Single Post

Tableau customer keynote – Data-driven marketing

Lior Barak - Tableau event

Click to accept cookies and enable this content

Hi everybody. My name is Lior and I’m with Zalando for a year. I brought a picture of me shaved so you will recognise me next week because summer is coming and it’s time to shave I think. I am working as part of the marketing app team. My position is basically the communication between the BI roll data so the channel managers and to create this synchronisation between what the channel managers need to do and what the BI supply. It starts with working out and discovering that you are not relevant anymore for our customers. Basically our state orders are stating that whatever we are doing for them is not good enough. We discovered that we are supplying them a weekly report while they are loosing a daily one. And we understood that something has to change. And this is basically our path. This is how we learned and we created this change to make sure that we actually supply what we told our customers. When we prepared the data, as I said we used to do it only once a week, it was a very heavy file. It was an Excel file. It was a very old file actually that we used to have since 2015. It wasn’t since than changed because every change of this file caused a crush. I believe everybody knows it. And than we decided that it is time for us to start feeding ‘OK what are the people or what are our customers not liking about us, what do we need to change to make it actually working’. And we started to do research and we discovered that we actually in our reporting we are supplying 25 KPIs. It was a premium researcher from the University of Bologna who found out that the optimal number for a channel manager to have is 10 KPIs. No more than that. When you’re supplying more than 10 KPIs he cannot focus actually on the data. He doesn’t understand actually what is going on under. He doesn’t know how to work with the data and than is getting lost. And this is the biggest key point that we found out when we started researching it.

We tried to figure out ‘OK so we have 25 KPIs how many of them are actually relevant, what do we need to do?’ And than we started to go over processes. ‘OK why are we so slow, why do we supply the data only once a week, why don’t we supply it everyday, why is there a blocker from us, why?’. Our channel managers on a day basis actually have data abbreviated and supplied to them by their own means. They have done meta processes for every single thing. They went to the dashboard of Facebook. They went to the dashboards of Google. They went to the dashboards of Twitter even. Downloaded the data. Put it in an Excel sheet. Turn it into abbreviation. Started to work with it and it’s made no sense that they are doing our job because for us it was the same. Now not only that. For us to actually supply the data to our customers at the end because there were so many of them we needed to aggregate the data with ours. Which was sometimes 1.5 Gb. I believe nobody can open that with Excel. This is where we realised that again we need to create a change. Which brought us basically to start thinking how do we take this Excel files and actually transfer them into the times of today that you need to consume data in a fast way.

And as you can see it’s an amazing report. It’s one of them. I’m sorry I cannot show you numbers but you can see that basically we had only a level of the quantity and no less. And that’s it. That was the only dimensions they could use. Not only that. As you can see we started from Histels [phonetics] is Histels county and than costs in CPI and CPU and so on and so on which didn’t make sense. It was too much for them to process and it was even for us too much to process and realised we didn’t change it. You can see here this amazing graph with the mouse that we used to present to them. I believe that everybody can understand exactly what they need to do tomorrow. Right? [laughter] This was where it went. ‘Cos we gave data that actually wasn’t usable for anymore. And this is where we decided that it’s a time to start to adapt and integrate it into the company and try to see how we do with it. And than we met an amazing book by Peter Grooker [phonetics] that says ‘there is this thing where you want to start something new you need to start…you need to stop something old’. And this is what [00:05] actually we have done.

We stopped reporting. We stopped giving Excel sheets for…I think it was almost a month that nobody received any report from us. And everybody asked ‘what are they doing these guys with the CPI guys with the purpose’. Everybody was coming around asking us questions about how we would not receive the data because they’re supposed to know what they are doing. But the truth was that we didn’t. We didn’t know what we were doing and we didn’t know how to learn it. And we sat together with our channel managers and started to understand ‘OK, what does this data actually mean, what are they using it for, what is this KPIs you’re using, why do you actually change RY [phonetics]’. An yes we are it for this company. We’re supposed to know what RY mean. But apparently our RY wasn’t their RY. And when they came to collect it the returned investment wasn’t the same. And we needed to understand why it is not like that and we needed not only that. We needed also to understand ‘OK so when you’re saying an RY what do you mean and can it mean the same for the Facebook channel and for the Google channel, do they all receive the same understanding on how they use the data’.

