Bosch’s Data Journey: From Chaos to Transparency – with Wolfgang Klein & Simeon Rilling, Bosch Digital

00:00:00: I started to understand that, hey the data products are something that brings semantic information.

00:00:07: That's then the foundation of AI can do their job and this was a momentum where the importance of data product and having a database really changed dramatically.

00:00:35: Hello and welcome to The Data Culture Podcast!

00:00:38: I'm Karsten Bange, founder & CEO of Barg and with me today is Florian Wiegelmeier And that tells you it's a data festival edition of this podcast again.

00:00:47: We invited two speakers from the upcoming Data Festival, Wolfgang Klein, director, data as-a-service and Simeon Willing chief product owner for The Data Mesh Ecosystem.

00:00:59: They're both at Bosch Digital and they are driving their data related IT landscape including data lakes, data products in much more for Bosch worldwide!

00:01:11: going on at Bosch, especially around data products.

00:01:14: I thought we found very interesting insights didn't we Florian?

00:01:18: Yeah absolutely Karsten!

00:01:20: I mean Bosch wasn't the situation as many big companies they had like plenty of data already available a few years back.

00:01:28: but there were really siloed and Wolfgang and Simeon Versukind to lead us through their journey.

00:01:35: how they took that challenge.

00:01:39: They talked about data ownership.

00:01:40: What does it need to make data available from a cultural and organizational perspective, the organization?

00:01:47: And we also talked about challenge of data products or the concept of data product which is really helpful for them To Make Really good chunks Of Data Available in The Data Catalog In A Data Marketplace To Make It Easy To Consume Than For The consumers in the entire Bosch organization, which is huge as you can imagine.

00:02:10: There's a lot more that we talked about but I think it's time to dive right into our conversation!

00:02:17: Hello and welcome to the DataKaiser Podcast Wolfgang and Simeon from Bosch.

00:02:21: It's really great to have your here.

00:02:26: Uh...I'm very glad too.

00:02:27: i am really happy To share our data journey at Bosch today And really happy to be on this round.

00:02:36: Thanks a lot, Kasten.

00:02:37: Really looking forward to our talk today and I'm really excited also share with you some of the insights that we gained on the journey that have already mentioned.

00:02:47: Absolutely!

00:02:48: We obviously talked about it up front data chaos to an ordered world where data has its place, everyone knows what there is and so on.

00:03:01: So let's talk about this journey.

00:03:03: I think it's super interesting.

00:03:05: first of all maybe could you give us an overview?

00:03:08: What Is This Journey?

00:03:09: Where Did It Start And Where Are You On The Journey?

00:03:15: Maybe Let Me Just Start And Then See Me.

00:03:17: Maybe You Just Continue the Journey.

00:03:22: There were two things.

00:03:24: First of all, we are all in silos data.

00:03:27: silos I mean We're always thinking really in specific domains and mainly driven by the business units By our different entities like mobility or power tools or automotive aftermarket.

00:03:42: Never one was having his own data silo.

00:03:46: It was also that times off the SAP Business Vary Systems Everyone region, every business unit they had their own system and just focused on their own stuff.

00:04:00: And this was good for them because it was stable.

00:04:04: They could do their monthly reporting and so on.

00:04:07: but On the other hand This was very much siloed.

00:04:10: It's not flexible at all.

00:04:12: Besides those silos there were a lot of things going on.

00:04:17: So For me it felt like everyone has his or her database under his or her desk and just doing their own data stuff.

00:04:27: Not being aware about the others, not really knowing where to get the data from.

00:04:33: And I mean as a result everyone was busy with getting data form somewhere putting it in His own data silo and then doing some data stuff not Knowing what others are doing.

00:04:45: maybe others were doing The same stuff Just next door Or In A different entity And basically, of course this was not efficient at all and transparency was missing.

00:04:59: We were duplicating data so many times!

00:05:03: This is really crazy time.

00:05:06: I think that's the time when Simeon stepped in... ...and maybe he could just continue.

00:05:13: Yes of course thanks a lot.

