How it works, in plain English.
The tooling we use comes with a lot of jargon. Here's what each part does, what it means for your business, and how to say it in a meeting. We build on Infinity Data AI's capability. Streem is its exclusive South African partner.
The map shows where each part sits, in the same five layers as our architecture. Click any part to read it.
Decisions get made, acted on and measured.
The top layer is where the value shows up. A decision is only done when it's been acted on and the result measured.
Closed-loop decisions
What it is
Every decision gets acted on, measured and fed back, so the next one starts from what the last one learned.
What it means for you
You see whether a call worked while there's still time to adjust, not in next quarter's report. The gains compound.
Decide, act, learn, repeat.
Dashboards
What it is
Dashboards and analytics built on the same agreed meaning.
What it means for you
The numbers on the dashboard match the numbers in the board pack.
One set of numbers, wherever you look.
People and AI work from the same answer.
This is where your people feel the difference. The questions and the actions all run on the same agreed meaning.
Plain-language questions
What it is
Ask a business question in plain English and get an answer built from your own definitions, with its sources shown.
What it means for you
Same question, same answer, every time. If it can't find the evidence, it says so.
It answers from your data, and shows its working.
AI assistant
What it is
An assistant that helps your people build and use the context layer. It works inside the rules you set.
What it means for you
Your teams get help without the AI going off on its own.
AI helps. Your people decide.
AI agents
What it is
Software that does a piece of routine work inside rules you set, with a named person signing off. In pilot with design partners today, with wider rollout on the roadmap.
What it means for you
Agents do the legwork. Your people handle the exceptions.
The AI proposes. Your people decide.
Business memory
What it is
A memory of what your business has learned, linked to the context layer.
What it means for you
Knowledge stays in the business when people move on.
What the business learns, it keeps.
You choose the AI models, and they run inside your rules.
Models change fast. The meaning lives in the context layer, not in the model, so you can change models without starting again.
Your choice of AI model
What it is
The AI models you choose, run inside your boundaries, locally or through a provider you trust.
What it means for you
When a better model comes out, you swap it in. Nothing underneath gets rebuilt.
The model is a part you can swap.
AI controls
What it is
One place where your team sets which AI models run, where they run and what they're allowed to read.
What it means for you
AI stays inside the rules you set, and you can see what it's using.
You decide what the AI can see.
The context layer holds what your business means.
This is the heart of it. It's where your data gets its meaning, so people and AI read it the same way. Technically, it's called the semantic layer.
Context layer
What it is
One layer above your systems that holds what your business means: what things are, how they relate and the rules that apply.
What it means for you
Every report, dashboard and AI tool reads the same meaning. So the same question gets the same answer.
AI can't read heads. This gives it your definitions.
Industry starting model
What it is
A ready-made model for your industry. It follows the recognised standard for banking, insurance, retail, healthcare, manufacturing or telecoms.
What it means for you
We don't start from a blank page. Most of the shape of your business is there on day one.
We start from your industry's standard, not from scratch.
Your business model
What it is
The industry model, extended for your business and for the decision you're solving. Your words, your products, your rules.
What it means for you
It fits how your business works, not how a textbook says it should.
The industry gives us the outline. Your business fills it in.
Ontology
What it is
The map of what things are in your business and how they connect. A customer places an order. An order holds products. A store stocks them.
What it means for you
The knowledge in your best people's heads gets mapped, so a machine can use it and it stays when people move on.
It's a map of your business that a machine can read.
Taxonomy
What it is
The way things are grouped into categories, like product ranges or customer segments.
What it means for you
Reports roll up the same way everywhere. A taxonomy groups things. An ontology connects them.
Taxonomy groups things. The ontology connects them.
Agreed definitions
What it is
Every key word, like active customer or stock-out, gets one meaning, an owner and a person who looks after it.
What it means for you
Finance, sales and operations stop counting the same thing three different ways.
One word, one meaning, one owner.
Reference data
What it is
The standard lists your business relies on, like country codes, product categories and industry codes.
What it means for you
"NY" and "New York" match before anything else runs, so your records line up.
Get the standard lists right, and everything above them lines up.
Master data gives you one trusted record for every customer, product and supplier.
Master data management, or MDM, is how a business keeps one trusted version of its core things. Almost every business has tried it. Most programmes stall, or slowly come apart after go-live. Here's why, and what we do differently.
Why most MDM programmes stall
- It's built as a separate hub. The records sit in the hub, but what they mean lives in documents, code and people's heads.
- The rules are written against tables and columns. When the business changes a definition, someone has to find and rewrite every rule.
