused on how can I use all this data from Poshmark platform to help the team build the best product, experience, whatever community. Similarly, when my head of growth and marketing team is working with his counterpart business partners. He’s very focused on figuring out how can we use the data to continue to retain our users, deliver them a great experience and continue to bring new users on board who are benefiting from Poshmark and telling into the Poshmark uniqueness of this social commerce platform.

Allison Hartsoe: 06:45 Nice. Do those different areas ever conflict where you have a product wanting to understand one thing from a particular angle and a different part of the business? Trying to understand something else that would essentially cause the two to work against each other.

Barkha Saxena: 07:01 So it does not happen here. I have been in multiple places, so I completely understand where that question is coming from. And I think they don’t, multiple reasons. I have not seen that happening at Poshmark. Going back to that, is this, what is the goal? Right. So it’s not that no function exists for the sake of its own function. We all exist in the service of our community. People can have different ideas for what onset or features will believe at us that value, but that’s it. I think data plays a beautiful load because data is the objective. You need to start questions with a context and some ideas, but then when you take those questions, and you start looking at from the data perspective, you always find the answer, which is very objective, not driven by any bias opinion. And my team is also independent teams, so we are not part of the product at all or not part of the goods and marketing.

Barkha Saxena: 07:55 So we also kind of just look at it independently without being biased with any perspective. And that’s pretty important. It’s a very healthy balance. And our product or marketing or all the business functions, they actually respect that. They post the question, but then they don’t try to influence how we are going to look for the answer to that question. So if a product manager is trying to build a feature that Hey, I want to be able to recommend a good personalized set of listings to my users, they just leave the question there, and then they let us figure it out. Like how we should be building that model and how it should be communicated to the user because that’s really can use the data to understand the types of users, how they have interacted with us in the past and how the social plays with the commerce.

Barkha Saxena: 08:39 Because even for building something which is as standard as recommendation algorithm these days. When you are building that for the social commerce spectrum, you have to look at it very differently. So I think that’s the value I would say the data team brings in. So first of all, because of the Poshmark culture, we don’t have even the business function conflicting with each other because they might come from a different angle, but then they all get together, and we look at it holistically about the user experience, and a user experience is delivered by each business function. If you look at the holistic picture, but then data also becomes a unifying denominator because data is looking at all the questions and coming up with a comprehensive story.

Allison Hartsoe: 09:18 And I think that’s where the structure is very different for Poshmark versus other companies and other companies are getting there. But I oftentimes run across executives who believe that, Oh, it’s just not that hard to get insights from data. You push it together and out pop the insights and yet behind that I think are two things that you’ve been underscoring here. One is the question the executive is learning to ask, and the other is the way the data is structured. Are your executives naturally thinking about the shape of their questions perhaps a little differently than other organizations might?

Barkha Saxena: 09:54 So I would say I’m hesitating in the judging or guessing how the other executives and other companies might be asking the question. But just thinking from the Poshmark perspective, our executive asked the question very broadly so they don’t try to narrow it down because if you narrow it down then what you’re asking for this, can you give me this and this data and I would say that the lack of reason that the data is in such a chunk position at Poshmark because Manish Chandra our CEO, he’s a very data-driven functional person and think of it there are not too many startups who hire a VP of data even you are a 35 people company. He did that because he believed in the power of the data. So he has actually instilled this culture. He doesn’t want to see the answer. He wants to post the question and then let the team figure it out.

Barkha Saxena: 10:44 I mean the whole reason, and actually I do the same thing too. I mean, I’m a data person. So if a question is being post to, for example, let’s say the product data team, I want to jump in sometimes and start sharing my own opinion, but I don’t do that because the reason we have really smart people who were very focused in that area because they have the expertise and they have insights about like a much thought of, but if you of the user behavior that they can actually do a much better job of connecting what the question is and then this thing, the hypothesis in partnership with their business partner and then breaking it down from hypothesis to what data should be pulled and how it should be structured and what tools should be used to give the answer you are looking for. So yes, we do ask a very broad question.