00:05:16: The interesting part about this journey fits perfectly to also my career.

00:05:22: I'm at Bosch because as Wolfgang mentioned, back then i joined as a SAP BW consultant and Then kind of to overcome this um in transparency And all the the caves that we have.

00:05:35: We started with a data lake approach?

00:05:38: And I think as many other big companies Um we were successful to a certain extent and of course you Kind of Get to know a lot of things also about technology, but also your company About data governance about But also about scalability and at some point we just simply faced the problem that me as a central Technology provider for whole Bosch could not scale fast enough by our capacity To keep up with the hunger for data did all customers have And I think this was like a turning point also for us.

00:06:19: We said, yes what's the next step from our data lake approach?

00:06:24: Of course then DataMesh came up luckily and since then we are building up a DataMash approach over all Bosch which of course is quite a task because it would be huge size of the company that comes along

00:06:41: with it.

00:06:42: What is kind of interesting in that story, just by observing the technologies you use.

00:06:49: You came from a siloed certainly but very structured way to look at data in BW and then shifted it away where all your freedom can do whatever you want with this data which in some cases ends up in a swamp if we don't do right And then went back to like a very domain oriented approach, which is data mesh.

00:07:15: Is that just because so-to speak there was the trend?

00:07:19: To do that and it was like a good way to do it for

00:07:22: you?

00:07:22: or is this something that reflects maybe also the journey little bit?

00:07:27: I think It's a little bit of both them basically started with demand coming from central logistics and central finance.

00:07:39: Basically, both of them had the requirement first to have all the finance or logistics data from all business units in one place.

00:07:50: To get an overall view on supply chain and their financials which basically was a need that we could not fulfill with SAP BW systems.

00:08:06: I think this was more or less simultaneously, so we started to build it up with the data lake approach.

00:08:13: And then obviously as Simeon said at a certain extent It's great but running into scaling issues and Then there is time when the mesh approach came up We had a lot of interaction with our corporate data architects and data management teams and then basically bringing in the trend from the market, but also defining for us what does it mean for Bosch?

00:08:47: So implementing a mesh.

00:08:49: And coming to point where of course it needs technology foundation But its more over an organizational impact that we need.

00:09:06: Basically, we have to change the mindset of some people in the organization dealing with data.

00:09:14: To incorporate this domain.

00:09:16: thinking right?

00:09:17: And that they are sharing approach.

00:09:19: absolutely so welcome to The Data Culture podcast.

00:09:22: That's exactly what you see all over place and everywhere.

00:09:27: So that makes a lot sense.

00:09:30: Before we dive little bit deeper into different aspects I would be interested in the drivers behind it.

00:09:40: You mentioned siloed lack of accessibility, lack transparency but on the other hand sometimes people like silos because they don't have to align with others that can kick their data.

00:09:52: so how How did you manage, in a way to get this all started?

00:09:57: Is there let's say a clear commitment for example.

00:10:00: To accessibility transparency of data and the business strategy?

00:10:04: or is that a clear need?

00:10:06: or was it a compelling event something happened where everyone said okay we cannot go on like.

00:10:13: so how Did You Get The Momentum For This Journey?

00:10:17: To be honest, I think it was more of like a process as well.

00:10:20: So i can't really remember the one event where we said okay now We have to change everything and um but It's more like when you are discussing with our counterparts in The business side what Are your needs?

00:10:33: What are your requirements?

00:10:35: And these obviously changed over time and Um we heard A lot More about.

00:10:40: You know we need to Be faster.

00:10:42: when you're being more flexible and i Think this Obviously, it becomes more important than ever now in the difficult times that we're in.

00:10:51: And there we saw... ...that the data lake approach does not really fit to these requirements anymore.

00:10:59: and also from a technology perspective We have to evolve ourselves From this centric data lake approaches More into like a data fabric that gives our customers and the respective teams a maximum of flexibility in speed.

00:11:16: Because at end-of-the-day only if business is successful, we as IT can be successful too!

