Rita Lefler is a global BI director at TOMS Shoes. In this podcast, she joins SAP’s Jeff Word to discuss the mid-size company’s decision to move to the SAP HANA platform, particularly SAP Business Warehouse (SAP BW) on SAP HANA. Rita and Jeff discuss:
- Why SAP HANA was a critical piece of the TOMS Shoes mission to bring business intelligence into the company
- What life looked like at TOMS Shoes after the BW on HANA migration was complete
- Where TOMS is headed next, and why SAP Lumira is a big part of the company’s future picture
Listen to the podcast, and read a transcript of the conversation here:
Jeff Word, SAP: Welcome to another episode of The HANA Effect. I’m your host Jeff Word from SAP, and I’m joined today by Rita Lefler from TOMS Shoes. Rita, welcome!
Rita Lefler, TOMS Shoes: Thank you.
Jeff: Before we get started, why don’t you tell us a little bit about yourself and your role?
Rita: I joined TOMS Shoes back in 2013 as a global BI director.
Jeff: You guys have been working with HANA for a few months now. You’ve had one really big HANA project. So why don’t you tell us a little bit about how TOMS Shoes got started working with HANA?
Rita: Back in 2013, TOMS Shoes implemented SAP. But it was lacking reporting. A business intelligence program wasn’t completed at that time. So when we started evaluating how to bring business intelligence into the company, how to make operational decisions in a more efficient way, we were thinking about SAP BW. And we decided that we wanted to build BW on HANA. We constantly have this challenge in the business where we need to analyze data at a very detailed level, and also be very agile with our analysis. We can’t really build very static models. We constantly change the way we look at the products, we add new products, and we thought HANA would give us this agility.
Jeff: So basically, you wanted to look at an enterprise data warehouse, but you needed the speed and capability from the HANA engine underneath it.
Rita: We needed both. We needed the HANA engine capability to give us speed. But we also needed the agility of the BW model, making sure that we can really quickly address new business challenges or business questions without going into the back end and changing our model. So with HANA, and specifically BW on HANA 7.4, there is a simplified approach in the data modeling — a lot of layers got removed. So what we could do now, our time-to-market is a lot faster than it would be with traditional BW.
Jeff: That’s actually very important. Especially in your business that has all of these really critical issues that come up from a retail perspective. You have to be able to see those and react quickly.
Rita: Yes, and it’s not just from the retail perspective. It’s also planning and forecasting, and how do we buy, what models do we have to invest in, and from the operational perspective, even from our own operations, we have to constantly look from the different angles.
There are so many other questions, too. So we want to know how the style is performing in the season, at the beginning of the season. Or how similar styles were performing last year. But because we’re an apparel and footwear solution, we don’t have the same products selling every year. We have new products, we introduce new products, and sometimes we have to look at them differently than we looked four months ago at a similar product. So we need that agility to be able to produce different results and produce different models in our BW system.
With HANA, with BW on HANA, we have this ability. We can really quickly connect the data and with new composite InfoProviders, an ability to join the data on the fly, which speeds up our delivery to the business.
Jeff: So it’s not just about the speed and the flexibility of HANA, but it’s about the speed and agility that the business users now have, and the flexibility that they have, to ask the questions and dig deeper to solve the business issues, yes?
Rita: Exactly. It’s that agility, and reducing that development lifetime.
Jeff: Can you give us an example of the before world and the after world once you went to BW on HANA?
Rita: In the before world, with traditional BW, we would have a really basic model. So we would load the delta into the DSO layer. Then we would create cubes, and then we would create multiple cubes on top of the DSO, so each one would be aggregated at the specific function. So the way that you would have, for example, a dashboard, you would have a specific cube attached to that dashboard for the improved performance of running that.
And then now, when you have multiple cubes in the same data, every time you want to change something, the underlying data, you have to either reload the data, you have to change all your transformations, you have to make sure all your cubes everywhere that have an impact in the chain, have to be changed. So it’s a long process to just make some small changes on the back end.