And than we decided to come with this great idea. No more Excel. Indeed it sounds like a story from a fairy tale. ‘How can you still give Excel to people and how can people still be using Excel’. But actually what we’ve done during this week was to learn how to use the data. What they aggregate. What they’re using it for. And actually to create the same structure of their thinking pattern in the database. So the data base knew exactly what it needs to do. We knew how to give them recommendation based on what they’re doing and to supply them the best information when they needed it.

When you think about it and the channel manager when he arrived in the morning and he saw the data from the marketing guys the first thing he’s doing is opening his dashboard on Facebook or Google. And he started to look ‘OK what I’ve done yesterday, how much did I spend, What is the return investment, what’s going on, how many users, how many…’ It’s a lot of numbers that they’re using but actually in the end you can do it in one number. You can focus your RY. You can gain…You can aggregate these data together and this is what we’ve done. We sat down and we took what is important for you. What are the numbers that you need to use. What are the numbers that actually make sense to optimise best on. And than aligned everybody. We took the entire optimal. I remember it was a couple of channel managers sitting there. They were all used to do a manual process and suddenly I told them ‘you don’t need to do it anymore because we’re gonna do it for you’. Everybody thought that it’s a joke. And than we pulled out our webpage which was evolved. And we tried to actually explain to them how they can use 3 or 4 KPIs and than write the hierarchy in the right dimension. And not only that. Instead of giving them the dimensions of the country we gave them the dimensions of the creative that could go with the level of creative that actually know what does it mean to gave the data. And than started our path.

Our path started actually with the desktop and the blow reader. We at that point at that we had a server we…Let me show how it’s gonna go. How people will take it. And I think that this is something that will interest everybody starting today with Tableau. Because you don’t show how the path is gonna go and you try to work around it and see how you can fit it. And this is exactly how we have done it.

We started with Tableau best of. We updated the data on the database every morning. We sued to come before our channel managers to make sure they had the data at 9 a.m. already. And we took this data afterwards and saved it on a G drive. Very simple. We put a lot of files but than discovered one issue that we had there. People didn’t actually download the latest file. And they used to come to us and they ask ‘guys but why my numbers are wrong’. And than we go with them over it and were like ‘OK show me what file you are using’. And we found out that ‘OK they’re using a very old file, it’s from last week’. ‘How can I use these KPIs if they’re form last week’. And I agree with you. You’re completely right. And than we started to understand that we actually need to do the Tableau server and we needed to move. And this is actually what we’re doing these days. We started to adapt our server. We wrote it out today to the app team and they started to use it as a tool that actually give the data on a live version and actually we don’t have these issues. That’s one of the key learnings that we had. We also learned that we separated our systems to 2 different types. We had the Hive adoop [phonetics] and Spar [phonetics] that we are using to process the data. We have a lot of data that we’re receiving. We are receiving data form Facebook, phones and Excel that we’re using. We’re using a lot of partners and all these partners are sending us a lot of information. We removed all the KPIs that sit under the bee Hive. And the bee Hive also processes information every morning and they finish it by 6 a.m. every morning. The application than decided a fraction of the data because not everything is correct. Sometimes they use the wrong data and you need to correct it. And [10:14] this is what we do. And than the final stage actually I’ll be getting the data so actually we can use the data at 8 a.m. already prepared and ready for our channel managers. Than when we finish the application on the hive and move it the operation to the Excel and this is the biggest issue for us at the moment because not everybody knows how to write it. And this is where Tableau is coming in. It actually was the big insight of last week. We actually understood how to start using the data in the right way once we moved it to Excel as an aggregated table. People can actually start using it from Excel directly from Tableau server and form Tableau server to adjust and start using it. It’s really simple. It’s really easy and we’re making sure that from now that our managing board are actually using it.

And not only that. We are seeing that it looks very good this time and it contains only 15 KPIs and not 25. We didn’t reduce it until 10 yet but we’re gonna write it. We did reduce it to 15 KPIs to understand that we can actually reduce the amount of KPIs that we supply. We try to think in the choose of the member in Zalando. Thinking about the issues of the customer. And this is actually what we’ve done. We’re starting to give them 15 KPIs and we’re actually gonna reduce it. We’re gonna take it as low as we can. Because we think that at the end as a channel management you don’t have time. You just don’t. And than we went to our channel management with this beautiful paper that we prepared. By the way this is my writing and I’m not allowed to write anymore in any meeting in the company. Nobody likes it. But what we’re basically doing is we’re sitting down together with our channel managers and we’re trying to understand ‘OK do you think that these dashboards with what you have here will be enough for you and be sufficient?’ And than we started the process to sit on the Tableau desktop and started to got though the board with actually leaving them a final version of what it’s gonna look like. We found out that there wasn’t much information. We tried to give it as simple as we can. This is how it’s gonna look like and this is the final version.

