«Infants with guns!»

Are we mature enough to track, collect and handle data responsibly, according to ethical standards? I talked with Director of Data Innovation at IIH Nordic Steen Rasmussen about the Business impact of Data Ethics.

Here are my key take aways:

• If were track, collect, and keep all data for any random opportunistic purpose, we put your companies at risk.
• This includes a «commercial curse» of budget-heavy tracking and budget-light management and business value creation through data.
• ROT, Norwegian for clutter, is an acronym for Redundant, Obsolete, Trivial - the dat that clutters your way to find valuable data.
• Collection and tracking of data is still too much dependent on people: If there is a change in personal, you get situations where new people «clutter the clutter»

Marketing & Sales

• For many companies it was Marketing & Sales that drove the data-driven agenda.
• The big value of Marketing & Sales is to add the market-dimension to the data.
• You can actually relate your product to the market, ship to where the market is.
• Analyzing market data is «putting a fixed entity on a moving target». The market changes to rapidly to provide good analysis.
• The more you push behavioral forecasting into the future, the bigger your uncertainty.

Business value & Ethics

• Corporate irresponsibility is an issue.
• Sometimes we get involved in a projects for the projects sake.
• Data projects have for a long time been theoretical, so the impact was not visible.
• Chat-GPT is a black box. Should we really give it more firepower, if we don’t know how it works?
• The market determines that there is a business value in being first.
• The speed of innovation doesn’t give time for reactive regulatory bodies to regulate efficiently.
• Companies need data ethical guidelines to say, how they will, shall and can use data.
• Who should define data ethical guidelines in a company? It is still done on a user-level, whilst senior management is looking at market situations and weighting them against ethical guidelines.
• We need regulatory and top-level guidelines that cannot be bent according to market situations.
• «Ideally, but highly unlikely we need a global set of data ethical guidelines.»
• The more trustworthy you are as a company, the more relevant data is shared with you.
• With that trust and data, you can understand the market better than companies that are not trustworthy and basically flying blind.
• Personal Data Literacy is important, and we need basic digital skills in our society.
• There is also a lack of understanding, when it comes to measures set in place for peoples benefit, eg. Cookie-banners.
• We are still lacking good privacy approved alternatives to the tools we are using on an everyday basis.
• Everyone has to follow ethical guidelines. We cannot have a DarkOps department in our company.
• Data Ethics guidelines should be something everyone can refer to.
• Ask yourself: What is the minimum of data we require to collect? Anything else should become an ethical question.

Data Protection Laws:

• There is a difference between the interpretation of regulations in the EU.
• Nordic countries interpret law relatively, is this just, fair, reasonable?
• Southern European countries use a more napoleonic or dogmatic approach, where «the law is the law, and the law must be obeyed.»
• Both Chat-GPT and Google Analytics have been handled differently by data protection authorities.
• Data Protection Authorities generalize to much, and don’t look at differences in technology.
• Is a strict, generalized interpretation creating panic, for users of eg. Google Analytics?

«There is a fundamental conflict between the essence of fashion and Machine Learning.» 

Fashion is always about change, innovation and identity. Whilst ML is good at making predictions based on historical patterns on those things, not change. How do those go together?

I had a fantastic chat with Celine Xu, who is Head Data Scientist at H&M Group, with a mixed background from Applied Mathematics and Business (MBA).

Here are my key takeaways:

Three things Data can achieve:

1. Automation of systems, where we have a clear understanding ion the process.
2. Sourcing, filtering, ranking of data input for decision making.
3. Ad-hoc analysis of all kinds of AB testing or correlation relationship.

Focus for ML in Fashion

• «ML can be embedded across the supply chain from product design to customer.»
• H&M focuses ML on two areas:

1. Future proof for customer experience: search experience, personalization, fashion inspiration, etc.
2. Demand-driven supply chain: reduce cost and be more sustainable through demand forecasting, cruise control, logistic optimization

ML in Product Design (some examples)

• 3D modeling can make the customer more comfortable in a visual way, in choosing the right size.
• Digitalizing fashion sampling is an interesting use case.
• Early detection of format, color, etc. that can generate trends.
• Optimizing logistics and storage through demand forecasting.
• Shop assistance for employees, with sales data, inventory and warehouse information.
• Language to image: use of stable diffusion for design patterns, on the material level.
• H&M is experimenting with trending words on TikTok or Instagram and transferring them to production design.
• Focus is not in understanding a sentence or the sentiment behind it, but instead use hashtag or product description to find out, what are the trends and link those back to a certain garment feature.
• Purchasing records (when and where did you buy what) and web viewing records (devices uses, pages that lead you to your purchase, etc.) are vital pieces to collect for behavioral analysis.