Allison Hartsoe: 11:28 In the process and I thought it was interesting was that you said listing the hypothesis, and I assume that comes from your team that helps the exec take that broad question and turn it into a hypothesis. Is that correct?

Barkha Saxena: 11:41 That is correct. But the way we do it, each of these data projects have a very set cycle. Right with the start to the broad question someone is asking, so data team, they’ll come up with some hypothesis, and sometimes even the business owners can put a few hypotheses there, but they never dictated that this is what you need to convince, here are some of my thoughts. What do you think are the additional things we should be looking here? Sometimes they will just give the business question, but together we come up with a set of hypothesis and then my team will go and review it with the business function owner because the truth is the business function owner have a very unique different types of insight because independent of the data and both pieces are very valuable. And actually one of the things which we do at Poshmark, which is I think very unique and very valuable, we actually also get the opinion of someone who is not from that business function and someone who does not belong to that specific functions data team to get a totally different set of opinion or the perspective which we might be missing because we are too much into that world.

Barkha Saxena: 12:39 So as someone who is working on the, let’s say the growth data team might just take the set of hypotheses and show it to a person in the product data team and say, Hey, this is how this was the question. These are the scenarios I’m thinking of exploring. What are your thoughts? Because unique thoughts come up when people are not immersed into that area, and they call it the review process.

Allison Hartsoe: 12:59 That is very unique, and it is also reminding me of the financial crisis where the companies got so narrowly focused on a particular value metric that they forgot to add common sense in. So in a sense, you’re putting a governor on the process by allowing someone to apply fresh eyes to it that might surface things that you just hadn’t seen. Right?

Barkha Saxena: 13:23 Exactly. And it’s a very codified review process we call it. So we have to write this, what they call the analytics effect in the cases of modeling our data science, we call it model spec, but once you have it, it goes to a business review cycle, not the technical review. Technical review comes after that�business review. And again, we do beyond a startup, so I’m calling it the process, but that doesn’t mean we take like these to do everything. It’s set up a meeting, 30 minutes meeting, go through it, get the feedback and implement there and then you move forward.

Allison Hartsoe: 13:53 Nice. That sounds very efficient. It sounds very agile.

Barkha Saxena: 13:56 We are a startup. So I’m a not a big fan of any project which anybody says, Oh, it’s going to take me three months. I like business, and the word would have changed so much. I mean finishing up the entire thing can take three months, but they have to be the pieces but just fully done in between so that we are starting to create the value and not waiting for this like the super great model or the analytics which comes solid that the train has already passed the fishing.

Allison Hartsoe: 14:20 Barkha, when you first started at Poshmark, did you have to wrestle with the data to get things to queue up so you could answer their questions or was it all greenfield and you could get ahead of it because you came in so early?

Barkha Saxena: 14:34 So I would say it was a positive and not so positive side of it. Then I, so Poshmark is my first startup. Before that, I had to spend my time in bigger companies, their data used to be dead, and my job was just to start working with the data. Poshmark was my first startup, and what I learned on my first day, we didn’t have too much of the data, which was a blessing because my first year at Poshmark was spent in just building the whole data infrastructure. And that was a blessing because as I built the data infrastructure and I built it from the very end goal in the mind that when I’m done with this process, what do I want to see, how this data will be enabling decisions that cause all business functions. So what tools will I need? So I basically started with a very high-level vision and then mapped it down to the execution steps.

Barkha Saxena: 15:21 Very practical, which were basically based on the resources I can have from 35 people to hundred people to 200 people company. But because we started with how this data will deliver value to the business, They were very intentional in deciding what we are collecting, how we are collecting various storing, what tools we are getting, and what data culture we are setting up. So that was, I would say a blessing because I didn’t have to struggle with a lot of things, which I hear from a lot of my colleagues who struggled. When you go in this legacy companies, there’s a lot of data, and it has been built with like just a next thing in mind as opposed to any holistic pictures which makes sense because they’ve exploded in last few years and not everyone had the benefit of starting fresh, but I did, so that was great. The other side of it was there wasn’t too much data on my, I would say first few months at least.