00:11:20: And I think these joint forces are having close discussions with their customer which aren't always easy to be honest but this really key for success also convincing people look into the same direction like really took a couple of years now because in such big company, there are lot different views on different things.

00:11:49: And everybody has always the best intention to bring their company forward.

00:11:54: but I would say that it's only discussions and only comment before what is helping us.

00:12:01: But maybe you can also add little bit from your perspective.

00:12:07: During this journey, sometimes I was really wishing for just a top-down decision and then we just implement it.

00:12:15: But typically that's not how Bosch works.

00:12:18: The business units have a lot of autonomy.

00:12:23: They are responsible to drive their businesses in the area.

00:12:29: they're free to decide on how they achieve their targets.

00:12:34: It is more convincing.

00:12:38: As I mean, Simeon had plenty of talks with the stakeholders in the business units.

00:12:44: Me as well and then bringing ideas coming from corporate, from architecture technology IT provider view.

00:12:54: bring this together to see a business unit you and getting the commitment that working federated approach makes more sense for all of us.

00:13:09: I mean, this really took a while but at the end i'm convinced it's a more sustainable approach.

00:13:15: Did AI play around?

00:13:17: Was that a driver ?

00:13:19: I would say yes and especially now The hype in the last couple years about Chennai was really opposed to Frost because like also the sea level management realized For the again how important data is because if you do not hand your data probably it's just trash and trash outs And you don't get much of gains from AI or Chennai.

00:13:43: Um, and also when it comes to topics like governance and giving Data a business background I mean this is probably obvious too all of us guys who are working in data environment for couple years now but Approaching this and giving data business a background, and also changing a background to interpret data.

00:14:05: I think that's the key for success at the moment!

00:14:08: Obviously here in the very center of what we're doing is establishing common sense on how should we handle it?

00:14:22: And maybe also what is not a data product.

00:14:24: I think this probably discussion that's going on in many companies and we did not want to miss it out, of course.

00:14:32: So i think one key discussions you're currently having?

00:14:38: I was just waiting for the kind keyword about their products because its so well matching with your planning.

00:14:49: So a lot of people have like, A lot of ideas what data products could be right now and whatnot.

00:14:55: And certainly two directions are to see it a bit more from the value perspective.

00:14:59: on The other one a bit More from the technical standardization perspective.

00:15:03: I don't want to get too prescriptive But is there anything specifically about Data Products?

00:15:09: you're thinking About how How your thinking On this kind Of Concept, what is like the most valuable part for you of this concept on this journey?

00:15:21: That would be interesting.

00:15:23: I mean we were starting to talk about data products and how important they are from our point-of-view And i felt Most of them did not really understand The purpose in the value add of a data product.

00:15:38: They said yeah that's something IT driven and it's something About standardization.

00:15:42: why do We need that And why do we need data products?

00:15:46: I mean, you can just use the AI and the AI will do everything.

00:15:50: Just give answers to all our questions... ...and I felt then sometime back that's not too far.

00:15:58: in the past it really did a change in the organization and perception of how important is this.

00:16:10: quality data products in the organization as a foundation for AI.

00:16:15: because I mean if you just feed the AI with just raw data, how should they do the interpretation of that?

00:16:25: That's when the organization started to understand.

00:16:30: Hey!

00:16:30: The data product is something which brings semantic information through data and then it's the foundation where the AI can do their job.

00:16:41: I think this was the momentum where the importance of data products and having a data marketplace really changed dramatically.

00:16:50: So in my personal opinion, it's probably both... Of course you need that business value because only if we have business value people will implement it and they'll take care about your data product Because everybody sees there is an investment needed for me to build a data product.

00:17:09: so what does the return on invest from?

00:17:11: And I think this is a really key question that you have to answer.

00:17:16: if you want to push data products in the data marketplace, and then of course like spreading You have to be a bit pragmatic if you approach data products because of course, is like a view on data that everybody, so let's have look at this.

00:18:03: Let's publish it!

00:18:05: Let's look how it works and bring to marketplace give transparency.