The Challenges of ML on behavioral and cultural data

• ML models are often evaluated on ML metrics, like accuracy, recency, etc. but not really business metrics, like revenue, profitability, etc.
• You fashion taste is influenced by your music taste, your interior design, and so much more.
• GDPR or webside viewing regulation are limiting the data source availability.
• Even harder to get the data accurate on a granular level, eg. Do you buy for yourself or doing you purchased a gift?
• With a lot of different dimensions in the data, you need to balance the accuracy with the cost.
• ML is always learning from something that already happened. If a new situation accrues, there is a lack of information and data.
• The complexity of behavioral data increases bias.
• You could artificially enrich your data set, eg. by drawing conclusions from certain data like income, area of residence, etc.
• Emotional and temperament data (what value do you hold to your fashion) can be important data to provide for your model, but also as an indication how you would react to recommendations.
• A small decision in your model or source data can lead to a butterfly effect.
• In fashion the definition of style is vague. But when you want to quantify certain feature in ML.
• Fashion trends last around 3 months on SOME, but production time is around 6 months. So how can you react in time?
• Images of material can be deceiving. Camera angle, light etc can influence how the color is represented.
• Presentation of textile is hard, because only a limited number of different textiles can be identified by image only.

«MLOps is a set of practices that bring people, process and platform together into a stream-aligned process to manage End-2-End Machine Learning lifecycles.»

MLOps is taken about a lot, so I asked an expert what we are actually talking about. Xiaopeng Li is AI business lead at Microsoft for the Western European market, located in Oslo. Xiaopeng is a passionate influencer in the field of Data and AI/ML, who was nominated as AI influencer of the Year at last years DAIR-awards in Stockholm.

Here are my key takeaways:

Patterns in AI adoption

• AI adoption projects are quite diverse, but with some patterns that are visible across. Here are use-cases that a lot of industries are working with:
  • Business Process Automation as an AI use case 
  • Adopting AI to process documents automatically and extract key-values
  • Natural Language understanding and processing, but also Natural Language generation
  • Chat GPT
  • Knowledge Mining
  • Unstructured data analysis

«Nordic countries are at the forefront when it comes to adopting AI and ML»

• Some of the most advanced search capabilities used in Microsoft are developed in Norway
• Nordic countries are typically quite tach-savvy
• Nordic countries have very good infrastructure

What is MLOps?

• MLOps is about agility, productivity, consistency and quality
• It is about creating scalability for your Data Science work
• MLOPs is a vage concept and you can probably find a variety of different definitions. Is MLOps at the intersection between DevOps, ML and Software Engineering?
• Scale ML development and deployment with constancy, with quality, with speed

The three elements that are most important are people, process and platform

• People:
  • 5 particularly important roles: Stakeholder, Cloud Infrastructure Architect, Data Engineer, Data Scientist, Machine Learning Engineer
  • There are many different roles involved in MLOps, from cleaning data to testing a model an implementing it. These roles need to be orchestrated
  • Domain experts and stakeholders play a critical role in defining the challenge in the first place. They can formulate what to achieve and what is good enough
  • Change Management is important, especially if your ML implementation triggers behavioral change
• Platform:
  • You are in need of a secure, scalable infrastructure to would your models on
  • Mature organizations who do ML at scale, have most an integrated architecture for Data Management, Analytics and Machine Learning
• Process:
  • Data collection,. Data processing and data management are processes you need to focus on in MLOps
  • You need a process and the right competencies to gather use-cases in the first place
  • Build a backlog of initiatives and then go through prioritization based on eg. Data availability, feasibility of solution given current etch-landscape, value for business, cost, time to marked,..

