 Hello, and welcome. My name is Shannon Kemp and I'm the Chief Digital Officer of Dataversity. We would like to thank you for joining this Dataversity webinar how a semantic layer makes data mesh work at scale sponsored today by at scale. Just a couple of points to get us started due to the large number of people that attend these sessions you will be muted during the webinar. There are questions we'll be collecting them by the Q&A or if you'd like to tweet or we encourage you to share highlights via your favorite social media platform using hashtag Dataversity. And if you'd like to chat with us or with each other we certainly encourage you to do so. And just to note zoom defaults the chat to send you just the panelists but you may absolutely change it to network with everyone to find the Q&A or the chat You may click on those icons found in the bottom middle of your screen for those features. And as always we will send a follow up email within two business days containing links to the slides, the recording of this session, and any additional information requested throughout the webinar. Now let me introduce to our speaker for today. I live to to I live is the global head of product for at scale, her most recent role prior to at scale was was as clicks vice president of innovation and design overseeing a global team of user experience designers, product designers and engineers, her innovations have led to patents for search and conversational analytics data analysis data management and more. Her research and technology development for augmented intelligence, a combination of data science and AI has led to the rise of third generation analytics. ALF is also a founding member of the innovation forum at forum ventures a leading venture group investing in early stage software as a service startups. In this role she serves as an advisor and mentor to co two founders and executives. She recently won the women in tech outstanding leadership award in recognition of her outstandingly leadership comfort contributions to the cloud industry. She was also named a winner of the business intelligence groups, artificial intelligence excellence awards program, acknowledging her work leading the charge to blend AI into analytics to further the AI human interaction with data while striving to eliminate bias. Very impressive bio and with that I will give the floor to a left to start today's webinar. Hello, and welcome. Thank you very much and it is great to have the chance to be with the audience and together with you to talk about data mesh and how a semantic layer makes data mesh work at its scale. So, I would like to start first of all that data mesh is still a relatively new concept. Well, I think it's beginning can be tacked back to 2019. But what I'm recently noticing is its popularity has especially grown over the past two years so I think this is a great topic to cover with this audience. In fact before joining to the session I was curious about the Google trends and I just did a quick analysis and it actually Google trends show that around a 300% growth in search for data mesh volume over the last 18 months. So it's obviously there is you know more and more interest around data mesh. And in the session I will be talking about why I see semantic layer is an enabler enabling technology for a data mesh strategy so with that let's get started. So, I think the first thing that I would like to cover is, you know, is a relatively new term and concept. I think there is some confusions and, you know, maybe it's sometimes hard to find a generally agreed upon definition of data mesh although I would say that in the last couple months there is more agreement on the on the definition. As far as I think there's agreement that data mesh is different than the, you know, it's not the same with data fabric. So, for me data mesh is really it's about creating an approach for organizations where they can build a decentralized analytics architecture, where really the ownership of the data is given to a business domain and business units. So that they can actually create value driven data products. And, and I think that really makes sense. And I will be covering in the rest of the session like some of the challenges that we have been seeing in the, you know, in the in the life of data, just giving this ownership to the group of users who are the closest to and best understand the data and their business needs just makes a lot of sense. And for a successful data mesh, I think there are a couple things that all organizations needs to pay attention. The first one is the technologies that you are using to enable data mesh and it can be a group of set of different technologies. They really need to kind of need to give the flexibility and agility that is needed to be able to apply new business rules and logic to the data. The second thing, and the, to enable really a data mesh is, you need to make sure that there is central governance, and so that you can ensure a single version of the truth, while you are giving the ownership of data to business units. But that central governance also requires a federated ownership of data for the business units. And then the final thing, this is all about user needs. And I will be talking more about, you know, data products that is, you know, where you need to have a user driven design thinking. That's really kind of requires to be able to abstract technical complexity from the user so that they can really, you know, flexibly work with the data to be able to answer their business questions. So I think this is kind of like a recap for me for the definition of data mesh and then the things that I will be highlighting during this presentation. Before we get into talking more about data mesh and the application of semantic layer to enable data mesh, I just want to take a step back and really, you know, walk us through about like orient ourselves like different types of data that we have been always doing in in organization so when we take a step back really the world of data is divided into two spaces. The first one is the operational data. And this is the data that is sitting behind your applications behind our services or the API's and this data really reflects the current state of the business and usually operational data is where we capture all the transactional data and then this is where we are really optimizing the infrastructure for this type of data sets for application logic like really this is where we are running a bunch of create update and delete operations. So this is all about you know how we can keep the transactional view of the data. On the other hand, you know, to be able to monitor the data we all need, you know, we all have business questions like business users right this is where we have the need to have analytical data and analytical data really is to optimize that data