In the company today we have I think in our office only in the desk that we’re sitting, 3 white boards. These 3 white boards have become our best friends. There is no more presentations in Zalando. In our team. The only thing that we’re suing is either the Tableau to present numbers or the white boards. If you need to present something you use the white boards. You’re not allowed to make any presentation anymore. And we actually try to do the same method when you come into our channel managers to paint with them together in the white boards and than transfer them into a paper that you can actually move it to the developers of the dashboard and build it.

This is for example our CRL dashboard. We finish it with them 2 months ago and this is what they receive. And they wanted a very special part of it. You cannot read it. I’m sorry. It’s basically asking to have a URL action. Meaning that every time that they are creative they are gonna recollect and they will go to directly to the campaign and be able to optimise it. And they will need to work on it. And this is something that was very possible in a very easy way with Tableau. And this is how we received the feature. Our channel managers want me to leave outside of the Tableau division this close divider. This sleek dashboard will basically have the numbers on top and it will be interrupted. It will move to a website. They will be able to optimise the campaigns and they will basically will be able afterwards to make action. And action I think is a key component of what we’re doing today. We are having key action for each team. We have for them…one is the main key that will optimise the best media creative and this was one enough learning that we have. That we need to agree on one KPI that will be enough for this key to actually optimise. Than we took another step and we said ‘OK what will happen tomorrow when our founder will come to the office and wants to see what’s going on with the performance’. And we found a very easy way to do it. We constructed our dashboard in an arrow key. So there is an arrow key starting from the top. You can see what the non painted organic has done yesterday. They can see afterwards ‘OK from the paint which channels had the most activity, which channels actually improved’. And we tried to keep it in all of our reporting that they created now to make these dashboards in every part of the company. Every fraction can actually open it and can find something that they can relate to. And I think that this is something that it’s another learning. From this Monday morning with this cap woke up and was not so happy about it. This was with us and they’re more happier now. Actually [15:10] we’re sitting today with the remote and basically we have a very very control with our dashboards. We know exactly what KPIs are under. We are able to fix it. We are actually understanding much better today our data than we understood it before. And this is our vision. To sit there in this huge piled room with a remote on the sofa and just push it when it’s not working. And this is the only option to do it today. With basically using the systems that can go and directly push the data from the database into the reporting. So we don’t really need to work on this stuff. We actually can focus on showing that the data arrives correctly.  We can actually work on beta attribution models. We can work much better. And I think that this is something that allows us to clean the way that we work today.

And in the future what are we gonna do. We’re gonna have a dashboard of 3 KPIs only at the end. Which will be what did you do yesterday in matter of session. What is the other line you drew and what is our recommendation to you. What recommendation. What is the idea of the recommendation process. At the end what we want to do is to replace the analysis part of the channel manager. And this is what we put ourselves as the vision. We don’t download data anymore to Excel. We want them actually to arrive in the morning and actually focus on what is really important to them. And it’s to optimise their campaign. They need to actually understand what they’re doing. And than in case they need to be the business cases we’re gonna have to process it so they’re gonna be able to really easily draw it out and be their own reporting. They’re not gonna need us. They’re gonna be more self served but we’re still gonna government the data that they receive and we’re gonna make sure that the KPIs that they receive are our KPIs. So when we’re arriving tomorrow to the managing boards meeting most likely we’re gonna report on the same number. When we’re saying that Facebook made 50000 sessions we’re all gonna have one number that is 50000 sessions and not 55 or 56 or 150 which once that it was.

At the end I will finish it with saying that our lives are too short. We have 8 hours of day at work. Make it as simple as we can and this is what we learned from all this process. We can make things much more easier for us and for our clients. And we can actually create a change inside the company.

Thank you very much.


Presenter (Lars): OK next on stage will…

Lior Barak


Lior Barak has been working with marketing data for over ten years, overseeing transformations of teams from using manual and nonefficient tools into a state of the art marketing automated reporting systems.

Your email address will not be published.