Path to MLOps

• Always start with assessing your current landscape and maturity
• Start by assessing your platform capabilities
• Ensure you have the right competencies and people
• If you want to operationalize MLOps, don’t look at it as a technological problem, but something that includes the entire organization
• Key is to bring key stakeholders as early as possible into the discussion

Oslo AI:
https://www.linkedin.com/company/oslo-ai/

https://www.meetup.com/oslo-ai/

Link to MS learning:

MLOps Maturity Model

«The real value in changing the status quo and getting more women to the forefront, is by having women share about the important work that they do, and not just talk about their gender as a topic.»

This episode was recorded right before Christmas, when I had the pleasure to chat with Alexandra Gunderson and Sheri Shamlou.

Alexandra inspired by a Women in Data dinner in New York, took it upon her to find likeminded people in Norway. That is how she came across Women in Data Science, the conference that was brought to Norway by Heidi Dahl in 2017. First meetup as a Community was June 2018, and this years WiDS event «Crossing the AI Chasm» is coming to Oslo (and digitally) on May 24th, 2023.

Here are my key takeaways:

Women in Data Science (WiDS)

• «Creating a meeting place, a place for people to connect and get inspired»
• Creating a platform and stage for outstanding women.
• Here are some of the events WiDS organizes:
  • «Champagne Coding» - hands on event
  • «Data after Dark» - after work event: 1-2 quick high level presentations
  • «Data for Good» - get together and solve difficult challenges for greater causes
• An important mission is to increase the number of role models in the community to look top to.
• The goal is to provide arenas to learn together, son it is as important to share stories about failure and collaborate around the learnings from those.
• WiDS is looking for sponsors, and one benefit can be, that trough events real-life uses cases can be solved.
• The focus for 2023 is «scalability» how to get unstuck from ML and AI pilots and bring your work to production?

The Quest for Diversity

• Diversity is a complex topic with several perspectives: gender, nationality, background, knowledge, expertise, and experience.
• Why is diversity important?

1. Leads to more innovative and effective solutions
2. Leads to more fair and just outcomes

• The starting point when working with diversity on a daily basis is awareness.

Diversity in the workplace

• Diversity doesn’t magically happen. You have to work for it.
• Awareness is a first step, but you also need to collaborate in broader groups.
• The value is gained when you are able to include everyone in your events and talks.
• For people to work together against biases of any kind, you need an inclusive culture from the beginning.
• Be open in your communication and foster a culture of collaboration.
• The «3rd shift» is an important requisite for women to be able to spend the same amount of time and intellectual capacity at work.
• The work for an inclusive work environment is never over. We have to continuously work on its and talk about it.

Diversity in recruitment

• You have to actively seek out to hire people with different backgrounds.
• In recruitment, be aware of how you write a job announcement.

1. Use gender neutral language (avoid stuff like «Data Science Ninja» or «Data Rock Star»).
2. There are online applications to check if your language is gender neutral and with suggestions for replacements of bias´ words.
3. You need to highlight more possibilities with a job, like growth and learning opportunities.
4. Minimize the list of requirements in a job-posting.
5. Be aware of you own biases and work with diverse teams also in recruiting.
6. When screening CVs, be aware that different people write in different styles.
7. In an interview process, eg. Women don’t like to do coding tests, with someone watching them.

Get involved: LinkedIn group, Meetup or https://www.widsoslo.com/.

«Data Mesh promises so much, so of course everyone is talking about it.»

I had the pleasure to chat with Karin Håkansson about Data Governance in a Mesh. Karin has worked with Data Governance and Data Mesh and is really active in the Data Mesh Community, speaking at podcasts and moderating the LinkedIn Group «Data Governance - Data Mesh»

Here are my key takeaways from our conversation:

Data in Retail

• The culture in retail is about innovation, experimentation, new products. So governance has to adapt to this environment in order to be successful. 
• If retail would do, what we do in data a fashion retailer would sell yarn instead of t-shirts.
• Retail knows what the customer wants, before the customer wants it. What would happen if we in data think like retailers?
• Its more about understanding the business better, then making the business data literate.

Data Governance

• Data Governance best practices in the DMBOK is still relevant, also in a Data Mesh setting.
• Data Governance has been on a journey from compliance driven to business value driven.
• Centralized Data governance creates a bottleneck. Decentralized Governance creates silos. So federated Data Governance is the middle ground.
• Create incentives to create trust.
• If you utilize your platform correctly, you can have high expectations towards computational governance.