So now with BW on HANA, first of all, we don’t have cubes anymore. We have just one layer, because it gives us the speed. And then we now have this agility to create virtual providers, which do not store any data. We can join the data virtually, so it’s really easy to just have one layer where the data is stored and if it needs to be changed, you change it in one place only. And then you virtually can produce results that you need for the business.
Another benefit of HANA and speed is the data loads. So if the data loads before would take eight or ten hours, now all of our data loads take within two hours. And certain InfoProviders we can load within minutes.
So what does it do for business now? So let’s say during the month-end close, users want to refresh the data during the day. We would never be able to do that before HANA. Now we can say, hey, OK, we can refresh the data, it takes a few minutes. No impact on the business. Nobody really even notices anything, and they can have data refreshed. So, the thing is that we don’t have real-time replication; we don’t have HANA with real-time data. But, even with BW, we’re almost able to provide them near-real-time results when they need it. It’s on demand. We’re able to refresh data on demand.
Jeff: That’s a huge, huge shift from the nightly batch, ETL, loading, and things like that. And then you come in the next morning and massage and restructure before it’s available. Now you guys can do the actual physical import very, very quickly, which allows you to spend most of your time on making the data more useful to the business, right?
Rita: We still do nightly batch reloads. We want to keep data consistent throughout the day, so our loads are still nightly. They’re shortened. So now instead of 8-10 hours, our window is two hours. In addition to that, in a specific situation or exception, when users want the data to be refreshed, in an exception situation, we could do it really quick. Or, let’s say we have an exception situation when the data load failed during the night, and we come in the morning, 6 or 7am in the morning, and we fix that, and we want to reload data, the data is available almost immediately after we fix it. In the traditional, old BW, without HANA, users would probably experience some downtime. Because we would fix it, and then they would have to wait for the load to complete.
Jeff: Let’s talk about the size of this data problem. Can you tell us about what the size of your warehouse is today and the data volumes that you’re dealing with now in HANA?
Rita: In HANA, we have currently 250GB. We have a medium-size HANA appliance. And we almost reached the max, so we are thinking about implementing next either near-line storage and also upgrading to more HANA memory.
Jeff: OK, so you guys aren’t really a massive big data problem. It isn’t necessarily that you have lots of data, it’s that you need the data quickly and it needs to be flexible in order to do the types of analysis that your users want.
Rita: So again, we’re a medium-sized company.
Jeff: Tell us a little if you can about the project, when you rolled out BW on HANA. What about the resources and the types of activities, and maybe also how long it took and some of the interesting things you found out during the process?
Rita: So we have started our project, migrating on HANA, in the beginning of 2014. Our challenge was that we wanted to migrate our existing BW system on HANA, but also bring up and develop new content. And we had a specific deadline for the new content, so we needed our plan to flow with migration and development and meet our deadline. So the way we decided to do that, because we couldn’t freeze our development environment for too long, we installed HANA on a sandbox. We spent about a month migrating to the sandbox and testing. And so the challenge we experienced at that time was that BW on HANA was so new, and the support pack was lower than we needed. So we basically spent a little bit more time to get onto the highest support pack. A recommendation for anybody who would go the same route, before you do it, you have to be on the latest, latest support pack for BW on HANA.
Jeff: That’s really good advice, and something also SAP highly recommends. Can you give us an idea of how long the project took?
Rita: It took us to do all the testing and the sandbox, implementing the sandbox, and making sure all our migration was successful, and all our reports are tested in the sandbox. And then, when we brought development down for the migration, it took us two weeks for the full HANA migration, development, and QA boxes, including the testing. Special focus was given to regression testing. And in this period, the migration itself would take only two days, but the testing is where all the time was spent.
We had to test all the reports, the data loads, and change management, moving objects from development to QA, which was really important, and object development. So you could find some issues where your reports are OK, but when you are trying to move the BW objects from one environment to another environment, you may find some problems, and we found a couple OSS notes to fix that.
After the two weeks, we just had to wait for the right time to move into production. So we actually did the production migration to HANA over the weekend. To all the users, it was treated like just a normal system upgrade. So we never ran in a parallel environment. We only had production unavailable for the weekend, Saturday and Sunday. But you also have to consider that our database size was not as high at that time. And development was only frozen for two weeks. That’s what it took us to get from development to QA and tested.