to transform that data so that it is optimized for analytics, where we can run queries do ML analysis and ML training on the data to generate predictive and prescriptive insights and this is where we are really transforming the shape of the operational data transactional data to become more multi dimensional. I would like to highlight the importance of multi dimension analysis because you know business users as humans. This is how our brain works right we don't just as a question about the KPI like what is the sales but we really would like to break that down by different dimensions such as region warehouses product customers. And also the other thing about the analytical data to be able to optimize for analytics. We are really capturing a big amount of data to be able to do historical analysis on that data and also use that with the amount training to the predictions. When we think about those like two different spaces of data, the place that is really this whole this approach or this definitions become very fragile is the how the data flows from the operational data sources to analytical data sources. And this is where we have been doing ETL. And so, and then I think we have seen different trends over the last 20 years in terms of how we approach ETL. So let me cover that very briefly. So when we look at the data journey and how we do the ETL I think two decades ago we have, you know, had all get excited about the data warehouses which we are still excited about it right this is where we capture create our data in the next few months to be able to provide analytical analysis like on the data. But the, you know, the journey of data transformations has started with you know, extracting data, and then based on the business users analytical questions, transforming and then loading it to a data warehouse so that the data can be accessed via the BI tools and APIs to be able to answer business question. Well, this approach, although it was very effective, very quickly it just become a bottleneck because one of the main challenges with this approach was, as you all know, business questions change. And now, especially right now, given the environment that we live in has become very dynamic, there is always something new around there's the war going on or supply chain issues, the business questions are changing literally early. So when you think about how the data needs to be transformed and loaded to a data warehouse to be able to answer the business question, this structure is not really agile and flexible enough to be able to reflect the business moment so that the data can reflect the business moment to be able to answer the question. Well, then I think in the last 10 years, especially with the technologies that we are seeing, moving the data from on-prem to cloud data storage, we decided to take a different approach. And then I think the whole concept of Data Lake came into place. And the idea here is that with the cloud data storage, as the storage becomes cheaper, the idea was, okay, let's load all of that data out of the operational data stores and then store it in Data Lake. And then with today's flexible cloud environments, it really make it possible for data to get transformed more easily. So once the data is loaded into Data Lake with less coordination with infrastructure teams, data engineers, and by the way, we come up with this concept of data engineer in the last probably 10 years, now they are able to script and automate transformations with tools like dbt, airflow, or another data warehouse automation tools, and then orchestrate and automate transformations directly into the Snowflake, Databricks, or BigQuery environment. So this way of doing ELT really created a new generation of data engineers who can then support the business requests for datasets customized for domain-specific users. So like with this logic, if the data doesn't have to be analyzed, it just sits in the Data Lake. But then if there is a new business request to analyze part of this data, then as I mentioned, the data engineer can actually use some of these tools to be able to transform it. So within this flow, still like there's a dependency to data engineers or actually, recently, we also have a new persona called analytics engineers. So they become a bottleneck in terms of like, because is the business identifies the need or opportunity for a new product, like when I say product like a dashboard or a report, they require request from the supporting data engineers to create that data. So whether these data engineers or the analytics engineers are embedded within a business unit or operate from a centrally managed team, they still like they do their best to be able to transform, translate the business request into transforms to be able to deliver the data that will to be able to answer those business questions. But again, within this approach, like there's a dependency to a central or a group of personas to be able to make those transformations on the data so that the rest of the business user, which is usually the 80% of the analytics consumers can start consuming that data. I guess for all of those two approaches, although we have been doing great progress, and by the way, a recent term is data lake and there's also lake warehouses as a new term that's coming up, there's still we are seeing low analytics adoption and still there's fraction data dream decisions in organizations. And I think there is mainly three reasons where we are still not able to achieve the optimum data driven decisions where the data can move as fast and transform as fast as the business is changing. So the first reason with all of these approaches, first of all, as I keep saying that it is very centralized and monolithic like as the data goes through its journey of becoming from raw to analytics ready. I think there's a set of users who are kind of creating the bottleneck. The other thing that I would like to say that, you know, in every approaches in terms of either doing ETL or EST, that supply chain of converting raw data into become analytics ready actually has involved many different types of personas like data engineers, data steward business users and recently, as I mentioned, the analytics engineers. So there's a set of hands off between those personas. And at the end, the business users are the ones who knows the business logic that needs to be applied to the data. And then all those requirements, you know, needs to be converted and then passed to the data engineers so that they can apply those logic to data for the business users to be able to consume. So that creates a bottleneck. And the final thing that I would like to say, although there's like more tools right now like air table and dbt or airflow and dbt to transform that data, still those tools, they don't provide the flexibility and agility to reflect the business