The concepts, the platform, and the master-data lifecycle — everything you need to be productive in MDM Studio and to earn the Associate credential. About 180 minutes of reading, six checkpoint quizzes, and the certification exam at the end.
Stuck on a concept? Ask in your own words — answers come from the course material with links to the source lessons. Signed-in learners only; daily limit applies.
None — this is the entry point.
Foundations has no course prerequisite. Every other track recommends it, and each of them opens with a recap module so you are never stuck if you arrive here second.
Why master data management exists, the method behind the studio, and the language the platform speaks.
Every organisation keeps the same real-world things — customers, products, suppliers — in more than one system. The CRM has one version of a customer, billing has another, the warehouse a third. Names are spelled differently, addresses go stale in one place and not another, and no system can say which version is right. That scattering is the master data problem, and it quietly corrupts reporting, compliance and every process that spans systems.
MDM Studio solves it by unifying scattered records into governed golden records: one trusted, continuously maintained version of each real-world thing, with full traceability back to every source that contributed to it. It runs end-to-end on the database platform you already operate — Microsoft SQL Server, PostgreSQL, MySQL or Oracle.
Keep this picture in mind for the whole course: many source records in → one golden record out, with evidence.
The platform is organised around four disciplines, which also name its method:
The four are sequential the first time you build a domain and continuous ever after — a running domain is always being re-mapped, re-governed and re-trusted as sources and rules evolve.
A handful of terms carry the whole platform:
Matching decides which source records are the same thing; survivorship decides which value wins for each attribute when they disagree.
The studio's left sidebar has ten destinations — count them, because the number matters when you are told to "check the eighth hub":
One screen lives outside the hubs: Messages, reached from the header rather than the sidebar.
The hubs mirror the DESIGN·MAP·GOVERN·TRUST method: Modelling Studio is DESIGN, Integration Hub is MAP, Governance, Reference Data and Administration carry GOVERN, and Mastering, Quality and Analytics deliver TRUST.
Signing in, finding your way around, working models and checkout, and what your role lets you do.
MDM Studio is a multi-tenant platform: each tenant is an isolated space with its own models, sources, users and golden data. If your deployment hosts more than one tenant, sign-in starts by picking your tenant; everything you then see and change is isolated to it.
Accounts sign in with a local username and password, or with Active Directory / LDAP credentials where an administrator has enabled directory sign-in for the tenant and the user.
At sign-in you also select the database engine serving your session — SQL Server, PostgreSQL, MySQL or Oracle. The same canonical model is served on whichever engine you choose; the platform translates its operations per engine, so your skills transfer unchanged.
Three areas structure every screen. The left sidebar lists the ten hubs — your destinations. The header carries your tenant, the active model switcher, checkout controls, global search, and quick access to notifications, messages, Help and About. The work area shows the selected hub as a tab-bar workspace: an Overview landing plus a tab per tool.
The sidebar is shaped by your role. Admins, developers and viewers see the ten hubs flat. For stewards, data owners and analysts, the three build-side hubs — Modelling Studio, Integration Hub and Administration — are tucked under a collapsed Engineering disclosure, so the daily work is on top and the build tools are one click away rather than gone. If a colleague's sidebar looks shorter than yours, that is why.
Two header details matter daily. The search box (Ctrl/⌘ K) jumps to any page, record or action. And the header tells you whether governance is actually running: when a model's review bypass is on, the header turns amber and shows an amber Review bypass ON pill — governance review is suspended. With the bypass off there is no pill at all and the header stays plain. Amber means something is switched off, not that everything is fine.
The header also carries the notification bell and Messages — person-to-person threads with @mentions and unread counts, attached to the work item being discussed. Messages is not a sidebar hub; you reach it from the header.
The Help hub (bottom of the sidebar, or the header ? icon) holds the full User Guide, the step-by-step domain walkthrough, and an assistant that answers questions from that documentation.
All structural design happens inside a working model. You open one from the header switcher; creating, cloning, importing and exporting models happens under Administration → Deployment Models.
Editing structure requires an exclusive checkout. While a model is checked out to you, you can change it; for everyone else it is read-only — they can view, export and preview, but every add, update, delete or run against that model is blocked until you check it in or grant them control.