Barkha Saxena: 16:11 I mean then again they all building in the building blocks so I started to get some data in six months, but then interesting thing is the data which we had even in the, like going back to your original question, did I have to listen to it? What we had done then I came at Poshmark they had started storing some real-time data to just keep track of how many orders we are getting, what is the GMV, how many units we are requiring. So we started with like this is the me at Poshmark. We had to start with tracking a few things. You should do as a startup and you start. But we have a very strong technology team but because they build this homegrown tool to do this field time data tracking, and they honestly did it before all this real data tracking thing became the thing.

Barkha Saxena: 16:50 So I’m pretty impressed. But because they had this homegrown thing, people get to adding things to it. And then on that data space, we had the problem which you see in all the data in like most of the bigger companies. So then I came to Poshmark. What I noticed in the big names that initially I was very excited that everybody forgets saw day when they’d see, small company, but they still have enough data. And when you go into the meetings, everyone is bringing the data to make a point, don’t have a discussion, or think about the next step. Well, pretty soon what became clear was that everyone was bringing in their own definitions of what the data should be like they could even argue on things, what is the definition of GMV? Because different people will add data into that data flow. So it was.

Allison Hartsoe: 17:30 I know exactly what you mean.

Barkha Saxena: 17:33 We had to start with the good point but, right, so even on that data, why that was building this more cleaner side of it. It was me and I had one data engineer in India, but we spent time in thinking a lot of discipline even though the data we had because you know whatever it was, it was still great and we were making decisions on it, so I spent some time in fuss to defining the metrics. Let’s just give the definition of first identifying here are the top party metrics we should focus on. They are very small. We don’t need to go a hundred metric yet. Here are the top 30 things, defining their definition, going to the code, cleaning it up, making sure it’s defined the same bit coming up fit the framework, and the metamodel soft like how these metrics should be evaluated because sometimes people will look at the daily data. Sometimes they are looking at the hourly data.

Barkha Saxena: 18:17 Even the line that is going up to the right. That trend can be interpreted in two different ways, depending on how you are looking at it. So I kind of just created like a structure that, hey, here is that we hit all the metrics we should be looking at. This is how we should be looking here at the dimensions on which we should be looking at. These are the trends which exist in the business when you’re looking at these trends here that a type of the seasonality, so kind of be that a playbook so that we have a framework and then we converted that playbook to actually build a UI based tool on the same data so no longer people could still download the data if they wanted to. However, Manish will rely on the tools where we had codified and defined and QA everything to make sure that the data is all correct. The definitions are consistent and be even applied some methodologies to give the interpretation of the data so that when we, once we get together we no longer arguing on what the data is, we’re arguing on the trends and the numbers and what does it mean and how we make our decisions. Basically, the conversations moved from arguing our data to the decision because data was the same.

Allison Hartsoe: 19:23 That’s kind of the company’s dream. Most companies get maybe one part of that or one aspect of that, but they don’t get the whole package together. And you’ve hit on so many themes in just that brief description. I want to circle back to a couple of things that you said. I’ve heard across other companies that the metrics that you pick and your framework for evaluation need to be trendable need to be extensible through the organization. And I’m hearing you check both of those boxes, is that right? Okay. I was a little bit surprised to hear you say the top 30 metrics, cause I’ve heard other companies go down to more like three or four, and then they unpack them from that base of three or four. But you mentioned 30 is that still the case or is that just where you started to just say let’s get a feel for what’s going on.

