When a chief data officer achieves excellence in master data governance, they’re transformed from someone who spends their time reactively fixing things when data gets messy or breaks a system, to a proactive, all-seeing business enabler. This series explores why master data governance is vital for: data modelling, data quality, rich content, industry standards and data enrichment.
Series introduction
Without a strong data model in place, your organization could find itself in all sorts of sticky situations.
It could lead to anything from customers not getting their orders delivered because you have the wrong address (angry customers=bad), to not having the right guardrails in place to comply with regulations like GDPR and HIPAA (angry regulators=very bad).
So, we’ve written this guide to help you define what your data model is, help you decide if it’s time for a change—and most importantly, show you why your data model needs to be built on a foundation of solid master data governance.
Data governance is essential for ensuring that your data model is fit for purpose, extensible and maintains data integrity.
Continue reading to discover how you can go from guardian to guiding light, by improving your data model with master data governance.
Back to basics—what’s a data model?
First things first, what is a data model?
In simple terms, a data model is like a map or blueprint that helps organize and structure information, defining how data is stored, validated and accessed within a database or system.
Think of it as a plan that describes the different types of data, how they relate to each other, and the rules for how they can be manipulated. This helps ensure that data is consistent, accurate and efficient to work with.
In simple terms, a data model is like a blueprint or a map that helps organize and structure information. It defines how data is stored, validated, organized, and accessed within a database or a system. Think of it as a plan that describes the different types of data, how they relate to each other, and the rules for how they can be manipulated. This helps ensure that data is consistent, accurate, and efficient to work with.
The three key elements are:
- Attributes: Individual pieces of information about a part, product, customer, supplier, etc.
- Entities: A collection of attributes that come together to describe that one part, product, customer or supplier
- Relationships: How individual attributes or entities need to interact with and behave around other attributes or entities across your systems.
Weighing up your options
Let’s look at weights, for example. Are you a grams kind of place, or do you prefer putting in the pounds? Is there 4kg of sugar in a package, or 4oz?
Seems simple, but get the measurement wrong, and you’ll have all sorts of trouble when it comes to loading items on a truck or making a cake! Your data model should stop issues like these from happening.
You could end up with millions of those attributes, entities and relationships, which is why you need to properly classify them. This helps you group your parts, products, customers and suppliers within a hierarchy so that you can find the information more easily across large organizations.
Defining your data model is the beginning of strong governance. Take, for example, any forms that employees need to fill in. Entering the wrong information into a form might break processes in another system—that’s why you’d restrict what they can enter into each field, while establishing which fields are mandatory,
When you have a data model in place, you can define policies and rules for collecting data to fit and conform to that model.
Establishing the right data model impacts future governance, which in turn impacts business operations and reporting.
It’s all connected.
Time for a new data model?
All large businesses already have a data model – in fact many of them!. But it can be affected by anything that happens to your business and/or the environment it operates in. That means you might need to make changes, standardize those models or map how they relate to one another.
It could be social issues. For example, if you need data to support ethical work practices. Or it could be new industry regulations like health and safety or ESG. Or economic and political situations that place limits on how and where you’re allowed to trade.
Here are a few of the most common scenarios where you’ll need to review or adapt your data model:
Implementing a new application or solution
If, say, you adopt a new data platform—such as Salesforce—it comes with a pre-configured data model. You’ll need to reconfigure this to work with your current data model.
Any new attributes, entity types or relationship structures need to be incorporated into the universal data model to accommodate any new business activities associated with the new system.
And you also need to consider the level of data governance embedded in the new system. You might have a fantastic model, but if you allow people to type anything they want into the system, you’ll quickly fall into chaos, disrupting existing business systems and processes.
Migrating to a new platform
Sometimes you’ll need to move your data from one platform to another—such as when the old platform becomes obsolete. SAP ECC is a good example here, as it’s being retired in favor of SAP S/4HANA.
These situations provide the perfect opportunity to review your current data model to see if it still meets your needs. Is it too rigid? Are relationships between tables too complex? Is your business having to use workarounds to get data transformed and properly flowing through the organization?
Each ERP has a different built-in data model, which can get in the way of straight-through processing (STP). And without the ability to reconcile multiple data models, migration can be costly and complex, with a heavy reliance on external implementors.
Making mergers or acquisitions
When two companies come together, you have two competing data models. But unfortunately, it’s not as easy as putting them both in a fight to the death to see which one comes out on top.