If you need a model someone else holds, you request control: the holder gets a grant/decline prompt, and the lock transfers only if they grant it. Check-in reviews your changes and releases the lock. This single mechanism is why two people can never silently overwrite each other's design work.
Six roles ladder the platform's permissions:
On top of the ladder sits an Access Control matrix (Administration) where an admin can fine-tune which role may do what per area — and sensitive-data clearances, which gate who can see classified values regardless of role.
My Account (Administration → My Account — open to every role) is your self-service page: display name, avatar, password change, and your sign-in method if directory authentication applies to you.
Sessions are protected by an idle sign-out: after a configurable period of inactivity you get a warning dialog, then an automatic sign-out — so an unattended screen never stays an open door.
When you're stuck, the Help hub is the fastest path: open the User Guide chapter for the screen you're on, follow the domain walkthrough for end-to-end tasks, or ask the built-in assistant — its answers cite the exact guide sections they come from.
Home → My Work is the unified task inbox. Instead of hunting through the Quality, Mastering and Governance hubs for things assigned to you, one list reads through ten kinds of work: change requests, exceptions, match reviews, campaigns, real-time reviews, workflow approvals, survivorship conflicts, comparison conflicts, expiring rule suspensions, and proposals about persistently noisy rules.
Every item carries an SLA ranking, a deep link that lands you on the exact screen and record, and a personal snooze so a task you cannot act on today stops shouting without disappearing.
My Work opens collapsed the first time you visit and afterwards restores whatever you last left open — a deliberate design so a long queue does not arrive as a wall of rows. The same behaviour applies to the grouped trees on Jobs & Runs, Change Governance and the rules browser.
Two of the ten types are easy to confuse and are kept apart on purpose: a suspension expiry is a deadline the system will honour by itself, while a noisy-rule proposal is a question with no deadline — nothing is suspended unless you decide it should be.
How a record travels from a raw source row to a trusted, reported golden record.
Every domain follows the same journey, and the walkthrough guide performs it click by click. In brief:
Re-running the pipeline is safe and repeatable — that is how a domain stays fresh on a schedule.
Trust needs evidence. The Quality hub provides it in four moves. Profiling examines what you loaded — completeness, patterns, distinct values — telling you which fields to believe. Validation rules, grouped into rulesets, encode what "good" means (an email matches a pattern; a customer number is present and unique). Running a ruleset produces a quality score with per-dimension breakdowns.
Rows that fail rules land in Exceptions, where stewards work them individually or in bulk campaigns that assign batches to people for remediation. Failed records are flagged wherever golden records are browsed, so quality problems are visible exactly where the data is used — not in a separate report nobody opens.
The Data Quality report itself is a tabbed workbench: Overview, Entities, By tier, By dimension, By hierarchy, Coverage, Targets and Domain Health. Two behaviours run through all of them: clicking a domain or entity in the scope tree selects and filters — the KPIs and trend follow the narrowest selection, with one scope row and a clear control — and every dimension letter (C·V·U·Cn·I·A·T, with a legend on the page) can render grey, which means not measured: not a pass, and not zero. Learning to read "not measured" honestly is half of reading the report at all.
Governed change is the GOVERN in the method. When governance is active, structural and governed data changes route through change requests: proposed, reviewed, approved or rejected — with the decision recorded. Policies set the guardrails.
A brand-new model usually starts with its review bypass on so first-time setup isn't blocked by approvals — and while it is on, the header is amber and carries an amber Review bypass ON pill, and the model's row in Deployment Models carries an amber Bypass ON chip. Turning the bypass off (from the pill or the row) is confirmed, audited and gated by the model's checkout; from then on the chrome goes quiet and governed changes queue for approval again.
Underneath everything, the audit trail records who did what, when, to which object — with business names, clickable detail, and export. If it matters, it's in the log.
Some values must not be visible to everyone who can otherwise see the data. MDM Studio handles this with classification and clearance. Every attribute can carry a sensitivity level — Public, Internal, Confidential, Restricted — set in the Data Dictionary and cascading domain → entity → attribute (the highest level in the chain wins).