Barkha Saxena: 20:11 It’s actually many more than that now, but we have a process on like how we keep track of it. And this is and the reason we have 30, first of all, we are the social commerce space, which means we have the many more metrics which we need to track as compared to the typical eCommerce company. And these 30 metrics are at cost, all the business functions. So there are metrics with the growth team needs to keep track of how different channels are performing. There are metrics which product team needs to keep track of to understand how users are engaging. And then there are metrics which our community service and operations team keeps track of to make sure they are serving our customers the best way and how they’re trading. It was tiny, but it was divided among different business groups who were looking at it. However, as a data team on it, it was my responsibility to do is to keep track of those 30 metrics, and the thoughts I had was every Monday morning, and though that was in the early days. I have to be used to look at like Sunday I will go to all the data and come up with the key insight and the Monday morning the whole executive team will go to get together and pretty much this is not that hard to go to if you are not arguing on the definition and the methodology and what is really up versus what is really down down. So it’s actually very fast to go to the data. Once you have alignment and all we are focusing on is what the trends are.

Allison Hartsoe: 21:30 I love that it’s very fast to focus on the data. Once you have alignment and you understand what the trends are, then you can argue about what it means instead of what it says.

Barkha Saxena: 21:40 Yes. And from there, but I’ve seen that we have lots many more now. Where we have gone to the flashback is a very data-driven company, and this is a culture we wanted to build. So we built it this way from the beginning that the dotting in the tools so that we put enabled every decision make that a cause the company at all levels to have easy access to consistent data. So again, be very clear that no two people should ever be arguing on what the definition of a metric is, and then they want to access it. The same source, the same data pops up. And by doing that, we have a major data literacy at Poshmark so people understand the metric, and there are different metrics that are responsible for it. However, back to still in show that that is an executive level ownership of all the metrics.

Barkha Saxena: 22:23 The has a process. So we have built a tool, it’s an ongoing tool. They call it Athena, just a name I came up with. What do we do in that tool is we have a structured, a set of dashboards for different business owners via B P Papa made data in the way it should be low bet. So we have the also trained each and every executive and their lieutenants to make sure that people are very data educated and then everyone has set of metrics and they are expected to, and those dashboards are data, is it daily, weekly, monthly, whatever the frequency you want. But again, that’s consistency, so you’re not as plugged into it. So everyone is supposed to give you their data but then every four weeks my team runs a day-long session where each executive owner comes with their team and each executive owner has a partner from my team so that they’re looking at this data together and not just wondering about things themselves, but in that day-long meeting, we together go over the metrics with each of the business owners.

Barkha Saxena: 23:21 But obviously let’s say a business owner has hundred metrics, you’re not going to go with a hundred metrics in like the 20 minutes I’m spending with that executive. However, you have been looking at those metrics since the long time that you haven’t met Amato Loco and what you’re looking for and where you see the trends, and you have this waterfall closet, right? So this is the top metric I look at. If I see some trends in dos, that top metric, here are the three metrics I’m looking at, and then you go down some there and in this monthly session which I have each of the executive owners we just focus on looking at the trends with that owner had identified his work thinking up to the entire A team and the tosses just work beautifully because we had a lot of metrics, but it’s not that there’s one person responsible for it. There are different business owners, but what my team does it feed think it all together because if you leave any of the functions in silos, you will miss the full picture. A user is a user. They are not what the good team is doing, and they are not what the product team is doing their expedience, and the value they’re getting from the platform is a function of what each and every one of us is doing. So my data function does this job of just bringing all the functions together.

Allison Hartsoe: 24:27 This almost sounds like a chief operations officer.

Barkha Saxena: 24:30 Well, we are just doing the data aspect of it. Right? So the CEO over who I’ve worked with, yeah. So his role is now from the data. You have to have the execution strategy. So my CEO actually sits in the day-long meetings. So there’s my CEO, and I are the two people and the person who owns the Startup, the other one who was sitting in that meeting the entire day because a lot of the execution steps comes out of that meeting. And then we have a whole process of like how do we follow on that process?

Allison Hartsoe: 24:56 This makes a lot of sense. And what I like best is it’s got a lot of sensitivity across the business if you see. So not only is the data structured well but if a business owner applying their subject matter expertise sees a trend that seems interesting, you’re able to maybe cross-reference it or understand it in the broader context as well in the organization.

