Use Data, Rally Team, Go Global

Julija Kosareva is a business intelligence analyst at Vinted, the world’s largest second-hand fashion marketplace, headquartered in Europe. Andrei Brasoveanu is an investor with venture capital firm Accel Partners, and he’s based in London.

Andrei: Thanks a lot Julija for taking the time to talk about Vinted! We’ve been fortunate to partner with your team ever since Accel Partners led Vinted’s Series A round in early 2013 and since then we've been excited to witness your impressive growth story. It would be great if you could describe Vinted in your own words.

Julija: Thank you for the invitation, Andrei. To start, Vinted is a social marketplace for women, a platform where women can sell, buy or swap clothing. This can mean new clothes or old clothes they don’t want to wear anymore. The company started as a simple website developed as a hobby by our founders Justas and Milda, and ended up growing into the full-fledged company it is today. The early joiners could easily identify themselves with the company’s mission, and the startup continued to grow fast even before receiving outside capital. Subsequently the company received an angel investment from Mantas, who is now our COO, and next Accel Partners came in and led the Series A.

Andrei: Could you tell us a bit about your background and how you came to join the company?

Julija: I joined as the company’s first BI analyst, when the company only had 40 employees. Back then our data processes were very basic, we were only using basic KPIs, still relying on Google Analytics and our internal database. We were just focusing on high level numbers. Most people on the team were working on product or development. Together with Accel Partners, our founders decided that the company should be more data-driven and take key decisions based on data. We needed to understand what worked, what didn't, and what needed to be fixed. In terms of my background, I’m an econometrics specialist, and I studied econometrics at the Vilnius University, at the faculty of Mathematics and Informatics. Econometrics has a lot to do with statistics in a practical context, analyzing real-time data.

Andrei: Interesting to hear the about the stage of the company when you decided to go all-in on data. What can you tell us about team design at Vinted and how business intelligence fits in the picture?

Julija: At the beginning we used to work as one common company: initially a few people were in charge of product changes and managed all the developers. As the company was growing it was hard to keep track of new features and be effective in overseeing how these teams work. So we divided all the developers into small teams and hired additional people to be product owners. These would be in charge of new product features and changes, e.g. whether we need to first improve retention, or UX, or conversion to first purchase. Nowadays we strive for our product management teams to work on a SCRUM process. Each team has 5–7 people working together, and as per SCRUM each sprint is 2-weeks long after which we gauge results. Each product owner is responsible for what feature they’re launching. In addition, our teams are autonomous in that they have all the resources they need to drive results independently.

On the BI side, originally all of our analysts were in a separate BI team that was supporting product teams on demand. This turned out to be a not efficient way as BI analysts would not see the full picture to be able to dig deeper into the problem. This is when BI analysts started working as a part of product teams. Now we want each product team to work with a BI analyst, to back decisions with data. We now have 6 product teams working on completely different problems and features, and we’re aiming to staff a BI analyst for each.

Andrei: So it seems from what you describe that BI analysts are an integral part of the product development process, rather than sit aside as a separate team. How has your role evolved over time?

Julija: I used to be part of product team, and now joined a new team built over the last 4 months, which is a research team. In the past when we were analyzing lots of data, we had situations when we couldn't see the underlying root causes behind specific drops in KPIs. Now we want to combine user research (e.g. interviews with members) with data, and see what issues members are facing and see how we can properly address those. For this we need to properly segment our user base, through quantitative and qualitative interviews and questionnaires, so that we can see everything from our members’ perspective. We see very different usage patterns on mobile and on web. Throughout all countries we have social profiles of people, to get a general view on second hand e-commerce, in addition to analyzing micro user data or acquisition channels. Now we know much more.

Andrei: What are the key challenges you face on a day to day basis, to what extent is your work reactive to issues you identify or preventive?