The chief data officer and their team will need to consolidate attributes, entities and relationships, and reclassify the data into one hierarchy. Then you can adapt the rules, policies and processes to create one single organization from a data perspective. Without a universal data model, it’s difficult for the rest of the business to get to a universal operating model.
Learning from others’ misfortunes
It’s up to the chief data officer to keep an eye on events taking place outside of the organization—because when things go wrong at other organizations, you have the opportunity to learn from them, and apply remediation before they happen in your own business.
This could be anything from fines for incorrect calorie or sustainability data, to a security beach taking advantage of improperly defined data access within a data model.
Three reasons bad data governance is bad for business
As we said in our introduction, when it comes to data governance and data quality, everything is connected.
Based on that, we have a theory at Stibo Systems:
Pretty much everything that goes wrong, anywhere in a company, can be tied back to bad data governance.
That means any governance issues need to be looked at across different business domains. Check out these three common examples below:
Example 1: Products
Product data contains all sorts of important information, from dimensions and imagery to specifications and pricing options.
It all needs to be present, correct and distributed to all the stakeholders who rely on having correct, up-to-date information—such as anyone involved in selling the products.
What’s the worst that could happen?
When the right people don’t have access to the right product information, there are all sorts of ways you’ll end up with disgruntled customers and unhappy category managers.
Just see what happens when a product is incorrectly discounted and promoted online with no imagery and poor specs. You’ll end up with lower-than-expected sales numbers due to poor conversion rates, with less profit than you should have per sale, while more customers are returning items because they weren’t getting the exact product they expected.
Example 2: Logistics
Logistics is another area that requires accurate information across a variety of fields.
First, there’s all the stuff relating directly to an item: height, length, depth, weight. Then there’s where you’re taking the item (the address) and how you’re getting there (has the delivery driver been given a suitable, efficient route?).
What’s the worst that could happen?
Having the wrong item weight and size means you’ll have poorly loaded trucks. Not that that matters when wrong addresses mean the customer or store will never even receive their items!
And, to top it all off, the truck no longer has a roof, because the route took Dave the delivery driver under a bridge that was too low.
You’ll have unhappy drivers, unhappy customers, reduced sales, and empty shelves.
Example 3: Customers
Businesses keep a lot of data on their customers—but is it correct information and available to the right people? If employees are able to enter lots of freeform responses into fields, everything can end up getting in a jumble. Easy to do when you’re trying to get through as many calls a day but can’t find the right fields for certain information.
What’s the worst that could happen?
You’ll inevitably have incorrect, duplicated personal details, addresses, and householding data on file.
If you’re sending out marketing materials, you might end up posting three flyers to three individuals who all live at the same address (that’s if you even get them to the right address).
You could send them an offer for car insurance on a vehicle they sold six years ago—or even worse, a life insurance offer to somebody who’s already died.
And in a more extreme (and potentially dangerous) scenario, a healthcare provider could end up prescribing the wrong medication, or booking people in for the wrong operations.
>> See how you to comply to industry standards with this handy checklist. <<
Why good data governance is good for business
No more drowning in data lakes
Businesses without a universal data model constantly battle to move data between their systems.
Many end up resorting to throwing all their data into one central data lake to transform, sort and analyze it—but how do you know the data you’re using in your data lake is correct?
With master data management in place, you can control, organize and structure the data at its source. Then you’ll know everything in the data lake is good to go.
Know your technical data quality from your business data quality
When assessing data quality, it’s important to make use of data profiling tools, to take a thorough look at the actual 'content' of your data—it's not enough to assume your data is fine, just because it conforms to the data model.
Just because a piece of information is technically correct, it doesn’t mean it’s useful to the business.
Here’s an example for you:
‘Wile’ obviously follows the rules of what a real first name could be. It’s made up of letters, and no numbers, ?’s or !’s have snuck in there.
And going by these rules, ‘Coyote’ could also technically be a last name. So, no problem here—well, other than the fact that your system has failed to pick up that it’s obviously not a real person. That’s assuming you’re not actually doing business with Road Runner’s archnemesis.
But there’s another potential issue. What if, say, your marketers are trying to sell your services to other businesses? You could get into a lot of trouble if the email address you have on your system is “Wile.Coyote@gmail.com” instead of “Wile.Coyote@acmeinc.com”.
Proper business data quality rules would reject the personal email to make sure you don’t fall foul of data privacy laws.
Getting it right with master data governance
So, we’ve shown you just some of the things that could go wrong with bad data governance, and some things on the flipside—what will go right with strong data governance.