Barkha Saxena: 25:20 And then we do the communication. We make sure that it gets communicated to the entire exec team, and then they do the communication to make sure everyone is obvious.

Allison Hartsoe: 25:28 Very nice. All right. Now I have to ask because we haven’t actually mentioned this metric. Are you using customer lifetime value or some of the components of customer lifetime value?

Barkha Saxena: 25:37 We absolutely do and I just to give you an example of that, honestly, yes, it’s a classic case of if we were making decisions just based on what user does in like the first day of the first week, we barely making so many not great decisions or at least not high ROI decisions. I was a cohort like the users who come on certain days to a certain platform. I mean we have a mini dimensions of the use of cohort, but our use of cohort go through the smiley curve as we call it internally, which means like any other eCommerce company or any, I would say a social company, any company which has the users, you’ll get a lot of users on board, and generally, they did some digging, right, and then the cohort has stabilized to some point. It’s a very standard cohort curve. What happens with Poshmark is, yes, there is an initial decay with some of the users don’t extend the platform, don’t have the same high value as on day one, but once they get to the stable point, and then we see them starting to grow, and that’s what we call the Poshmark love effect, or you can call it social network effect.

Barkha Saxena: 26:38 That is something, at least I have not come across in many places and be called them a smiley curve because if you can visualize it comes down, but then it starts to go up. So if we were not looking at it from the customer lifetime value, we will have a very different picture of different cohorts and different tax forms in different channels as are the different products each does interactions, how they lead to it because the picture is not complete. Just look in the beginning. So we absolutely use a lot of lifetime value for almost all decisions.

Allison Hartsoe: 27:04 That is awesome. I love that name of the smiley curve and also the fact that after that initial decay, you can grow customers through the right, not just through the data, but through the actions you take that reflect perhaps a relationship. I mean, when you talk about it as a social network effect, to me, that sounds like two ways back and forth. The organization is listening and supporting the customers as much as the customers are giving back to the organization. You’ve got a two-way model there.

Barkha Saxena: 27:34 Yeah, and we don’t think of it, it’s our jobs. We bring in people with the promise of this great experience and for our sellers a great success, and it is our responsibility to make sure that we delivered on the promise.

Allison Hartsoe: 27:45 But it’s not just a great experience. I have to call that out because so many companies are using that term about, Oh, I just want to have great CX, and I always think about that as kind of a red herring because they don’t respect the heterogeneity of the customer base. Oftentimes when they design these experiences, it’s just kind of a flashy thing, but what you are doing is different. It’s not just a great experience. It’s a great sustaining experience. Maybe it’s a series of micro experiences along the way from how I act as a new customer versus how I act as maybe a customer who bundles or a customer who has a boutique or any other number of things that you could lever along the way.

Barkha Saxena: 28:26 Absolutely. I mean, that’s what we call. So when you look at any data, or even then they build a model, it’s never one ping because the user has so many dimensions and you get different insights.

Allison Hartsoe: 28:37 So can you talk a little bit more about that? When you say it’s never one thing, you mean like within any user, there are so many dimensions of detail behind them that whatever model you form could reflect differently depending on the question you’re trying to answer.

Barkha Saxena: 28:52 Yes. So to give you an example, right, so there are classic demographic dimensions to the user, right? So there’s agenda and, and uh, but beyond just the, the reason I stopped this I second because we actually don’t collect a lot of demographic information because.

Allison Hartsoe: 29:06 It’s not very predictive.

Barkha Saxena: 29:07 Right? So it’s all about that. So, but the way we defined the users is, so which of the platform you use that actually matters? The people who are on one platform behave differently than the other platform. Uh, the users who are very seller focus was the by focus. They do behave differently depending on the depth and the best of the social networks. Different user groups can have very different behaviors and requirements. Then are sellers who have become, for example, the power sellers, they have very different asks in the requirements then. The support which is, which a new seller might need. So we have to think about both of than get thinking about the feature and not just focus on one versus the other. Even from the bias perspective, we have people with different passions and different reasons for coming to Poshmark, and we have to make sure that when we are thinking about any changes to the feature, what we are showing in, you’ll see is we are personalizing it based on the signal so you are leading and those signals could be the people you are connecting to, the items who might be liking on the platform, the items you might be sharing.