moment to the data. And I really would like to highlight the importance of being able to apply the business moment to the data. Because this is again, as I mentioned, we live in a very competitive dynamic environment where things are changing so fast. And that that change requires, you know, some of the things to be reflected to the data. Maybe there's a new data warehouse or, you know, because of the supply chain issues, you know, a couple of the vendors are not available. So that you have to reflect that same that those need logic to the data. So overall, like there's a need to be able to apply the business moment to the data. So like when we think about all of those challenges, I think what has happened in the years is there has been a lot of effort and technology innovations that we have been doing to really convert the raw data to analytics. A common example that I give is, you know, during pandemic, every organization, they had the data warehouses, we all had the data maps, right, in star schemas to be able to answer the questions. But yet, you know, many business users where they were still not able to get the questions or the answers that they needed during the pandemic. Because again, to be able to get the right insights to the right users require the need to have business-readed data. And so first of all, let me explain what I mean by the business-readed data. Business-readed data, like, you know, you can think of the business-readed data as the last mile of data transformation before it's put in the hands of insights and data consumers. So data engineering pipelines transforms the raw operational data from transactional systems into a format that is more applicable for using query logic to ask questions of data. That is where I see the transformation of raw data is happening to make it analytics-ready. But then again, as I mentioned, analytics-readed data is not really reflecting the needed and the timely business logic and business context to the data. And that is where I see the gap that is happening. Like, again, there's many different personas and many technologies that we use to produce the data. But on the other hand, there's the data consumers or insights analytics consumers. They're still not able to find the data that is needed to be able to answer the business question. And I think this is because maybe we think that there's a big overlap when it comes to analytics-readed data. But when I think about it, there's really less overlap when it comes to business-readed data. So again, the business-readed data is the final transformed version of the data that has the timely business logic and business context applied so that right insights to the right user can be applied to the right time. So this is something that I take it from heart. Again, just like we have been always making technology innovations, solving problems to make the raw data analytics-ready with data warehouse automations, change data capture, and many other technologies. But we really didn't pay enough attention on how to make analytics-readed data business-ready. And this is where I think we have a chance with the semantic layer to really enable the business units and business owners to create value-based data products by using the semantic layer so that the analytics-readed data can become business-ready. All right. So why I keep saying that achieving business-readed data requires to have a semantic layer. I think when you think about having a business-readed data that truly enables the data mesh concept, again, you know, data mesh concept is all about how you enable the business units and business owners with the creation of the data products. There are like four criteria that you need to pay attention. The first one is, you know, decomposing data around domains. So when I say domain, it is usually the business units or business domains. So just a couple of examples, like it can be, you know, the different line of businesses that you have, like finance, human resources, operation, sales. And it's about like the data mesh is really distributing the ownership of the data with governance to those business units and then creating data domains. The second one is really, you need to think about serving data as a product. This is a very important thing for me because I spend my career out of time with design thinking and then thinking about, you know, doing innovations always with the user in mind. So when I start talking about, you know, creating a data product, it's all about delighting the consumer of that, of the data product. So and if you think about the user needs when it comes to consuming insights and analytics and data, the user needs to really trust to the data. Like when it comes to consuming that data, they should first of all be able to easily discover what data assets are available or data products available to them. And then once, you know, they search and discover the data products and they should be able to easily really understand, you know, and trust where that data come from and what type of transformations has been done to that data. And now what is the best use cases for that data? So again, I really would like to highlight, like if you want to achieve a successful data mesh practice, you really need to approach the data as a product and then have user driven design thinking on that. The third important criteria is really, you need to think about enabling autonomy. And again, when we think about the data mesh, it's all about enabling the business units and the business owners with capabilities so that they can actually do the necessary transformations to convert the analytics ready data to make it business ready. And to be able to enable those users, you really need to kind of abstract the technical complexity so that they can more easily work with the technology without requiring a lot of technical skills. And then the fourth one is you have to think about how to build an ecosystem. And I will be talking more during the session about, you know, what are the things that will enable you through the semantic layer to create an ecosystem. But at the high level, it really becomes, you know, how you can have a central governance, but then make those data products through the governance with governance, searchable, discoverable, so that the business units and the business users can easily search and find and reuse the data products or the components of the data products to be able to create their own semantic models and the data products to be able to use it in the analytics consumption. So really like it's about, you know, creating how to create an ecosystem. So when we think about those four important criterias to achieve the business ready data, there's a need to have, you know, a semantic layer. So before I get into like how semantic layer really enables the data mesh concept, I just want to