Users see classified values only up to their clearance; above it, values render masked (••••) everywhere the platform serves data — golden records, reports, exports. Administrators grant clearances per domain, entity or even a single record, with optional expiry.
Whole records can also be restricted: hidden outright from anyone below the record's level, indistinguishable from records that don't exist. Every restriction and every unmasked view of classified data is written to the sensitive-access audit.
Several MDM Studio features have names that promise more than the feature does. Knowing the difference is part of being competent on the platform — and every one of these is stated plainly in the product itself.
semantic:false for that lane. It only becomes semantic if the deployment configures an external embedding model.The habit to take from this lesson: when a screen makes a claim, check what it actually computes. The platform is built to tell you — status panels report which engine produced a result and what it did not evaluate.
Master data does not stop at the platform edge; and the studio changes shape to fit what it has to show you.
Everything else in this course is about data MDM Studio masters. This lesson is about the things it does not. A finance warehouse consumes your Party golden records; a downstream report is built on a table nobody in your team owns. Those matter to your lineage, and Modelling Studio → External Assets is where somebody writes them down.
Two ideas are worth carrying out of Foundations. First, nothing crawls — the platform never goes looking across your network to discover what exists. A steward records an asset because it matters, and that record is a person's assertion, not a scan result. Second, external assets are global: they carry no model, because a warehouse is not a fact about one version of your design.
They link to your own assets three ways, and the direction is the meaning: CONSUMES (it reads from you), FEEDS (it supplies you), DESCRIBES (it documents you). Get the direction wrong and the lineage answers the opposite question.
Two behaviours confuse new users until someone explains them, so here it is up front.
Grouped lists open folded. My Work, Jobs & Runs, Change Requests, the rules browser, the Report Studio report list — the first time you open one, everything is closed. An empty-looking tree is a collapsed tree, not an empty one. Every folded group shows its count, so you can see how much is behind it before opening it, and after that first visit the page restores whatever you last left open — including "everything", which is why expand all sticks.
The tab bar changes shape to fit your window. Where a hub's tools all fit, you get them in order. Where a hub is large enough that tools would have to be hidden — Administration, on most screens — the bar carries the groups instead, each with its tool count, and the selected group's tools appear on a slim row beneath. Clicking a group reveals; it does not navigate, so you can look without losing the page you are working in. Widen the window and the flat bar returns on its own. Nothing is hidden behind a bare number.
The third meaning of "hierarchy", and how the cleansed value becomes the value the whole platform speaks — two ideas that make every roll-up and report trustworthy.
Tiers can say that PROVINCE is level 1 and CITY is level 2 — but nothing in a tier records which province contains which city. A value hierarchy does: it is a declared chain of containment between values — South Africa contains Gauteng, Gauteng contains Johannesburg — and that one recorded fact is what makes a roll-up trustworthy. A Johannesburg record filed under Western Cape stops being an unnoticed grouping quirk and becomes a named contradiction someone can fix.
Every level of the chain is a governed reference domain (a code set), so the links are governed too: a link to a value the parent's code set does not hold is refused, a link that skips a level is refused with the levels named, and a code set can never contain itself. Depth is declared with a stated cap that refuses rather than truncating.
Two of its abilities are switched off until somebody chooses them: derive (showing a record under Gauteng because its city resolves there — display only, labelled, nothing written) and the IN_HIERARCHY rule operator (a data-quality check that flags records whose two values contradict the declared containment). That restraint is a pattern you will meet across the platform: a capability made possible is not a capability switched on.
You meet value hierarchies on the ERD as containment edges, in the GR Explorer's Hierarchy rail as the fourth source, on the Record Inspector as a per-record path line (South Africa ▸ Gauteng ▸ Pretoria), and in Data Quality's By-hierarchy view.
Standardization rules produce a cleansed value — consumer and CONSUMER become one form. Matching always uses that cleansed form. But whether the golden record carries it is a separate, per-attribute decision: the master standardized value flag. Off, the cleansing is invisible to every consumer of the golden data.
The platform now defaults that flag on the first time a rule is attached to an attribute on which nobody has yet decided — so a new rule's cleansed value reaches the golden record without a second step. Three states stay distinguishable forever: undecided, on, and off by somebody's explicit choice — and a recorded human "off" is never overturned by the default.