About resources, all of it’s been done with a small team, with one internal resource and two consultants. And with those two consultants, we also were developing new content, so it wasn’t just a HANA migration, but we were also doing data modeling, and building and developing new data models.
Jeff: So this was a pretty bare bones team that did a huge amount of effort in a very short of time.
Rita: Yes, so that’s the BW team. We had another team within IT, who were hosting our hardware. And so we had the team there that was a Basis and database team who actually did the database migration.
Jeff: So what kind of things can you tell us that may have been surprising to the team as you guys were going through that project?
Rita: A couple of things were surprising. We had some setbacks, because we needed to upgrade to a different support pack. We had to go through an upgrade of support pack in NetWeaver 7.40 first before we did the migration. We had to go in and upgrade the whole landscape to the support pack and then later on we found out that we needed a higher support pack, so we had to go through the whole upgrade again. But I don’t think this would be a problem now because BW on HANA was too new at that time. I think now you could safely go to the support pack once, and then do your migration as a second step.
The other surprise was that after we went with BW on HANA, was learning how the HANA database works, and how memory allocation works. And so your Basis team would have to do a good job to make sure it’s all set up properly, also making sure that you do the system clean-up before you go to BW on HANA, making sure that your tables maintain only the data that you need. So making sure that all the log files are clean, and keep it clean, so you don’t get data that’s not needed in the system. Also making sure that’s what’s stored in the column store table versus row store tables. So all this has to be taken into consideration for better performance and better use of the appliance that you have.
Jeff: That’s really, really good advice. And I might also add that since the time that even you guys went live, there’s been a huge additional development in the Cookbook for advice for going through these projects and preparing for them. For listeners that are interested in this, there’s a whole page on that on the SAP community, on SCN, that’s a great place to start to get ahead learning about these things.
So let’s move on to, now that you guys have this up and running, and it’s been going for a while, where do you see TOMS next evolutionary steps in the HANA world?
Rita: So now when we’re on HANA, this is just the beginning. In our evolution now, we want to get more into visualization and business intelligence. We are installing Lumira right now. We are playing with Lumira and how we can start telling our data discovery stories through the Lumira visualizations and having users explore the data through the fast ability on the HANA database.
Jeff: So going way beyond the standard reporting into visualizations and other really high value-added data discovery?
Rita: Yes, we’re going into data discovery, and we are enabling faster time to insight and analysis and accelerating decision making. We also know that we have the ability to replace some complex data models with new ability, with built-in composite InfoProviders, so we’re taking full advantage of that.
The next thing we also want to do is have some insights in real time, and we can build some data model on HANA that will be independent from BW schema.
Jeff: OK, that’s phenomenal. What are you guys planning on the more transactional side, with ERP?
Rita: With ERP, we would like to build some analysis, especially for finance. We got a lot of requests from finance during the month-end close, when they analyze the data and the line items, and we need this in real time. So we would build some financial reporting with a deep-dive analysis for finance as a first step. Then we could see some of the customer analytics so that we can connect to our e-commerce database and run some customer analytics to get a 360-degree customer view. That would be really impactful.
Jeff: Any other words of wisdom you’d like to share with other companies that are considering going down the BW on HANA route?
Rita: I’m glad you mentioned the cookbook because we actually used it, and it was really, really helpful. That’s something that should not be ignored. The team experience is invaluable, especially Basis team experience. Because the way the HANA database is managed is different from other traditional databases. So you want somebody who has experience and can help you get through that and through the system migration.
In short, the hand off of responsibilities should be understood by your infrastructure and Basis and BW teams. Even if you look at the Cookbook, there are different tasks done by different functional groups. Some of them have to be done by the BW team and some Basis and some infrastructure. That needs to be clear and understood before you start your implementation.
Jeff: That’s absolutely critical and very, very good advice for other folks considering this. So I want to thank you so much, Rita, for joining us today. This has been phenomenal.
Rita: Your welcome, it’s my pleasure. Thank you for having me here.