kind of talk about the dinosaur in the room. So as you, you know, you may be aware of already what's a semantic layer. Obviously, it's not a need concept. Semantic layer has been, you know, around maybe, you know, two decades, I think it has been first introduced with business objects. And it's really like creating a semantic layer is creating a business representation of the data that help and users to access data autonomously using common business term. At the very high level, the way that I refer to it is if you want your data to talk the language of business, you need a semantic layer. So it really kind of map the complex data into familiar business terms. And then as I will be covering, it really enables the organizations or the create the data products where the data products become discoverable, trustable, and then, you know, composite like that can be, you know, can can put together like Lego builds, Lego blocks to kind of really enable the different business units. So again, like semantic layer, it's not a new term. I think it's the dinosaur in the room. But as we look at the, you know, the modern cloud structures and also like thinking about the data mesh and how to enable data mesh, it's the must to thinking about semantic layer. So why am I keep saying that, you know, semantic layer is really enabling the data mesh because I think so far in the presentation, I repeated a couple times that data mesh requires a central data governance. So today what is happening is if you don't have a semantic layer, you want to enable your business units with the necessary data transformation and the consumption. And then, then what is happening is, as those business users are touching their data or the touch point of the business users to the data is the business intelligence tools or the analytics tools. This is where all of the transformation of data is happening to make that analytics ready data business ready. So again, the operational data maybe has been already transformed and then is being made analytics ready. And then by using AI or BI tools, now the business users are having access to those analytics ready data, but there is still tons of transformations that are happening. And the worst thing is all those transformations to make that analytics ready data business ready is happening at the edge in each BI or AI tool. Of course, again, many thing about data mesh, you don't want to lose the central governance, the single version of the truth while enabling the business users. So that's why it is very important to think about the semantic layer. The advantage of using semantic layer, you know, although it provides the central governance, one place where, you know, there's one definition of customer and, you know, the transformation logic to make data, you know, customer data. There is also the benefit for the additional benefit for the data consumers, because now, again, as I mentioned, all of the business units can have access to the, you know, single version of the truth. It is trusted. They can actually see the lineage of the data, understand what transformations has been data, where the data is coming from. The data becomes discoverable right away. And then, more importantly, I see that, you know, semantic layer to achieve data mesh and also to achieve the ultimate self-service. When you have the semantic layer, you know, you can let the business users to really use their BI tool of choice to access the single version of the truth, to be able to do their analytics. And for me, again, I spent almost two decades in my career, like building and innovating self-service analytics. And I think one of the important things when we think about self-service analytics is the, you know, the ultimate self-service analytics is we can actually let the business user to use their BI or AI tool of choice to consume that data. And that is what the semantic layer enables. So I think this was a kind of a good recap of what's the data mesh, what's the semantic layer, and where I see the overlap between semantic layer and data mesh and how semantic layer really enables the business units. But again, there's like more things that I would like to cover when it comes to the capabilities of semantic layer. There are fundamental things that we really need to think about to achieve data mesh success. So first of all, a semantic layer, it really manages the translation of analytics-ready data to business-ready. And this is again, as I mentioned, where you can really enable your data to speak the language of your business. The second thing is, you know, if you are thinking about the semantic layer, the platform should be really, you know, easy to use, where it can actually abstract the complexity of data and technology so that it really simplifies the creation of new business ready views with pre-built and composable building blocks. And then again, semantic layer should be the ultimate logical place where you can apply the governance policies, where you can put the guardrails on data usage and also ensure that there's compliancy and trust and consistency that is happening on the data. So when we look kind of closer to the like more detailed capabilities that I advise, you know, to the customers or the users that I talk to, you know, what are the things that you need to think about when if you want to apply a semantic layer approach to enable data mesh. The first thing is, you need to really have a practical and agile approach for semantic modeling. So semantic modeling is where I see that, you know, the application of the business logic and business context happens to the analytics-ready data. This is where the data, you know, becomes more, you know, more dimensional and the semantic modeling really needs to enable different modeling personas. Like the persona doing the semantic modeling could be a BI developer with a graphical user interface or semantic modeling can also be done via a markup language or a code-based approach so that, you know, data engineers or the analytics engineers can actually create semantic layer or semantic model at the end of the data pipeline so that the data pipeline doesn't have to end with just a table, a data table, but it can actually end creating a whole semantic model where the user can actually create, you know, a metric store and then define all of the dimensions and the breakdowns or the relationships where that set of metrics can be analyzed in a government manner. The final, the other thing that I would like to mention about the semantic modeling capabilities that you need to think about is, you know, the semantic model should support the composability with confirmed dimensions. I think this confirmed dimension concept is very important for data mesh. So let me first of all explain what I mean by the confirmed dimensions. Confirmed dimensions are you can think about them