Microsoft Purview Unified Catalog - Simply Explained

0:00 Your company's data isn't missing, it's just impossible to find. Most teams keep their data spread across different systems, and when someone needs the right numbers, they end up asking around instead of searching. Welcome to the M365 FM podcast, where we explore the people, ideas, and technologies shaping the future of work. Microsoft Purview. Unified catalog is Microsoft's answer, a plain language catalog of your data that anyone on the team can browse. By the end of this, you'll know what it is, what problem it solves, and who actually uses it. And you won't need to be technical to follow along the data problem. Think about the last time you needed a number for a meeting. You knew it existed somewhere, but you weren't sure which system held it, or whether the version you found was the current one. So you asked a colleague. They pointed you to someone else, and eventually you got an answer you hoped was right.

0:42 That's the quiet mess a lot of companies live in every day. Now multiply that across a whole company sales keeps its numbers in one system. Finance keeps its own version, and marketing keeps a third. The same customer shows up three times with three different spellings. Reports disagree, and meetings stall while people argue about whose number is right. Nobody trusts the dashboard enough to act on it, So the work slows down right when it should speed up. Microsoft has a name for the fix, and that name is data governance. That phrase sounds heavy, but the idea underneath is simple. Data governance means making sure the data you use in reports and decisions is findable, accurate, trusted, and protected. When it works well, a person who has never met you can find the right data set on their own. Understand what it means and trust it enough to build a decision on it.

1:23 The cost of skipping this is bigger than it looks. Microsoft says data professionals spend most of their time finding, cleaning, and organizing data instead of using the clean, trusted data they already have. And that's a lot of skilled people doing detective work. Governance used to sound like a lock on the door, and it still keeps data safe. The modern version is also about making data visible to the people who need it. That shift matters more every year, because AI tools and reports are only as good as the data behind them. Feed a chatbot messy data and it will answer with confidence and be wrong. Microsoft ties the urgency to two forces at once, which are the spread of AI tools and the growing pile of rules about how data has to be handled. The easier your data is to find, and the safer it is to use, the less those two forces fight each other.

2:05 There's one idea that holds all of this together, and it's the thing. Every piece we cover shares keep it in the back of your mind because it explains why each part works the way it does. Two groups feel this pain in different ways, and it helps to see both the people who need the data lose hours hunting for it, and the people who own the data lose hours answering the same questions over and over. A good catalog takes pressure off both sides at once. What it actually is. To understand the fix, start with the tool. Microsoft Purview is the family of products Microsoft built for data governance and data security. Inside it, two pieces do most of the work, and they're easy to confuse. There's the data map and there's the unified catalog. Think of the data map as the part that does the legwork. It connects to your systems, scans them, and collects information about your data.

2:46 That information is called metadata, which is just data about data, a file name, a column heading the data table was last changed. All of that is metadata. The scan is what keeps the map current. Microsoft's tools read your systems on a schedule. So when a new table shows up, the map learns about it. You can scan everything at once or scan only. What changed since the last pass? And on a big data estate. That second option saves real effort. The scan can also flag sensitive information like fields that might hold personal details. Metadata is the thing that makes search possible in the first place. Without it, finding a data set means knowing someone who knows someone, which is exactly the problem we started with. The most important thing to say out loud is this everything in the data map and the unified catalog is metadata, not the actual data.

3:27 These tools describe where your data lives and what it contains, and they don't move it or copy it. The permissions inside them control who can see the catalog entries, not who can open the underlying data. That distinction clears up a lot of confusion, so hold on to it. That distinction sounds small, and it changes how you think about the whole tool. A map that describes your data is a very different thing from a place that stores it. If the data map is the survey, the unified catalog is the searchable result. It's where people browse what exists, understand what it means, and ask for access. Microsoft runs it as a cloud service, and it keeps an inventory of your data assets, their metadata, and their lineage. Lineage is a fancy word for a simple idea, which is the trail of where data came from and how it moved. You can search the catalog by keyword, by business area, by data product, or with an AI assistant that understands plain questions.

4:13 Microsoft also uses one word that sounds corporate, but isn't. It calls the approach federated in plain terms. A central team sets the rules and the standards, and the people closest to the data handle the day to day governing. Nobody has to wait on one overloaded team to approve every single thing, which is the difference between a catalog people use and a catalog people avoid. The catalog aims at three different needs at once for the people who consume data. It offers discovery, secure access, and enough understanding to use the data well for the owners and stewards who manage it. It offers tools to curate the data and keep it high quality for the leaders who pay for it. It offers a way to tie data to business value and standardize the whole estate domains. A catalog with everything in one flat list is still too much to handle, so Microsoft gives you a way to organize it.