00:18:11: then of course you need also these more scientific approach defining company level what attributes it has but I think for me its more like a phased thing.

00:18:23: I think that for us, like a data marketplace was key turning point.

00:18:32: Because it gives transparency what is already there and For us at least It also worked out pretty well That we did not start with the most complicated Marketplace ever But just simple web application where you have tiles With your data products Where can click on you're taking also the data domains and on a journey with us, developing further from there.

00:19:04: And at that moment we are looking not only in marketplace itself but like other topics around data products like data policies carrying all of regulatory information needed which become more important than ever and continue to catalog strategy because change the way how we think about data assets and data catalogs, like foundation for data products I would say.

00:19:35: And then of course integrating it more deeply in our enterprise IT and also business processes.

00:19:41: so this is all next steps that are currently working on with quite a high pressure because

00:19:52: Could you elaborate a little bit further, what do mean with the relationship between data catalog that changed?

00:19:57: So data product, data catalog meta-data marketplace.

00:20:03: Obviously it's all intertwined in a way but super interesting to hear the effects on the data catalog made.

00:20:10: for my perspective The interesting part is if you want to approach a data products and I think Quite easy to argue why you should look into data products follow usability transparency.

00:20:22: You can have governance, but then the next step is of course what do I need?

00:20:28: To build a data product and there you come two data assets an you come through the question who owns these data assets.

00:20:35: And What Is This Data Assets All About?

00:20:37: and um i think it's quite obvious that if your data catalog is not maintained well and you just have like, maybe the technical SAP table name there.

00:20:52: It's pretty tough from there to start to build a data product.

00:20:56: but if it's well-maintained then do we have a description?

00:21:00: If you know who owns that data... Maybe there was also a DataStuarts already maintained with deep expertise in what this table means.

00:21:11: I think much easier to build up a valuable data product and high quality data products.

00:21:19: And so I think for us, the shift is really that it's not enough just have simple data catalog as stand-alone application but its more like an environment or out of a data catalog link closely with your marketplace regulatory information like what are you allowed to do with the data and not allow it together in one platform.

00:21:49: I think that's really currently changed, like Simon mentioned before we had a data catalog or several data catalogs for each technology platform.

00:22:01: We have some data marketplace And maybe somewhere else we had some systems to define who has access to which data.

00:22:12: Now bringing these three components very close together, making them as an integrated system and just seeing this from a holistic view I think that is the biggest value add in this area.

00:22:27: This gives us lots of transparency.

00:22:31: so which data is where which data products available and they are based on which of the data assets.

00:22:37: And who is responsible?

00:22:39: I mean, whose responsible to keep that data product up-to-date?

00:22:43: Who's responsible To ensure a certain quality level for the data Assets?

00:22:50: and then at The end who should get access to Which part Of the Data and what Is here her allowed to do with That?

00:23:02: I think this adds a lot of value, and i think that is now more understood in the organization.

00:23:08: Maybe it's time again to zoom

00:23:10: out

00:23:12: after talking about interesting details on how you made progress both from the organizational part or technical parts?

00:23:21: When we talk data products... We just had a survey which was published two weeks ago And found those who established Data Products company-wide are much further ahead in terms of their agenda, AI and also other AI projects.

00:23:37: Is that maybe putting it as a question?

00:23:40: Also away how to bring forward an incentivize data owners?

00:23:46: think about data products and put there investments in the air.

00:23:50: And do you have already positive returns on this journey?

00:23:57: I would say what you have just mentioned, like using data products for GNI is probably the most tangible return that we see at the moment.

00:24:05: And from... and i think The Best Example is probably talk to Data where of course if You Have Data Products which are kind of standardized in their technical implementation but also In Their Metadata Like The Business Description Of Course This Is A Very Good Foundation.

00:24:27: Talk to data not only on a single data product but also to be able to skate it, which is then obviously more in our focus.

00:24:35: so at the moment.

00:24:37: Our idea was to say let's apply this talk today they're already in and the data marketplace itself.