as the common way of analyzing data, like common master data that exists in the organization, like things like product, customer, time or warehouse or store. Like those are the things that you would like to have a consistent way of analyzing metrics. So those are, you know, what I'm referring as the confirmed dimensions and your semantic layer or the platform should enable you to define those confirmed dimensions once, right? Because you want to have one definition of customer, one definition of product in your organization. But then once those confirmed dimensions or the master data has been defined, then it should become searchable, discoverable, so that, you know, if a sales department has already created a definition of customer to analyze the metrics, then the marketing department should be able to easily use the same confirmed dimensions to be able to have the same type of answers to achieve the single version of truth. So I really think that the confirmed dimensions are a, you know, the needed, the stone to really achieve a successful data mesh because I see the confirmed dimensions as the connected, like, are the connecting points to create that connected tissue of data mesh in an organization. The second area that I would like to highlight is the power of providing central governance via the semantic layer. Lately, as I talk about governance, especially in the age of cloud analytics, modern cloud analytics, I actually encourage the customers that I talk to always think about, you know, not only data governance, but to go beyond the data governance and then start thinking about performance governance and also financial governance. Because as you all know, we now live in the age of cloud data, and it is very important to think about the, you know, cost and the financial governance. Recently, especially with the, you know, economic station that we see this year, I really think that the CFOs are the new CEOs. Congratulations to the CFOs. It's very important for the organizations to manage the cost of their data warehouse, cloud data warehouse. And this is where I think, you know, a true semantic layer platform should enable you to provide the visibility and also actually help you to optimize the cloud data usage so that actually you can start easily, not only monitor, but optimize the cloud data consumption as well. The third area that I would like to highlight for the key capabilities of a semantic layer is to really the opportunity to create decentralized data products. This is, again, very important to think about user-driven design thinking to create the data products. This is all about enabling the, you know, the user with the right tool or the tool of choice to do the analytics. So if you're a financial analysis, you can, you know, you should be easily use Excel to be able to, you know, analyze your data that is on the cloud via the semantic layer, but still, you know, have the same definition of metrics in a high-performance way that you can analyze it. Or if you want to do more dashboarding, then you can use, you know, your favorite dashboarding tool. Or if you're a data scientist, you should be able to use your choice of ML tool to be able to consume data and do ML modeling. So this is all about, you know, enabling the right persona with the right data tool so that they can actually consume the data product to generate insights. All right. So I've been talking, you know, what is the semantic layer and the key capabilities that you should be thinking if you consider a semantic platform to enable data mesh. I just want to kind of highlight, you know, where semantic layer sits. I think so far, it has, you know, become obvious that, you know, semantic layer sits between the cloud data platforms and then the analytics and then AI tools. This is where I really think that you can enable the business units where they can actually come to the semantic layer where they can actually do, you know, with drag and drop, with a nice easy to use graphical user interface, they can define their tables and the relationship between the tables. They can define and create a metric store and then they can actually define confirmed dimensions like that one definition of customer, one definition of product. And when I say definition of a customer or a product, I don't only mean like a business definition or a text. What I mean more is like, you know, a customer dimension may actually have four different data tables where, you know, this is a data steward can actually define what's the best way to relate those four tables so that the user analytics consumers can have the right way of drilling down that they want to analyze their metric by customer. So, and then once the, you know, all those semantic layer or the model definition has been done easily, then the user can define the metrics and the dimensions. And as you can see, like on the semantic layer, you can have a metric store, but then those metrics and then those hierarchies and drill down paths, the same ones in a government manner become available on the BI tool of choice. And this is where I'm referring as the ultimate service. And anything about data mesh is all enabling to enable the business units. And I think this is very important to really enable the analytics consumers where they can actually have access to the government set of metrics and dimensions. So that could be different layers in a semantic platform. I think for me, you know, there are four important parts to think about semantic platform. The first one, which I've already covered like semantic modeling, like what are the things that you should be thinking if you are looking to have to do semantic modeling, as I mentioned, like providing multidimensionality, confirm dimensions, being able to create those confirmed dimensions, but easily share them, become make them searchable. The second thing that I would like to highlight is, you know, the semantic platform should enable query virtualization. And this is very important is when you think about data mesh, right, as it is all about enabling the business units and the business users, it is important for the business users to really analyze the data with the power of the, you know, cloud elix elasticity and then the scalability. So this is where, you know, to enable data mesh, I don't think that you should replicate data so that you can actually, you know, empower your business users with data. You should really think about the technology where you can actually leave the data where it is, especially I'm pretty sure there are many enterprises right now, they are going through their journey to move the on-prem data to cloud so that they can have the elasticity, right? So that is one of the goal. You don't want to, you know, replicate data and extract it from cloud environment and put it in a tool. This is where you really have to