Data Mesh

• Data Mesh comes with a cost - you need to invest in Data Mesh.
• But more than anything Data Mesh implementation is an enormous change effort.
• If you don’t know why Data Mesh, you will implement something else
• Implement Data Mesh in an agile way: «start small, fail fast and iterate»
• To start with Data Mesh, work with a business team that is eager to get started and sees the benefits - «You have to have business onboard, otherwise its not going to work.»
• Always check if you get the value that you expected.
• When you do it, make sure you get Governance, business and tech teams to work together and a re aligned on the why.
• Make sure two upskill for Data Mesh - its is fundamentally different: talk about it, have debates, book clubs, ++
• The 4 elements of data mesh: Can you implement those in a sequence or should you look at them as a unite to implement within a limited scope? 
• Start finding ways for people to work together (eg. Common goal, and environment where it is ok to share).
• A good first step is to find an example of data with a certain issue or limitation and talk with the business user about exactly this.
• Data Governance, as much as Data Mesh, is about change management: You need to get close to the business and collaborate actively.
• Your first two steps should be:

1. Start with one business unite, an early adopter.
2. Find their most critical data and talk about actual data.

MDM and Data Mesh

• Are we still hunting for that golden record? How do we work with MDM in a mesh? This is not solved yet.
• You can refer to data, instead of collecting data in a MDM system.
• Maybe the best approach so far are global IDs to track data cross domains, but how you link your data might become the new MDM.
• «You still need to connect the data, but you don’t need to collect the data.» - MDM in a Mesh.

Domains, federation and responsibility

• If you federate responsibility to the domains, they also need the resources and competency to fulfill these responsibilities.
• If the domain data teams are successful in abstracting the complexity, it will become easy to create data products.
• If you scale too fast (faster than your data platform), you might end up having to duplicate teams.

«If we think of Data Mesh as an evolution of data lakes, knowledge graphs are an evolution of Master Data Management.»

Data overload is becoming a real challenge for all types of businesses. With all data that is gathered, both in multiple formats and huge volumes, has created a need for connected, contextualized data. Combined with continuing developments in AI, has resulted in increasing interest in knowledge graphs as a means to generate context-based insights.

I had a fantastic chat with Mozhgan Tavakolifard. Mozhgan describes herself as «incubator and alchemist». She worked with PhD-research on trust and SOME. The techniques Mozhgan used to collect data for her research introduced her to Data Science. On this episode of #MetaDAMA, we talked about Knowledge Graph enabled Data Mesh.

Here are my key takeaways:

Data-driven transformation

• Personal transformation and transformation of businesses are quite similar.
• Very few companies can blame that they are industrializing data and are transformed in a data-driven way.
• For a transformation to be successful, you need to have a holistic view and invest in a practical manner according to your business case.
• Don’t change your business culture to match a data culture, but rather let data be an enabler for the business.

Data Mesh

• Each domain can have a different culture, produce different data products.
• 4 indigents:

1. Give data back to the domains where it is produced
2. Create self-service data infrastructure
3. Federated governance
4. Data as a product

• The centralized data and analytics platform has failed.
• Knowledge Graph can basically be considered as Data Supply Chain for Data Mesh.
• KG can be used to semantically link data products.

Knowledge Graphs

• KG enable a human brain-like approach to derive new knowledge. 
• A KG is quite simply any graph of data that accumulates and conveys knowledge of the real world.
• Every consumer-facing digital brand, such as Google, Amazon, Facebook, Spotify, etc., has invested significantly in building knowledge and the concept of graphs has evolved to underpin everything from critical infrastructure to supply chains and policing. 
• There is a difference between Knowledge graph and graph data store. The semantic layer is what makes data smart.
• KG is when you have a dynamic and rich context around knowledge.
• KG can be used to semantically search for data.
• KG are a very important part of data mesh.
• If you want to start with knowledge graphs: find your business case. What is the purpose for you?

«RIP Semantic Web! The Semantic Web is dead.»

• Maybe part of the problem was that is was academy focused, more than practical, industry-focused.
• If we manage to implement Data Mesh on a society level, we might have taken a big step realizing some parts of the semantic web.

Trust

• Trust and explainability is based on context. Knowledge graphs can provide context and connect between them, this can ultimately generate trust.

«How can you use AI algorithms to make your city more sustainable?»