4:58 The tool is called a governance domain. A governance domain is a boundary that groups your data by business area like marketing or finance. Microsoft describes it as a mini catalog inside the catalog, and that's a fair way to picture it. Each domain has a name, a description, and owners who look after it. This is where ownership gets solved and it fixes a common frustration. The old complaint was that it stored the data while the business actually used it, which left a gap between who understood the data and who looked after it. Domains closed that gap by handing ownership to the people who know the data best. The boundary earns its keep in daily use. If someone needs supply delivery schedules, they can search just the supply chain domain instead of scanning everything the company owns. They land in the right neighborhood and skip the rest. This is where the catalog stops feeling abstract and starts working like a filing system you'd actually want to use.

5:43 Microsoft lists a few common kinds of domains. There are functional units like sales and finance, lines of business built around specific products, core entities like customers or employees, regulatory areas tied to privacy rules and project domains for cross team work. The guidance from Microsoft is to mix them because a single giant domain works fine for a small company and falls apart as the company grows. Domains also hold your business goals. Microsoft calls these OKRs short for Objectives and Key Results. In plain words, an OKR is a goal. Plus the numbers that show whether you're hitting it. Attaching a goal to a domain tells everyone which data matters most right now. Which turns a pile of files into a set of priorities. Domains give governance somewhere to stand in a flat catalog, you'd have to govern each asset on its own, which gets expensive fast.

6:27 A domain lets a policy apply broadly to everything inside it, so the rules scale as the data grows, instead of multiplying the work data products domains organize the data. But people usually don't want a pile of tables. They want an answer to a question. That's where the next piece comes in, and it's the one that changes how people actually work. A data product is a bundle of related data assets such as tables, files, and power BI reports packaged for one clear use case. It comes with owners, a plain language description and a list of everything inside it. That description is written in business language, so a newcomer can read what the bundle is for and decide in a minute whether it fits their job. Picture the difference this makes without a data product, someone building a report has to find and request access to fifteen separate tables, one at a time with a data product.

7:08 All of those tables sit under one use case, so the person finds the bundle and understands the whole picture in one place. That's the moment the catalog saves someone an entire afternoon. A data product lives under one governance domain, but it can be discovered from others. That balance keeps ownership clear without hiding the data behind a wall. The access flow is where it gets practical. A user opens the data product clicks request access, agrees to the terms of use, and the right approver signs off. Once approved, the user gets the whole bundle at once rather than one piece at a time. The terms of use matter as much as the approval, because the data arrives with its guardrails attached. Instead of a warning buried in a wiki, nobody reads. Owners win here, too. When they change a policy on the data product, it flows down to the assets inside it so they don't have to update every table by hand.

7:52 Microsoft gives a clean example. A data scientist builds a set of assets for a model and wants others to reuse them. They package the assets into a data product once, and everyone after them finds the finished set. Instead of repeating the same research, the company stops paying for the same work twice. There's a governance benefit tucked into the same idea. Because the assets are grouped, Microsoft can score the whole product and tie it to business goals so owners can see which data is healthy and which data needs attention first. There's a limit, though. A bundle only helps if everyone agrees on what the words inside it mean. That's the next piece, and it's the quiet one that makes the rest hold together. Glossary terms. The idea we promised to come back to is context. Every piece so far is really about putting business meaning on top of technical data, and glossary terms are where that idea becomes concrete.

8:37 A glossary term is a shared business word with a definition that you attach to your data. Take a table with a technical name that nobody outside the database team understands. Link it to a term like account, and suddenly a business user knows exactly what they're looking at. The word does the explaining, so the person doesn't need to read a manual. The twist is that in the unified catalog, terms aren't just definitions. They're active, which means a term can carry rules and policies. So the meaning and the handling travel together. Microsoft gives an example that shows why that matters in a human resources domain. A term called feedback results marks company feedback as sensitive. A review team handles access requests, and the policy attached to the term applies automatically to any data product that carries it. A steward sets the rule once and it follows the word everywhere.