think about, you know, how you can leave the data in the cloud in that elastic data warehouse, cloud data warehouse, but then you need to have a semantic platform where it actually can provide all of the advantages that I just covered with flexibility, agility, multidimensional speed of thought analytics, but then by using the query virtualization can actually optimize the consumption without caching or replicating data. I also covered, like, the importance of performance optimization and also cost optimization that you need to think about. So that's why, like, as we kind of think about all of those needs when it comes to looking for a semantic platform, I think there are two things that you need to think. The first one is really, you need to have a semantic layer and this is more the traditional, you know, thinking for the semantic, you know, it's about the semantics of the data and this is all about having a passive metadata, where you store the passive metadata. So definitions of the data, who has access to data, where the data come from, which has, you know, those are all of the important criterias to really to be able to create a data product. But then the other important thing that you should be thinking about is having a semantic engine and this is where I think most of the organizations or technology providers are missing this piece. When I mean the semantic engine, this is where the platform, semantic platform should be able to capture active metadata. Active metadata is the metadata about the analytics usage and consumption so that the platform should provide you like what's the data usage, what is the data popularity, what metrics and drill down paths are used together. And by using this information, you can really empower the data mesh practice in your organization because this is how you can actually have the visibility and monitor what business units are creating, what data products. And again, this all comes about the usability of the data products, but then also like optimizing the consumption of those data products where the semantic engine actually can create, can do automated data engineering to be able to create the right data aggregates in the cloud data so that the consumption of the analytics and data products become much more faster. And through the semantic engine with the usage of active metadata, you can actually also achieve a lot of cost savings on your cloud data consumption because the semantic engine is able to realize the query patterns, the most commonly asked business questions to create those automated aggregations on the data source so that your overall consumption of cloud data doesn't have to be at granular level at the same all the time, but it can actually start using every bits data. So I just want to kind of quickly kind of do a recap in terms of have the semantic platform fits into the modern cloud analytics consumption and then just tie it back finally again to the data mesh concept. As I mentioned, this is all about providing governance of service analytics that can be both descriptive and predictive. You need to think about a semantic engine that leaves the data at the source, doesn't replicate the data, but then it can become the natural gateway for all of the analytics and data product consumption. And then just if you are making investment to cloud data because of elasticity and scalability, you want to leave the database results and so that also the users they can actually have a 360 degree view of all the data that's available in your cloud data platform and they don't actually create their products, data products on a silo data set. I think I have already kind of covered how you need to think also governance in the age of cloud analytics, although for when you talk about data mesh, central governance and data governance is the core of the governance that we have to think for data mesh for a successful data mesh, but you also have to think about performance governance and financial governance in the age of cloud analytics. So to kind of wrap up, I think if you really would like to achieve data mesh through my experience working with customers, I think there are five things that I always suggest if you want to achieve a successful data mesh and what are the main points that you have to think about. The first thing is really you need to define data domain and then have an alignment with business domain for those data domain and a good way of a good way or a good approach to achieve that like again you can look at your line of businesses that exist in your organization and then start aligning the data domains around them. The second thing is you need to think about how we can put business context to the data domains and again as I mentioned data mesh really requires you to kind of the user driven thinking on the data product and this is where you need to be able to have a flexible and agile way to reflect the business logic and business context to the data so that it can become business ready and then once you have those data products you need to be able to easily register those data products so that they actually become reusable, discoverable and searchable and that reuse is very important because this is where you can really enable the central governance with the reuse by achieving a federated ownership that's been done by the business units and that reusable data products which I was referring as part of that is the confirmed dimensions really they create the data mesh tissue because then you will be able to connect different data domains via the confirmed dimensions so that maybe the sales department has already a semantic model to answer questions related to promotions and then now your marketing department can actually do easily search discover the components of that data product and then like a Lego blocks they can actually put them together with the support through the metadata and then just create that connected tissue so that marketing data can be analyzed together with the sales data to create the to provide a 360 degree view on the organization and then the final thing that I want to highlight you have to think about central governance with a federated approach where you can actually give the responsibility to business domains to kind of create those data products I think this is kind of like where I would like to wrap it up in terms of you know what was the data mesh what is the need what we have learned over the last two decades you know doing different approaches to make data analytics ready and then you know where we have been failing because we haven't really put enough attention on the last mile of data transformation where the data should become business ready for the analytics consumption and this is where I see the application of the data product thinking comes in place so just I want to briefly kind of mention you know