- Governance gets added at the end, in a spreadsheet that's out of date within a month.
- Merges are one-way. Merge two customers by mistake and you can't cleanly undo it, so trust drops with every error.
The industry treats master data as a consolidation problem. It's a meaning problem.
Golden record
What it is
The one trusted record for each customer, product, supplier or store, built from every place it appears.
What it means for you
One record everyone uses, built on an agreed definition of what a customer is, not on a column mapping.
One customer, one record, one meaning.
Matching
What it is
Finding the records that are the same thing. Exact matches, close spellings, names that sound alike and AI-assisted matching, used alone or together.
What it means for you
Duplicates get found without a consultant tuning rules by hand.
We find the duplicates. Your people decide.
Stewardship
What it is
A named person reviews and approves matches, merges and definitions, in one place. Their decisions tune your own matching, in your own environment. No model is trained on your data.
What it means for you
Matching gets better from your people's judgement, and your data stays inside your boundary.
It improves from your stewards' decisions, inside your walls.
Reversible merges
What it is
Every record keeps its full history. You can restore an earlier version or undo a merge.
What it means for you
A mistake can be fixed, so trust in the golden record grows instead of wearing away.
A bad merge is no longer a one-way door.
One accountable owner
What it is
Each area of data has one person who answers for it. That's kept apart from who's allowed to change it.
What it means for you
When something's wrong, you know who owns it. When people leave, the accountability stays.
Two owners is no owner.
| Traditional MDM | How we do it | |
|---|---|---|
| Where meaning lives | In documents, code and people's heads | In the context layer, agreed and owned |
| Golden record built on | Table and column mappings | An agreed definition of the thing |
| Merges | One-way. A bad merge is permanent. | Reversible, with full history |
| Quality | One average score that hides problems | Measured by segment, against the agreed meaning |
| Governance | A project at the end that goes stale | Carried with every record |
| Over time | Drifts back and needs fixing again | Gets easier to extend |
Traditional MDM consolidates records. We govern meaning. Master data is what that meaning looks like for your core things.
Back to the mapQuality and rules keep the answer right.
These parts check the data against the agreed meaning, and keep your policies running the same way everywhere.
Data quality
What it is
Checks that come from the agreed meaning, so they apply wherever that meaning appears. Quality is shown for the whole business and segment by segment.
What it means for you
One average can look healthy while one region has been broken for months. Segment views show the real problem. It finds problems and routes them to the right person. It doesn't quietly change your data.
It doesn't hide a bad region inside a good average.
Business rules
What it is
Your policies, written in business language and run automatically. Credit limits, eligibility, pricing.
What it means for you
Change a rule once and every decision that depends on it follows. No re-coding.
Change the policy once. Everything follows.
Drift watch
What it is
A check that flags when your systems or rules start to move away from the agreed meaning.
What it means for you
Problems show up early, before they reach a board report.
It tells you when things start to drift.
Your systems and data stay where they are.
Everything starts with what you already run. We connect to it. We don't replace it.
Connecting your systems
What it is
We connect to your ERP, point of sale, core banking, CRM and spreadsheets, read-only. Where a use case needs it, we stage data in a controlled way, inside the boundary you set.
What it means for you
Nothing gets ripped out or migrated. The systems your teams use every day carry on as normal.
We read from your systems. We don't replace them.
Where it runs
What it is
In your own environment behind your firewall, or hosted and run by us.
What it means for you
Your data and IP are protected whichever way you host it.
It runs your way, and it stays yours.
Governance and lineage run through every layer.
These two aren't a layer of their own. They travel with the data from the bottom of the stack to the top.
Lineage
What it is
The trail from any answer back to its source, the rule applied and what changed.
What it means for you
When the board, an auditor or a regulator asks where a number came from, you click through instead of starting a three-week scramble.
Every number can show where it came from.
Governance
What it is
Rules on who can see what and how data may be used: sensitivity, where it may be stored, allowed use. They travel with the data.
What it means for you
Audit evidence builds up as the work happens, not as a project before the audit.
The rules travel with the data.
What it isn't
Four things people often assume, set straight.
- Not a replacementIt works alongside your warehouse, lakehouse and data pipelines, and gives them meaning. It doesn't replace them.
- Not a black boxAnswers come from your agreed definitions, with sources shown, not from a guess.
- Not another hubMaster data, quality, definitions and rules all run on one layer. You don't maintain five separate tools.
- Not a compliance stampIt helps you show how your obligations are met. It doesn't make you compliant on its own.
Want to see it on your own data?
Start with one decision or outcome. We'll show you what each of these parts does with it.