00:24:45: I'm sorry that people can go to my data marketplace find the data products of business insights are looking for and let them all ready there.

00:24:53: start your talk with the data.

00:24:56: For me, actually this is kind of a milestone when it comes to data and data analytics because in previous times you always needed some kind of IT expertise.

00:25:07: You know build a Pobii dashboard or write a SQL statement To get the data and derive decisions from there.

00:25:17: And now we open up the whole lot of data.

00:25:19: basically everybody who's able type something into a chat.

00:25:24: And I mean, that's obviously everybody in such big company and this kind of gives us scales data like for every employee in the company.

00:25:36: That is really big.

00:25:37: so i'm looking forward to what we will be seeing there.

00:25:40: also if you look at market how fast it develops Yeah, kind of make it available for everybody.

00:25:53: Absolutely maybe can share a little bit numbers.

00:25:56: how many data products are on that marketplace and may be which domain has the most?

00:26:02: And How Many users do you have?

00:26:05: so overall like I think we Have to give a Little Bit of background Of our how We Are organized because It's not Everything at Bosch is organized in data domains.

00:26:18: And so, of course we have a governance when it comes to processes and alpha-data domains but also as Wolfgang mentioned earlier the GBs where again with the money is made are off course.

00:26:33: one other dimension.

00:26:35: So um at the moment We have around sixty to seventy data products there I would say company master data.

00:26:45: I mentioned the cost center, for example earlier which is important that everybody can use it and doesn't have to start from scratch to build it up.

00:26:54: but there's also different GBs.

00:26:57: they are of course on a journey when it comes to data products And some of them are little further.

00:27:03: Some of them may be having to catch-up in the future again But we definitely have our front runners there, it's really great to see how fast development was from like the initial discussions until publishing twenty-twenty five data products.

00:27:25: So this is why as I said overall sixty seventy around that number of users at the moment is still a bit tough question to answer because we don't always see.

00:27:44: So of course we provided that it might be used from other technological platforms, which also then again have like a lot of end users.

00:27:53: That's really tough question to answer at the moment but maybe you can do some research until we talk again?

00:28:02: Let's do that!

00:28:05: One other question that often comes up, you separate basically the effort to create a data product from their usage where they're often value and benefits are created.

00:28:15: And some organizations started now think about maybe like building structure or how to reimbursed the efforts of producers of data products.

00:28:27: have I recently had a discussion or we have in the podcast, yeah.

00:28:37: We had Fiona Kambais from Siemens and she said well it's not a problem because we have top-down decision that data products has to be implemented Yeah?

00:28:44: And no one is asking about reimbursement.

00:28:48: So how was it at Bosch?

00:28:49: you mentioned very decentralized company.

00:28:51: Is that the dispassion you have for us?

00:28:52: just everyone happy to share our data?

00:28:55: input into effort produce data product.

00:28:58: This discussion definitely behalf and we had several requests.

00:29:03: can we add the price tag for usage of a data product, so that one who is providing a data products make profit with it.

00:29:12: But basically... We decided to say sharing data and making available data to others should be free.

00:29:22: To foster the usage as much as possible.

00:29:28: In addition I have to say that OSH is complex company worldwide with many different legal entities.

00:29:38: And when we started to discuss about this also some corporate departments, I mean first questions came regarding tax relevance transfer prices

00:29:50: etc.,

00:29:51: so he said no let's keep it for free to boost the usage of data products.

00:29:57: and at the end if you think further I mean, Siemens mentioned currently it's a prompt is integrated in our data marketplace so that everyone can just enter some questions there and then get answers based on the data product.

00:30:14: And then the underlying data.

00:30:17: we have some kind of OSH internal JetGPT which was used across the whole company.

00:30:24: basically next step here going now as to connect our data ecosystem and the GNI agents which are able to interpret data products, then also have access to underlying data.

00:30:37: I mean connect both of them so that at end any Bosch user worldwide can just use the let's call it Bosch internal JET GPT And ask questions from there whenever its a data related question.

00:30:55: it's connected to our data ecosystem and then the data marketplace, to the data products into the underlying data.