Hierarchy tiers align with the same idea: a tier bound to a standardized attribute groups by the standardized form, so one node holds what used to be two. Where the attribute does not yet master its cleansed value, the surfaces say so — "standardized for grouping — stored value is raw" — and offer mastering as the way to make tiers, grids and exports agree.
One reading habit to take with you: when the platform cannot measure something, it says unknown — which is not zero. An offer whose collateral is unknown still requires an explicit acknowledgement. Nothing on these surfaces rounds ignorance down to nought.
Report Studio, dashboards, marking a column so it reads at a glance, and the two habits that keep a number trustworthy once it is on a screen.
Mastering produces golden records. Report Studio is where the rest of the business meets them — and it lives inside the hub, so answering a question about mastered data does not mean exporting it somewhere less governed.
A report is one visual. There are seven, and the wells you get change with the one you pick:
Authoring is the same three panes every time: Fields on the left, the canvas with its wells in the middle, Properties on the right. Drag a field into Rows, Columns, Values, Filters or Small multiples and the preview redraws. Learn one visual and you have learned the others.
One operator vocabulary runs across every visual, and a filter can be required, exposed as a parameter, or carry a default — the reader is shown which. Relative dates like last 30 days resolve when the reader opens the report, not when it was written.
Two behaviours to expect, because both are easy to misread:
On a dashboard, one filter can drive tiles that come from different sources by binding to the field each source actually uses — and a tile it cannot reach is flagged, not left quietly showing unfiltered numbers beside filtered ones.
Most reporting tools draw something no matter what you feed them. A picture of the wrong thing is worse than no picture, because nobody checks it. MDM Studio refuses instead, and names the reason on the visual.
Two more habits protect the number itself. A gauge or bullet prints the ends of its scale, so you can see what the needle is measured against instead of guessing. And a subtotal over an average is weighted by the records behind each group — an average of averages is not an average — while a distinct count refuses to be re-aggregated at all, because distinct counts cannot be summed.
Values follow the display format governed on the attribute in the Data Dictionary, everywhere: grids, data labels, axis ticks, KPI cards. That formatting is display only — filters, copies, exports and every API payload carry the raw value.
Bands are where an honest gauge is decided. The band editor warns when bands leave a gap or overlap, and its Suggest from data reads the measure's own distribution — where the values actually cluster, what the extremes are — and proposes bands from that. The AI may narrate the proposal; the numbers come from the arithmetic, and you can edit every edge before saving. Nothing is applied until you accept it.
A value can also be shown as something other than itself: % of total, % of a grouping (the row, column or a named level), a running total, or a rank. The show-as is a window over the aggregate, so the raw value is still there underneath — and exports carry it.
A grid of numbers makes the reader do the work. A column tool does it for them: you write one set of rules on the column, and choose how that rule is drawn.
The important part is that the rules belong to the column, not to the mark. Change a threshold and every mark on that column moves together — a bar and a badge on one column cannot disagree, because there is only one place the answer lives. The same marks reach matrix and pivot measures, and row and column headers.
Rules are not only for numbers. Text takes contains, starts with, is one of and wildcards; dates take on, before, after and relative windows like “in the last 30 days”, resolved as the reader opens the report — so a rule written today is still true next quarter. A rule’s value can also name another column, which is how each domain gets judged against its own target instead of one number for all of them.
Two rules of thumb, and they are the same two you have met elsewhere in this course. First, colour never carries meaning alone: every rule takes an icon and a label as well, because a reader who cannot separate your green from your amber still has to be able to read the report. Second, a blank cell matches only “is blank” — it will not fall into “less than 50”. An unmeasured score painted red as “0% — at risk” is a different and much worse claim than “not measured”.
Marks reach total cells only if you switch that on, and the panel says why: a rule reading “≥ 95 healthy” is a claim about a percentage, and the sum underneath it is not on that scale. Turn it on when the totals share the values’ scale, and leave it off when they do not.
MDM Studio Certified Associate — Exam — 35 questions, 50 minutes, 70% to pass.