Barkha Saxena: 30:09 So we use all the data to make sure that the experience is very personalized. So it’s not that we need to build like different apps to cater to different users. It’s making sure that then you have already left so much information about you in the way you are using the app. They are preventing you with the experience, whether it’s a movie they recommend you that you should connect with, and the people recommendation, all recommending the items you should participate or thinking about what seller tools they should be building. We need to think about a different segments of the bias or the sellers as opposed to just looking at them as this one lens.

Allison Hartsoe: 30:43 I love it. You’ve really taken heterogeneity through data into an exponential level and found not just ways to collect it but ways to collect it meaningfully so that you know what behaviors are actually driving a difference for your business. And I think right there is where you quantified what you call the love effect, and you can see people so happy to have a new tool that matters to them or the ability to share it with their friends or whatever personalization or recommendation that you’re layering in seems to again have that two-way cycle. That’s really great. Okay, well, Barkha, are there any other examples that you want to share before we talk about how somebody should think about their data structures?

Barkha Saxena: 31:29 So I would say as the up picking off all the data structures then it goes to the data structure for the data tools or the culture, I think it’s starting with at the very top, not just from the business question I’m asking, but what is the purpose of this business function and but I talked to the business function I’m talking about like within the data team. So if you are trying to build something, what is it supposed to do so that you are thinking it holistically and from the getting perspective, whether that be you build the data in the right way and you build a tool in the correct way. An example of that when we built our AB testing platform, we wanted to build it in such a way that we can get insights into multiple phases. So the early early on and then the data matures and like when the other, the confidence we wanted to be able to do this in a very automated way.

Barkha Saxena: 32:18 Because again, the IB still, I’m like a startup, so we need to be very efficient. The still needed to have all the statistics of measure sense. So making sure that when you were polling something, yes or no, there’s a statistical significance, but we needed that a statistical significance to be communicated in a very business way so that as we start to pull out these reports, first of all, we wanted all the product managers like a sample to be able to get the reports, and even they get the reports, we need them to be able to understand what it’s saying without really trying to understand what the heck P value means. So once we thought about from this perspective, doing these like one-off AB tests of any other Nina test is not a solution which can scale because then there are only so many tests you can run unless you have building like a massive organization.

Barkha Saxena: 33:02 Ditch is not the maybe function, the function of that, the scaling mindset. So we built an internal AB testing tool, and we looked at a lot of external options as well. But there’s a lot of things that are unique about Poshmark from the data customer perspective. I myself being a big believer on, I need everything to be running on our data because I don’t want the data from the three places coming in because again, I don’t want the data confusion. I want the discussions on the insights. So we had to build a thing internally. But the benefit of building it internally while there was an investment in building it, but once it’s done, we have become so efficient. So just comparing from last year to this year, We were able to output 75% more AB testing analysis as compared to last year because like on the stuffing, the whole thing, because we thought all this efficiency in the process and we also just educated our entire product engineering community so that as they start to send out ugly insights, there is no confusion, people understand what’s coming out, what it means if they are curious in between, there’s a place that they can go and pull the data.

Barkha Saxena: 34:01 So again, just I think it’s using the, I would say the same underlying principles that the purpose of the data is to deliver value in helping making decisions. Depending on what business problem you’re trying to solve. You have to think about the data structures accordingly and have to build the appropriate tools, and the only way things can scale is then we’d go outdoors like the constant customization and doing one ad hoc pink to just building these the P table, what I call the data product. So instead of trying to do the projects, they tried to work on building the data product so that whenever you do something the first time, you might doing it to answer the question, but then you are doing it think of what are the additional usage your team versus the other data teams can get out of it so that you build this VP table data product so that you can go on to do the next exciting thing and not just be pulling the same data lov building the same model and not having as much fun. There’s just so much data like even at Poshmark. I mean, it’s let’s have more fun with the data as opposed to keep working on the same thing again and again. So I think man, I’m going, but that is like when you started talking about the scaling startup with the data, I think that’s the mindset we operate with that let’s build data product.