How can we use AI to work more sustainable and optimize our operations for less pollution and more efficiency? I talked with Umair, Head of Data Science, Data Warehouse and Artificial Intelligence at Ruter AS, the public transport authority in the Oslo region, the Norwegian capital. Umair is also a associate professor at OsloMet, teaching AI to bachelor level student, as well as founder and CTO of Bineric Crowdsourcing and founder of the volunteer organization Offentlig AI.

Public Transportation;:

• «Public transportation is complex, because data is not coming in from many different internal and external sources.
• «A bus can have minimum 20 sensors.» All theses sensors are sending realtime data.
• A huge amount of data is collected through external sources.
• Ruter is going away from centralized Data Management teams to more of a mesh approach.
• Data Mesh will give you a more complex data function, with need for more people and more organization and coordination.
• One team cannot have the full ownership for the entire data-driven prerogative of a company.
• 2 factors that helped Ruter succeed with Data and AI:

1. E2E responsibility for the whole AI algorithm
2. Create in Production and don’t overdue PoCs

Sustainability & AI

1. Capacity prediction

• Predict capacity 3 days in advance, but possible up to 1 month in advance
• This gives passages the possibility to plan their trips better
• An operations team that sees a pick in traffic realtime, send additional busses to ensure enough capacity
• Through the prediction algorithm, fleet capacity can be reduced and it is easy to plan for balancing the load beforehand

1. Fleet management

• The long term vision for Ruter is to work more with Order services. In the future you shouldn’t have to walk to a bus stop, but can order a bus to your home.
• The existing solution is for seniors (67+) and is tested in the Viken area
• But how can you ensure that the busses are close to an eventual future order?
• Ruter is training an algorithm to predict where orders might come from to ensure a bus is parked close by. This results in less driving and less emissions.
• To train the algorithm Ruter uses mainly historical information, but combined with e.g. weather information

1. Analysis of customer feedback

• Sentiment analysis to see how happy/unsatisfied a customer is

Explainable AI

• «AI is just statistics on steroids!»
• Its hard to explain how a probability output is achieved
• That makes an AI algorithm a black box
• Developers create a set of tools and applications which can give insight on different factors how a decision was achieved by an AI algorithm

Quantum Computing

• Traditional computing is expensive and needs more time and resources to reach a specific output
• Volumes of data are constantly increasing
• This was a research project together with OsloMet
• Quantum computing was cheaper to work with then traditional computing
• Quantum Computers are more sustainable and energy efficient

🗓️ 1 Year
🎙️ 2 Podcasts
💡7 Lessons
👨‍💻 50 Data Leaders

Together with Loris Marini, host of the great data podcast Discovering Data, I had a chat live for a retrospect on a fantastic podcast year 2022. Combined, we have had the pleasure to talk to over 50 Data Leaders in 2022 and shared their message with our listeners.

Here are our 7 lessons for 2023:

1. Lead: Data is a shared asset with shared benefits. Data quality is a shared responsibility.

2. Network: Tips to build stronger relationships with your stakeholders.

3. Sell: Map the compelling business need behind your data management initiative.

4. Be memorable: A simple framework to increase your storytelling 10x

5. Support: Enable the business to take lead. Provide guidance.

6. Observe: Don't jump to conclusions, but observe the demographics, ways of working, environment.

7. Have fun: provide a framework to ensure collaborative readiness, physiological safety, and encourage failure and try again.

«Its not just about the use of data, but the use of data in a cross-functional setting.»

What a fantastic conversation with Nina Walberg. Nina has been with Oda since 2019 and has a background in Optimization and SCM from NTNU.

Oda strive  to create a society where people have more space for life. Make life as hassle-free as possible. And to achievetis with help from data, Oda has created its 6 principles for how they create value with data.

Here are my key takeaways:

Business model and use of data

• Odas business model allows for a better and more cautious way of thinking sustainability. The quantum of products can be tailored to the actual need.
• Climate recipe: Oda provides data on the climate footprint to its customers, when ordering products.
• Use data to provide not just inspiration to your customers, but also help them to create a complete basket of groceries for the week, to avoid any additional trips to the super market.
• The use of data needs to be combined with all functions, also business functions. All should be part of the development process

6 Principles

• 6 principles recently updated, but originally created by the entire team in 2019.
• Oda believes in autonomous teams and that trust and responsibility is given to these team.