you know who is at scale and what we do as I've been kind of talking about the value of semantic layer at scale is the leading semantic layer platform and really what it does is it enables a metric layer supporting different analytics use cases that can be augmented analytics or bi consumption or AI consumption but the important thing about you know semantic layer and our platform is we really leave the data at the data source at the cloud data source we don't do extract or we don't do caching and it's a highlighted you know it is all about having a semantic engine where it can actually optimize the consumption of those data products where it becomes actually you know speed of thought with the performance optimization but then while ensuring that you know with the automated data engineering the semantic engine can actually create aggregates at the data source so that the cost overall data cloud consumption costs can actually decrease by time so this is kind of like a you know whatever we are very proud of like we have many enterprises today using at scale as their semantic platform and then I've been you know working in the last couple months a lot with all of those customers to really help them with their data mesh strategy and how they can really empower their business units with a governance set of metrics that's been powered by at scale so if you would like to see at scale in action you can go and visit our website we have a lot of demos and then also I've been writing a lot of blog posts about data mesh practice and also just overall how to create data products there's very good data practices that are available at our resources so with that I want to open the the time for Q&A Shannon thank you so much for this amazing presentation so many great questions coming in here and just to answer the most commonly asked questions just a reminder to everybody I will send a follow-up email by end of day Thursday for this webinar with links to the slides and links to the recording so diving in here I live in a BI scenario and not necessarily big data does the semantic model require the enterprise or organization to have the data warehouse with a star schema or snowflake schema on or will flat tables with no star or no snowflake data model schema work yeah it can work this is the you know the one of the advantage of using a semantic platform where you can actually convert the tabular data to a you know multi-dimensional where you can on the fly you know on the graphical user interface you can define a customer and what fields what drill down should happen when you want to analyze the metric by customer and then the system generates the right sequel logic or whatever the dialogue the backend requires so yeah the the short answer that is the advantage of having a semantic platform where you can actually make already like that star schema and then define additional transformation on that star schema to make it business ready or you can have a tabular data you know you can still define the business logic and then importantly you know provide multi-dimensional analysis on that tabular data quick so it seems like the vocabulary management is missing is that some of it no yeah that's a great question so I'm assuming that the question when they say vocabulary management it is like how you define the business terms on data and this is very exactly the semantic layer you know this is you know by using that graphical user interface you can actually define the business definitions and business term on the data fields this is where you are actually defining or making analytics data business ready is what I've been referring like when I say making analytics ready data business ready is you know there are different activities that you conduct like one of them is defining the business terminology on the data the other thing is you can actually create new transformations in a agile and flexible way by using the semantic layer without requiring you know writing how to write you know sequel or complex different scripting language so those are the things that you can you know do easily do on the semantic layer to really enable data management and make the data business ready is the semantic layer a single unified tool or a set of tools that work together to provide all this functionality it's usually like a single tool that where you can actually do the semantics of the data but I think you know you have to think about a semantic platform or the layer as part of your whole data fabric strategy that is one of the things that I have been highlighting at the beginning of the call right when it comes to the definition of data mesh versus data fabric data fabric is the utilization of multiple technologies in combination to enable a metadata driven implementation and augmented orchestration design and a data mesh is a solution architecture that can guide the design with the technology agnostic framework so coming back to the question this is where I see you know semantic platform is part of that multiple technologies that really enables the you know a metadata driven implementation of data fabric and then you know data mesh is really an organizational framework and practice to achieve you know with the use of technology so it is part of the you know multiple technologies so um data mesh and data fabric are different are they different concepts or also different technologies and how they're implemented and can a semantic layer be uh the technology for both I think I just was answering that question Shannon but that's again a great question and it's just great to see those things again just to repeat like uh data mesh and data fabric are not the same concepts uh a data fabric is the utilization of multiple technologies um in combination to enable a metadata driven implementation um and different orchestration orchestration techniques to really kind of create that connected tissue via the technology so the data fabric is you know it's all about the application of the data where those technologies can share the metadata to create that connected tissue of data mesh and data mesh is actually again as I as I mentioned it's an organizational practice um to really um help the business units to create their data products uh so when it comes to semantic layer semantic layer a platform is part of the technology right and and when you think about the whole uh data supply chain it starts from raw data to make it analytics ready and you have a set of uh technologies that they use like dbt airflow um and others um and then to make the analytics ready data business ready this is where I see the semantic platform technologies to be that needs to be used um but then really to augment uh the business ready data there are other technologies like data catalogs analytics catalogs uh that you can use so that you can make the data products more