Julija: We try to focus on the future, and segment the market to understand what the needs of our members are. Still, all the time there’s something we’re currently working on we can’t just ignore. We work as much with features that are already launched, to make sure they perform well, as with features we’re planning to launch shortly. As an example, an initiative we’re dealing with at the moment is that we’re launching our internal payment system in all of our countries. In the US and UK we introduced payments right away, where the seller would be charged a fee for the transaction. In our old markets we first needed to get both our buyers and sellers on board. There are a lot of benefits about our payment system, in terms of saving cost and reducing risk, but we need to explain to users how useful it is and how they can use this to transact more efficiently. This is an example of the continued conversation we have with our members: if we changed something, would they feel better?

Andrei: From what you say it looks that leveraging data is deeply intertwined with listening to your customers. What are some of the key KPIs you track, around areas like marketplace growth or customer engagement? How do you think of improving those KPIs?

Julija: There’s a ton! But first, I want to emphasize that simple general metrics, such as new members, new transactions or revenues would not tell us much. We need to know how to split these metrics by some important dimensions, to see how the product is actually changing and what is happening. These dimensions can be different depending on what you are currently changing or improving in your product and sometimes it requires a lot of analysis and experience to find them.

Initially we always start with simple metrics, e.g. new members acquired, conversions into different metrics (if they became a buyer, a seller, a lister), MAUs, WAUs. Very important for us is customer retention. We look at this by analyzing weekly cohorts and monitoring how often users are coming back. Besides active members we look at transactions and listings. We want our content to always be fresh, as we want members to upload new content and transactions to happen. Only after this we start looking into second-order metrics. As we've increased monetization, we've started to concentrate on transaction metrics: GMV, lifetime value of members, and revenues. What we've learned over time is not that KPIs are bad, but that sometimes we track too many KPIs, and then we can get lost in data and miss a trend. Also the general trend can sometimes hide an underlying issue, so we need to segment each metric by platform or channel or geography. We are also currently collecting data and working on new methods to get more customer satisfaction metrics as KPIs.

Andrei: That’s an interesting problem, how do you balance digging deeper with not getting lost in KPIs?

Julija: It’s complicated! When product managers work on features, they start tracking more KPIs, and more specific ones. As BI analysts we track more general ones, and dig deeper if we see anything alarming. We spend a lot of time looking at why certain metrics are decreasing, e.g. which members stop engaging. This way can go back to them and discuss. In addition, we analyze a ton of different features for PMs, helping them plan future features with research. Sometimes we happen to look at metrics we shouldn't have tracked at all in the first place. For example we might be tracking a metric and notice deterioration, but what’s really happening in the background is that we’re testing a new market channel, or a ramp in user acquisition, changing overall user engagement metrics. In this case, members acquired through the new channel might not be as active, or if they’re acquired through vouchers they might buy more, so we need to split adequately our analysis by each dimension.

Andrei: What are some of the key tools you’re using internally to help with the above?

Julija: We have our internal analytic system, very useful in building new reports, where anyone can ask and get a dashboard with only the metrics that are relevant to them, e.g. general dashboards, product dashboards, community support dashboards, marketing dashboards. There are always areas where we want to improve in, especially around reporting. Previously we would do weekly reports, to see what has changed from a week to the next, but that ultimately grew into a huge presentation, and we then decided to drop it, as the dashboards were much easier to work with. But there’s room to improve here, e.g. we could have better tools to dig deeper into significant changes over a period or generate more sophisticated graphs. On top of building a great infrastructure, it is also important to market this internally to make people aware of how powerful this is in making their jobs easier. Our data warehouse team is also working hard on building our own Hadoop-based data cluster. Besides our internal systems we use few third party tools, and while we've trialed a few e.g. around mobile customer funnel analysis, we realized that ours work better for our needs.

Andrei: How do you think of international expansion? Was there something unexpected you've learned when you expanded into new geos?