Allison Hartsoe: 35:13 Yes. And I think that’s the problem when you part starts putting in all these different tools and trying to cobble together a stack from external tools. You don’t really control where the tools go. You don’t really understand how things are collected or built, or things change, and essentially, your ability to have more fun with the data is directly related to the way you’ve built the stack.

Barkha Saxena: 35:35 Yeah, absolutely. And that’s why we have made very heartful decisions on what tools whenever we got anything from the outside will fit Poshmark ecosystem.

Allison Hartsoe: 35:44 Yes. This has been a very well thought out structure, and I appreciate you spending time with us to talk through how you did this. If you had to boil it down into first, second, third, what would you do? I heard you say scale and I’ve heard that from other senior executives who had a lot of progress, but what would you say first think about this second, think about that third, how would you frame it?

Barkha Saxena: 36:08 I would say all this begins with the end in mind. That’s very, very important. The whole point of thinking about the vision and the mission of anything you are trying to do, I cannot overemphasize the importance of that and doing it well actually takes time. You have to fit for hours to dive that one-line statement of what do I want to really do and then mapping it back from there and when you are thinking about the data, changing the mindset from how to what because generally there’s a tendency to just start collecting the data and start going from there. I think even though it’s an addition which have a lot of data, even for the further data collection, if we can just start with what do we want to do with it and not the next month but next two to five years vision of what do we want to do with it?

Barkha Saxena: 36:51 And that will help you collect the data very differently that will help you think about the tools very differently. So again, the very first step will be think about what is it I want to achieve in the next three to five years from the data and then map it from there to I mean assuming that the answer will be somewhere in creating the value for the business. I would say more the streamlined or more personalized to specific organizations, but mapping it from there to be able to do that. What data do I need and what frequency and how should we structure. So we do have to think about the data, and then from there, we think about what tools we need to be able to. And in the end, your top house statement. We also have to think about the culture because at least in my biased opinion, data is powerful when the entire organization is enabled to use the data as opposed to just the data team and then some, once you identify what we want to do, How will I enabled organization, then what data we wanted to collect.

Barkha Saxena: 37:48 Then we think about the tools we need to be able to scale data to deliver the value. And then, after that, we need to think about the specific data functions development and what types of, I would say initiatives they should focus on. And even in this last year, I think we should think about that very broadly because data partners with business function in multiple ways. So a couple of months ago, I wrote an article about measuring the ROI of the data team. And in that article, I have shown a two-way metrics, and the purpose of that metrics is the one-sided shows the different business functions where the data team adds value and the other dimension shows the different ways data adds value to the business. And we have to really think about all those functions. For example, it is so much excitement around the AI and the machine learning that a lot of times people don’t think even today 70% to 80% the value gets delivered without that because there are lot of other ways you add value to the, whether it’s the insight, so finding the types of meetings I was talking about when everybody’s informed doing a lot of sophisticated analysis where you bring in the statistical concepts, but the modeling part is only needed when you are going into the automating decisions, which at least in the startup world comes to much later.

Barkha Saxena: 38:59 I mean even at Poshmark like although plus machine learning production in Whyton was built just last year and honestly very proud of that, that’s not late for me. That was a plan from the very beginning because there was so much we needed to do. Like why would you jump on the like I always give this example to people. Like you never use the same tools and the same ingredients and the same recipe for all dishes.

Allison Hartsoe: 39:20 Exactly. Yes.