9:18 Microsoft's bigger point here is scale. A steward used to apply policies to one asset at a time, which simply doesn't keep up. As a company grows with active terms, the steward defines the policy once and lets the term do the work across everything it touches. Terms are created inside domains, so sales and marketing can use the same word with their own meaning, and the domain keeps the two apart. Once a term exists, you can attach it in several places to a data product, to a single data asset, to a specific column, or to a critical data element. And it brings its definition and its rules along wherever it lands. There's a close cousin worth knowing. Microsoft calls them critical data elements. In plain words, they group the same important field across different tables. If one system calls it custody and another calls it Sid, a critical data element maps both to one customer ID, so the field is governed the same way everywhere it appears.

10:04 Data quality so far you can find the data. You know what it's bundled with and you know what the words mean. There's still one question left, and it's the one that decides whether any of this is worth the effort, which is whether the data is actually any good. Data quality in the unified catalog answers that with rules and scores a rule is a check, like making sure a column is never empty or that every value is unique. You set rules at the column level and you don't need to write code to do it. There are ready made rules you can pick, and AI can suggest rules for you. The reason Microsoft puts so much weight here comes down to AI. Its own framing is blunt, which is that unreliable data undermines trust in AI insights, and once people stop trusting the answers, they stop using the tools altogether. Microsoft's ready made rules measure six things completeness asks whether values are missing.

10:47 Consistency asks whether the data matches across places, and conformity asks whether values follow the right format. Accuracy asks whether the values are correct. Freshness asks whether the data is up to date, and uniqueness asks whether duplicates have crept in. This is the part that protects you personally, because the next report you build depends on whether those checks passed. Then the scores roll up, a single rule gets a score, and those scores climb to the data asset, then to the data product and finally to the whole governance domain. A leader can look at one number and see how healthy a business areas data is. Microsoft also adds profiling, which means the system studies your data and suggests which columns are worth checking. It shows a snapshot, like the range of values in a column and how many are missing. That's a health check that points you to the parts that need attention.

11:28 Alerts finish the job when quality drops below a threshold you set, the owners and stewards get notified so they can fix the problem before a report or an AI tool builds on bad numbers. Catching a bad number early is worth more than explaining it later. You can also schedule the checks to run on their own hourly or daily, so quality becomes a habit instead of a one time cleanup. One more piece rounds this out. Microsoft scores how well you're governing with something called health controls, and it gives you a list of health actions, which are concrete steps to improve. That turns a vague goal, like doing better into a checklist you can actually finish. Microsoft's guidance is that quality is never really finished, because a one time cleanup fixes today's problem and quietly returns next quarter. So the catalog is built around checks that keep running.

12:09 Who uses it? If you've been picturing a tool that only IT admins touch, that picture is wrong. The unified catalog is built for several roles, and most of them aren't technical. The reason it spreads across so many roles is on purpose, because Microsoft's goal is to use your whole team and match your data with the people who know it best. At the top, there's the Central Data office. These are the people who set the rules, the standards and the definitions the rest of the company follows. Then there are data owners. They register the data, decide who can use it, and make sure it meets the bar. Data stewards sit close to them and they keep the quality high, the glossary accurate, and the lineage clear. If something drifts, the steward is the one who catches it. Finally, there are data consumers. They're the ones who search the catalog, find a data product, request access, and get on with their work.

12:52 Many of them sit in finance, marketing, or operations, and plenty of them have never written a query in their life. A normal day in that world looks like this. A consumer searches for revenue figures and finds a data product. A steward reviews the request, checks the terms and approves it. The consumer gets the data, and the steward moves on to keeping the glossary and the quality rules healthy. One point clears up a common mix up the permissions inside the catalog. Decide who can see a catalog entry and they don't decide who can open the data behind it. Those are two different gates, and keeping them separate is on purpose because it lets a company share knowledge without handing out access it hasn't approved. Roles also come in two ways reader and owner. So someone who only needs to look at data carries a lighter permission than someone who manages it.

13:33 If you're a beginner, the consumer role is the one you'd likely grow into first, and it needs no coding at all. You search, you read, and you request. The heavier roles come later if you want them. If you're eyeing a move into data work, this is a friendly place to start because the consumer role teaches you how a catalog works from the inside, and it's the natural first step toward the steward and owner roles later on. Recap in one question. The short version is this the data map scans your systems and gathers information about your data, and the unified catalog turns that into something. People can search domains, organize it by business area while data products bundle the pieces people need for one job glossary terms give the data shared meaning and data quality tells you whether to trust it. Here's one question to leave you with if you could fix one corner of your company's data mess first, which corner would you pick?

14:18 And would anyone else on your team agree with your choice?