Julija: With new markets your experience can be very different, as there’s large cultural differences. I’ll point out a few examples:

Product familiarity. The first new market we entered after we raised our first round with Accel Partners was Poland. Poland already had an existing early company in our space, so it was easy for women to get on board with Vinted, the concept was not new. The next market for us was France, and there was nothing similar to our product previously in France, leading to confusion for our users and ultimately poor engagement. Members in France didn't know what this product is all about. Simply by tracking metrics you would not find a solution for such problems. The answer was in educating our French user base on what Vinted was all about.

Customer satisfaction. In one country we have members who’re always happy with the transactions, while in a new market we've recently entered, there are more frequent negative reviews. In this instance we came to the conclusion that there are different cultural expectations around peer-to-peer transactions, which we’re making sure to incorporate in both our tracking and product development. Another example is how buyers handle delays from senders, as there are different expectations in different cultures.

Mobile usage. In Western markets we have a very high mobile usage, so on the flip side our users are less social than in Eastern Europe. They read less on the website or forum, and we’re using that knowledge to improve our app experience to increase engagement, but there will always be differences to the online-only experience.

Monetization. With marketing we can target acquisition towards young members, but that comes with different monetization strategies. For example young people in Poland with less money than in the West want to swap rather than buy clothing, such that if they don’t have money they still want to give something in exchange.

Market adoption. In some countries people just like to buy new clothes and not bother with second-hand. Of course you can still attract a dedicated user base in those countries with better targeted marketing. Previously we would market to everyone in a similar way and wait for results, now we've become more specific.

Payments. We were thinking of ways to boost our user engagement in France, and noticed that members would struggle with paying. In France it’s not popular to do money transfers through the bank, and our users would sometimes send money in an envelope or via checks which created friction and delays in the process. After we launched payments this helped our users a lot, as buyers can now pay with a credit card or via Paypal.

Andrei: How do you think of the right team you need to lead BI efforts? What kind of people are you looking for?

Julija: Even though we’re based in a small country like Lithuania, we have access to great technical people who've studied mathematics or CS. We’re looking for very skilled people, who've worked on their own and who don’t need much help in getting up to speed. The company is moving so fast, you can’t have much micro-management. For the BI teams we try to find technical people who know how to do work with data and DBs, and run their own analyses. What’s difficult in Lithuania is to find people with a product- and user-oriented mind. Even for those of us working with the company, it’s hard sometimes to shift your mind from thinking about product features to putting yourself in the members’ shoes and understanding their decision-making process. You need to think broader, to take a step back from analyzing data and think about how you can produce valuable insights. It’s very important for new joiners to have an open mind and have the creativity to brainstorm and generate new ideas. When a person comes for a trial or interview, we try to ask them lots of questions to understand their thought process, beyond just looking at their performance on technical tasks. We want to see how people react when they’re challenged. In terms of core skills we look for, it’s experience with DBs (SQL at least). Then it doesn't matter what tools you use, as long as you’re handy with processing and analyzing data. In general people are working with R. If you’re an expert in some tools, you’d be able to switch. Usually we hire people with a science background.

Andrei: What sets Vinted apart in terms of culture?

Julija: Vinted is very different than all the companies I know in Lithuania, and I’m sure it stands out in Europe as well with its strong culture. We’re very open and we share everything, it’s a flat culture. Everyone can access any kind of information they need, and everyone is invited to share their opinions and add their suggestions. We have good people and strong professionals, and we’re trying to hire the best people in the Lithuanian market. It is very fulfilling when you can learn from your peers and you feel that you are growing as a professional, working very hard and evolving in any direction you like. It is also easy to switch roles in the company and challenge yourself: everyone tries to learn as much as they can, not just in their field, but also more broadly around what could be useful for the company. The harder you work the more opportunities open up down the road. At the same time Vinted’s also very friendly, we have a fun and young community, I always find myself in a happy mood at work!

Andrei: Julija, thanks for your tremendous input, and congrats on what you and your team have achieved so far! We look forward to following your journey!

*******