00:31:02: And I think this will boost the usage of data products even multiply the benefits which we are providing.

00:31:12: but if you start talking about charging the usage for a product that gets more complex let us not do.

00:31:24: Most of the data providers are really proud to share data products and make it available for others as well.

00:31:32: So I think maybe add two aspects because at end-of-day our mind is quite simple, if you question every time we want to use data will be charged... ...I'm not sure that's contradicting with your idea about sharing data or the initial idea of a product itself?

00:31:54: If you look at our organization, we would probably charge from left to right-to-left.

00:32:00: At the end of day it's one company and probably everybody brings in some data products.

00:32:05: Everybody consumes data product.

00:32:07: so I guess that overall they will balance out nicely.

00:32:12: And then of course the Christians always how much organizational overhead do you really want bring into this?

00:32:23: In this example quite nicely, shouldn't it be just fun that you consume data and that you derive business insights out of it by being able to chat with not thinking about how much my customers will charge for.

00:32:38: So overall I think at least for us its the right way more to go in direction of being proud to share data products.

00:32:49: It

00:32:50: already

00:32:51: paths perfectly the way

00:32:52: to

00:32:53: looking into next steps, how things are continuing at Bosch.

00:32:59: So Wolfgang Siemer we're really happy you have a data festival in June and Munich as speakers on our stage where you dive for little deeper especially the product parts of it.

00:33:16: Well, your next steps both of the sites off developing organization bring more commitment to ownership.

00:33:23: Maybe also on a technical side.

00:33:24: what do you think is only road map right now?

00:33:28: What?

00:33:28: we will definitely show us more details about how our data mesh approach look like and how be kind of also adapted some aspects of it through hour I would say truly due their reality that we face in our everyday life.

00:33:46: And of course, also in addition like our journey that is Also more going to the cloud I think as many other companies as well and it's really amazing To see the very fast development sometimes.

00:33:58: It'll be terrifying too Be honest but you know all the all the new capabilities on your functionalities That come up there?

00:34:05: And i think it's Really important to Think ahead how You want to approach this especially from technological perspective.

00:34:14: end Of course.

00:34:16: What we are heavily looking into is how can we apply Chennai.

00:34:23: We talked about it a lot now, and our end users can use Chennai.

00:34:29: But of course also How Can We Optimize Ourselves, Our Data Fabric Using Agents And Applying It On A Bigger Scale?

00:34:39: Because From My Perspective This Is Where A Lot Of Efficiency Gains Come from.

00:34:47: I think you covered it already very well.

00:34:50: Maybe just to add one aspect, our vision is maximise the value of data for Bosch and so far we are mostly looking on the commercial data part And basically next step is also to sync data in a broader view.

00:35:09: So there's also data coming from shop floors So all this kind of manufacturing data and machine data.

00:35:17: And there's also a lot of IoT data from devices, e-bikes et cetera.

00:35:25: in addition There is also engineering data.

00:35:29: The thing to take our framework or concept the OSH Data Fabric approach then see how we can extend it through other data tapes as well and making the access to data as easy as possible.

00:35:46: Of course, always keeping compliance in Data Security, Data Protection in mind that's for sure but Making the Access To Data As Easy As Possible And Not Only For Those Users Who Are Familiar With Power BI And Tableau And All The Other Tools But Also For Those Who are Just Able To Type Their Question Into a Prompt Or Even Just having an app on their mobile phone while being on the way to our customers, then just asking questions.

00:36:17: Hey can you please tell me the turnover of the customer where I'm just driving through from last month?

00:36:23: Which are most important products and please propose some strategy?

00:36:28: which kind of discount counts i should give this customer increase my contribution margin?

00:36:35: right

00:36:36: Excellent, that sounds really exciting.

00:36:38: So good luck on your journey!

00:36:39: I'm looking forward to seeing you in June at the Data Festival and hearing more about your journey from data chaos to accessible transparent data.

00:36:52: See ya there,

00:36:57: bye-bye.