And if you’re still not convinced, here are some more benefits:
Strong data governance means a high-performing organization.
- Reduce costs by eliminating rework
- Improve conversion rates on product sales
- Decrease the number of items returned by customers
- Enable more efficient deliveries
- Build more trust and goodwill in your partner relationships
- Reduce the cost inventory management
- Empower your service team agents with the up-to-date information they need
Perhaps most importantly, strong master data governance makes for a very happy data team. With good governance enabling a strong data model, they’ll spend less time reshaping, reformatting and fixing data issues to make the data useable.
So, by leading your organization to a strong unified data model, you’re the guiding light using data to steer the business into faster growth and higher profitability.
How to make it happen: establishing strong data governance
Data governance is indispensable to the success of data modelling initiatives, and to maintaining high technical and business data quality within your data model.
Without it, you won’t get the insights needed to make informed business decisions, meaning you’ll lose the strategic value of your data assets.
With a strong data governance model in place, you’ll establish a solid foundation of standardized definitions, while setting expectations for data quality management. It goes a long way to ensuring a good security posture while promoting accountability that enhances the precision and reliability of the data itself.
Before you begin: top tips for your master data management project
According to Gartner, 75% of MDM projects fail to meet business expectations. Why? Because organizations often jump straight to implementing technology, failing to first consider how important people and processes are.
So, here are some of the things you need to do before bringing in technology:
Set policies and define processes
Good data management relies on good policies and processes in order to conform to your data model. Then, once these are all planned out, the technology can enforce them.
Engage the right stakeholders
Executive level sponsorship is essential to keep accountability high and align the importance of the project to its impact on business results. It’s also vital that the chief data officer creates a data culture to transform attitudes towards your data model.
Subject matter experts from different business domains (product, logistics, etc.) and a business analyst should be included and consulted, to capture which data points impact which parts of the business process.
Then you get the techies to turn those requirements into a technical data model, as they’ll have the right context to understand the difference between technical and business data quality.
Accept that data modelling is a continuum
Existing and current systems and their requirements today will need to adapt to include future applications and all the business requirements that come with them—meaning it has the agility to support the vision of the company, not just its current operations.
If you’re on an M&A spree for example, your data model will need to grow and adapt to bring on the data from the other businesses.
Let’s get started: a roadmap to a stronger data model
- Start with a data model review:
- Look at the governance placed on the attributes, entities and relationships within your data model.
- Ask:
- Where is it broken?
- Where is there inefficiency?
- Where is there waste or a poor customer experience?
- In other words, what are the problems manifesting in the business on a repeated basis?
- Understand the lineage of the data in the poorly performing processes:
- Where did the data originate?
- Which systems did it pass through before you had the problem?
- Which humans had the ability to manipulate the data, and in which system?
- In which system does it start looking different?
- Where else does that data move to once the problem surfaces?
- Then rewind, look at the data flow from source to final destination, and figure out how to fix the issue with better data governance.
- Struggling to think of the fix?
- Look at how individuals in the business solve for that problem now.
- Then write the code version of that solution to implement an automated fix.
- That might be an algorithm or rule to transform the data. It could be things like a data validity check, a computation of two fields to get to a useful result that the business can use in its operations, or a scheduled cleansing rule to keep the data useable.
- Implement the guardrail you’ve come up with into the data model for any other areas where a similar error could occur.
- Rinse and repeat for your other data issues to make your data model more robust.
We like to think of this process as ‘moving the data firewall further upstream.’ The goal is to minimize opportunities for things to break by ensuring better data governance across more of the data’s journey through the organization.
What to look out for in a master data management platform
When you have the right master data management platform, that supports good data model governance, these are just some of the things you’ll be able to achieve:
- Give you the ability to apply a wide range of governance rules and policies, ensuring that data models fully support both technical and business compliance rules and standards—this is crucial to ensuring critical master data is collected, stored, managed and accessed in a robust manner.
- Capture data lineage automatically for analysis, so you can see where the data may have gone wrong.
- Apply data model fixes, allowing you to build a policy to fix the issues with the data model you found in your review.
- Leverage pre-configured data model blocks to give you quick value. The STEP MDM platform, for example, has over 200 attributes preconfigured into a data model for retail customers, based on what’s known to be most practical for them, alongside all the rules needed to correctly govern those attributes within entities, across relationships, in proper hierarchies and classifications.
Get a head start on with your data model, by downloading our handy checklist here.