searchable more discoverable um again uh it's all about kind of like thinking about the metadata um and then how we can create the connected tissue of data by using those metadata and then semantic metadata is an important part of that thank you so where do I store such a semantic later uh sorry Shenak could you repeat the question where do I store such a semantic layer so um one of the important thing that you have to think if you're well to consider semantic uh layer that you know the semantic layer shouldn't require data storage um again when you think about you know we live in we live in the age of cloud data um and then elastic data platforms in the cloud um you know and many enterprises are moving their data to clouds to provide that elasticity and scalability and performance so that's why it's very important if you are considering to have a semantic platform where you can actually you know apply the business logic make data more business ready um you really have to think you know look at the technologies where there is no caching no replication of data or no data moments um the system that's where I I've been talking about you know ask about semantic engine um because the semantic engine uh should be able to really uh answer the questions that are coming from multiple bi tools should have the unique capability to convert the right dialect to back to sql and then run those queries on the data source it's very important like I really like if there are a couple things that you are taking out from this uh seminar I I just want to kind of really highlight that you know if you are looking to be successful with data mesh yes data mesh is all about creating data products for your uh where your business units can achieve that but that shouldn't mean that you have to replicate data to have data products that are available to business units if you have a strategy to move the data to cloud you know there's a reason why you have started to that journey so look for the technologies semantic layer technologies where it can leave that it should leave the data at the data source and then you know have a unique semantic engine to enable business strategy data without moving data there's so many questions coming in any questions we don't have time to get to I will be sure and get over to uh to you afterwards so but continuing on here any tools out there that could um well let me actually move on to this so do you have an example of a completed semantic layer that you can show uh I don't have it but if again if you if the audience can go at skill.com there are a great set of demos where you know we can really see how easy to create a semantic layer on top of your cloud data and then have your business units create data products make the data business ready and then to make it available to be consumed with the any bi tool or bi tool of choice yeah so unfortunately I haven't planned in a demo on this session. No worries so and you know for in terms of data governance are there any tools that fit into the architecture in the context of data mesh do traditional tools like you know any data catalogs or stuff like that work well yeah yeah I highly suggest again that is a great conversation that I'm having with our customers as well as a again it's all about using multiple technologies uh data catalogs is where you can actually have um you know more business definitions on the data they can provide more uh visibility on the data lineage and the way that I see you know semantic metadata is part of that whole overall metadata um fabric that the organization has so definitely I think there are additional technologies that you can consider uh data catalog and in fact many of our customers they have a data catalog product um and we have an integration to uh of the semantic layer to data catalogs so that we can actually reflect the metadata of the semantic models to the catalog uh so yeah I definitely uh you know um encouraged organizations to look at those technologies as well. There's one question that I like what's the difference between semantic layer and headless bi that's a great question in fact I think you know we have to think about what's the difference between semantic layer headless bi and metric store I think those are all terms that you know define the same thing uh it's all about having a layer uh that sits between the analytics consumption or ML consumption and the cloud data warehouses this is rare you know having that layer where you define you know a common definition of uh metrics and dimensions and all the business transformation so semantic layer headless bi metric store they're all the same concepts um and you know I think headless bi we have started seeing it more and more when it comes to embedded analytics or process automation again you know our semantic layer has apis and this is where I refer it as a semantic platform where you know it's not only about human analytics consumption you know we are all going through a digital transformation there is more automated processes so all of those processes should be able to have access to the semantic platform or the headless bi uh to use those you know business ready data for the processes as well they're really nice and uh just one more architectural question here I think we've got time to slip it in does your uh semantic layer in at scale need to be relational or do you support graph semantic layers and knowledge engineering oh that's a great question so uh it doesn't like um the way that we we have done it uh it's actually we have a graph within our technology stack this is how we are able to understand the query patterns in an organization um like for me the use of graph is important for active metadata because active metadata is all about understanding the analytics usage patterns in an organization and from that perspective like its scale semantic platform has a unique graph database that we use on the back end with a lot of IP on it uh this is how we are able to optimize the analytics consumption so again you know part of the semantic platform yeah graph is important for active metadata but then you know the tabular or relational structure is also important for the passive metadata this is all about how the data relates to each other well I know this has been a fantastic webinar thank you so much but I'm afraid that is all the time we have scheduled for today again just a reminder to everybody I will be sending a follow-up email by end of day Thursday for this webinar thanks to the slides and links to the recording and again I'll get all your unanswered questions uh over to at scale uh so those can be viewed as well I love thank you so much thank you at scale for sponsoring today's webinar and thanks to all our attendees for being so engaged much appreciated hope you all have a great day thank you very much and thank you bye thank you