Barkha Saxena: 39:22 Right. And you can apply the same thing everywhere, right? You don’t use all three walls to build your house. You have different things to do this in parts of house building projects. Why would you not remember the same concept for the data, and this is a discussion I have more of it. I would say the new young graduates, right? Because a lot of the schools I think selling this whole machine learning in such a way that everyone thinks that that’s all the mathematicians do and you have to generally visit that that’s what we do for the problems where that makes sense and even in doing that, the business judgment and the human intelligence you need to bring in, you only learn that by actually working on the data to what the classically would call the analytics or call the data science projects because without that you can Toine the garbage data.

Barkha Saxena: 40:07 You can have as long things that you’re trying to model all. You might end up building a model which is so biased. It’s okay. Going back to the Poshmark example. If we would have built our recommendation and guide them with the goal of, let’s build the best recommendation algorithm from the data we have. It can be a different answer than let’s build an algorithm which will delight you with experience at Poshmark because they come to Poshmark to find a variety of inventory we build a very different thing, but we added human elements to it. They were places where the massage the data or transform the data because we know there are things which makes more sense because historical data would only tell what people have been doing in the past what you expose them to.

Allison Hartsoe: 40:47 It’s like driving a car, looking out the rearview window.

Barkha Saxena: 40:51 Exactly, and you learn these what I call, you can call it the business common sense, or you can call it like I always tell people like show the respect to the data before you tow it to the tool.

Allison Hartsoe: 41:01 Yeah, that’s a nice way to put it. Well, Barkha, if people want to get in touch with you, I’m going to suggest that they reach out to you on LinkedIn. You can find her at Barkha Saxena and I did see that Poshmark is hiring and I’m sure you’re always looking for just the right people on your team, so I’ll just put that shout out there, but if people haven’t experienced the Poshmark platform, is there a way they should get started with that?

Barkha Saxena: 41:29 Yes, absolutely. I would say install the app or the web, the other multi-platform company, and if you want to venue, join the Poshmark. I do have my vessel code. I can give out that there’s my first name, Barkha, B, A, R, K, H, A two zero one four, again, my first name, and then for the zip, two zero one four. When you join the Poshmark, You get $10 credit to get started with. Well, anyone who has anything to sell in your closet, I would highly recommend for you to come and experience the Poshmark and see how the selling experience looked at Poshmark. We have 8 million people selling at Poshmark. Yeah, the 16 million users for discharge. So if you want to sell at think of the audience, you’re going to get at the platform. And for people who might be interested in purchasing, I would say we have 38 million items which are shared on Poshmark every day.

Barkha Saxena: 42:25 So these are the sellers who are actively promoting their items because there’s the legal inventory. Yeah. If you have almost like 2 million likes, which happened on Poshmark every day, which means that users are finding to almost two million items, which are so beautiful. And think of it, a user spends 23 to 27 minutes in our app every day. And they opened the app seven to eight times, which I am from the data for some perspective, I will think that as a signal that people are having a delightful experience, that they are so engaged. So I will invite if anyone who has not joined Poshmark come join us and to be this on this beautiful journey thereon.

Allison Hartsoe: 43:02 That is a fantastic way to wrap up our show. Barkha, thank you. I’m actually gonna go see what I can sell from my closet. You’ve inspired me, especially with the customer love side. As always, links to everything we discussed are@ambitiondata.com/podcast. Barkha, could I get a link to your article that you mentioned as well? You put that in there?

Barkha Saxena: 43:25 Yes. I will send you the link. It’s a Forbes article, but I’ll send you the link.

Allison Hartsoe: 43:29 Excellent. Barkha, thank you for joining us today. It’s been just a real pleasure to hear how a startup was shaped correctly with data to maximize the customer love.

Barkha Saxena: 43:40 Thank you, Allison, for having me. It was fun talking to you.

Allison Hartsoe: 43:43 Remember when you use your data effectively, just like Poshmark, you can build customer equity. It is not magic. It’s just a very specific journey that you can follow to get results. See you next time on the customer equity accelerator.