1. Domain knowledge and discipline expertise
   • 70% of Data and Insight people are embedded in cross-functional teams.
   • Data is connected across the company and across domains. So you can not work exclusively in an embedded model. You need some central functionality.
   • Data Maturity differs between domains. So really the embedded model depends on the circumstances and how Oda applies that.
   • Data Mesh has been an inspiration.
2. Distributed data ownership, shared data governance
   • Processes and parts of the product for Oda are developed by the domain teams. To federate the ownership is a natural move.
   • Domain boundaries need to be explicit. «Every model that we have in dbt is tagged with a team.»
   • Ownership of certain products is harder or sometimes not right to distribute. It can either be core products with no natural domain team to own, and that many are dependent on, or the core data platform itself.
   • Customer data first, but without proper product data you can’t live up to that
3. Data as a product
   • Make sure you don’t just deliver a product, but that it meets the customer need, not just the customer demand.
   • Consistency matters, structure matters, naming matters when it comes to data products.
4. Enablement over handovers
   • Enabling others to do what they should be able to do themselves.
   • Oda has established a segmentation model for five levels of self-service with different expectations to different user groups.
   • Self-service needs to be tailored to different roles, maturity, and needs of the internal users.
   • Data University, Data Hours and many other initiatives help to create a learning culture and improve data literacy.
5. Impact through exploration and experimentation
   • It is important to test and see how a solution actually provides value to the expectations.
   • This provides insights and information you can act on.
6. Proactive attitude towards privacy and data ethics
   • Data ethics needs to be incorporated and can’t be an afterthought.
   • Company values can and should be directly linked to the work with ethics.

This episode was recorded in September 2022. Click here if you what to know more.

«If you do guidance correctly, people will follow it. People want to do the correct thing. Nobody wants to do things wrong.»

From fisherman on Island to Data MVP in Copenhagen! Ásgeir Gunnarsson has been through a fantastic journey and we had a great conversation around PowerBI Governance. Asgeir has his own blog about the topic.

Here are my key takeaways:

• Think of platforms at a translation: You are own a journey and platform is where you stand on, from where you push off, start your journey.
• A lot of organizations are only on maturity level 1 when it comes to Power BI, even though they think they range higher, or even if there are different levels of maturity in different parts of the organization.
• Buzzwords help shape opinions, and to have discussions. With good help, these opinions can be shaped into actions.
• If you have the possibility to do a green field approach to data mesh - consider it: It will not get easier than this.
• The parts of Data Governance, that can be solved with a focus on technology, and process are much more mature and easier to handle than the softer parts, that are concerned with people.
• It is easier to lock out intentional mistakes than unintentional.
• Governance is changing focus, from being compliance driven to providing a framework of how we can get more value from our data, it’s about making people productive.
• By making people more productive and efficient, you can pay off the cost of governance really fast.
• If you use data more proactively, you are moving the power from people to the machine. But there needs to be a balance, its not either or.

PoweBI

• In PowerBI implementations we should have talked Governance from the start.
• PowerBI enables people in your organization to use data on their own. There is always value ion that.
• To use self-service tools correctly, you need to either have people formally trained or/and have guidelines in place BEFORE they start doing things.
• Not everyone in your organization will become a PowerBI expert or is a data person. Don’t expect everyone to go there.
• Governance is about giving people a framework and a good chance to do things correctly from the beginning.
• Most low hanging fruit: Build a report inventory!
• Try to keep the usage of PowerBI to the M365 environment and meet people where they are.
• Know your requirements before you start using tools.

The 5 pillars cover all of what your governance strategy implementation should cover:

• The people pillar is about having the right roles in place and train people to perform in these roles.
• Because PowerBI is a self-service tool the administrator role is often allocated randomly.
• The processes and framework pillar is about what proper documents in place to make it work so users can use Power BI correctly and be compliant. This covers administrative documents as well as end user documents.
• The training and support pillar is about making sure everyone that uses Power BI has gotten the required training. This is also about deciding what kind of support mechanisms you need to have in place.
• The monitoring pillar is about setting up monitoring of Power BI. Usually, it involves extracting data from the Power BI activity log as well as the Power BI REST APIs for information about existing artifacts in your Power BI tenant. 
• The settings and external tools pillar is about making sure Power BI settings are correctly sat as well as how to use other approved tools to support Power BI (extensions to Power BI).