IFS Cloud data governance · Independent IFS Cloud practice
ERP data failures start with data nobody owns, and governance closes that gap before go-live
An ERP data failure starts when the data changes and the software processes it as configured: a dataset that changed without its consumers knowing, a modification nobody owns, or an integration tested for the first time at go-live. Each of these has a governance control that can be put in place before the next release update.
Key takeaways
- ERP data failures come from four gaps: data without a named owner, changes made in one area without checking who consumes the data, integration testing pushed to the end, and training on screens instead of the end-to-end flow.
- Federated governance from the data mesh model closes those gaps: the business domain that creates the data owns it, and a small set of enterprise rules applies to every domain.
- A data contract fixes the fields, meaning, quality rules and owner of a shared dataset such as customer, supplier, part or invoice. A change that breaks the contract is caught in testing, not in production.
- In IFS Cloud, every configuration, report, integration and modification (CRIM) should be treated as a governed item with an owner, a consumer list and a review date.
- Access rules are part of the same governance: permission sets decide what a user can do, and Custom Data Access Control decides which records an API returns to that user.
1.Why do ERP data failures happen when the software works?
ERP data failures happen when a change in one area breaks data that another area depends on, and nobody was responsible for checking. The software does what it was configured to do. The configuration was decided by one team without asking who else reads the same records.
I see four gaps behind these failures in IFS Cloud projects:
- No owner: the supplier master is maintained by purchasing, read by finance and reported on by management, and nobody is accountable for its accuracy as a whole.
- Isolated changes: a team adds a modification or changes a basic data setting for its own process without a list of the downstream consumers.
- Late integration testing: each area passes its own tests, and the end-to-end flow from order to invoice is run for the first time in the last test cycle.
- Training on screens: users learn which fields to fill, but not which process downstream fails when a field is left empty.
The general causes of failed projects are covered in why IFS Cloud implementations fail. This article deals only with the data side: who owns a dataset, what the consumers can rely on, and how a change is checked before it reaches them.
2.What does federated governance change in an ERP project?
Federated governance moves accountability for data to the business domain that creates it and keeps a short list of rules that every domain must follow. It is one of the principles of data mesh, next to domain ownership, data as a product and a self-service data platform.
- Domain ownership: the people who create the data own it. The warehouse owns inventory transactions, purchasing owns supplier and purchase order data, finance owns the chart of accounts.
- Enterprise rules: the company decides once, for all domains, how shared records are identified, which of their fields are mandatory, how data is classified and who may access it.
- Automated checks: a rule that can be tested by a script is tested by a script, so that compliance does not depend on someone remembering a meeting.
The series on this site explains the principles step by step: what data mesh is and how to implement it, applying federated computational governance and ownership and quality standards in IFS Cloud.
3.Which governance control answers which failure?
Each classic ERP failure has a data mesh counterpart and a governance control that addresses both.
| Classic ERP failure | The same risk in a data mesh | Governance control |
|---|---|---|
| Over-customised modules that make every update a retest of the core | Domains publish the same entity with different fields and meanings | One data contract per shared entity, with the owner and the allowed changes written down |
| Integration testing deferred until the last test cycle | A domain changes its data without checking the consumers | Contract tests run in every test cycle, so a breaking change fails early |
| Training on buttons, not on the flow | A domain optimises its own process and ignores the enterprise measures | A cross-domain review of every change that touches a shared entity |
4.How should CRIMs be governed in IFS Cloud?
A CRIM list is a risk register: every configuration, report, integration and modification is something the customer has to keep working after each update. Treated as a checklist, the list records what was built. Treated as a register, it also records who owns each item, who depends on it and when it is reviewed.
For each CRIM I keep the same fields as for a data product:
- an owner in the business, not only the developer who built it
- the datasets it reads and writes, and the consumers of those datasets
- the reason it exists, so that a new standard function can replace it when one appears
- a review event, such as the next release update, and a retirement date when the reason disappears
A modification without these fields is the one that turns an update into weeks of retesting. The inventory before each update is described in IFS Cloud upgrade readiness, and the difference between update types in service update vs release update. Where a rule can live in configuration rather than in code, IFS Cloud custom defined events are, in the words of the documentation, “defined as parts of the customer solution configuration” and fire when an entity is created, updated or removed (IFS Cloud Events).
5.What are the steps to put this governance in place?
Governance that prevents data failures is built in four steps, and each step leaves an artefact the project can check.
- Write the contract. For each shared entity (customer, supplier, part, invoice, shipment) write down the fields, their meaning, the mandatory fields, the quality rules and the owner, and give the contract a version.
- Turn rules into checks. Every rule that can be tested by a query or a script is tested in each test cycle and after each update. Access is part of this: a permission set is “a collection of access grants assigned to objects and privileges” (IFS Cloud permission sets), and Custom Data Access Control adds record-level rules on top, because “Permission Sets grant users access to Projections (APIs) but do not manage data access” (Custom Data Access Control).
- Name the people who connect domains. One experienced analyst per pair of domains that share data, for example sales and finance, reviews each change to a shared entity before it is accepted.
- Measure the time from request to data. Track how long it takes from a request for a new field or report to its delivery, and how often an existing dataset is reused instead of rebuilt. A long lead time shows that governance has become a queue.
The rhythm in which owners meet to accept these changes is covered in IFS Cloud governance cadence. Standing rules that outlive the project go into the enterprise book of rules.
6.Where do data quality and metadata fit in?
Ownership and contracts decide who is accountable for a dataset. Two further disciplines keep the content right and findable: data quality rules and master data management, described in ERP data quality in IFS Cloud, and metadata, which records what each field means and where it comes from, described in metadata management in IFS Cloud.
When a project needs help setting up data ownership and contract checks for its IFS Cloud data, that work is part of my services for IFS Cloud.
7.Frequently asked questions
Why do ERP projects fail on data even when the software works?
ERP projects, including IFS Cloud projects, fail on data when a change in one area breaks records that another area depends on and nobody owns the check. The gaps behind them are data without a named owner, changes made without a list of consumers, integration testing left to the last cycle and training on screens rather than the end-to-end flow.
Is data mesh only for big data platforms?
Data mesh is an operating model for data ownership, not a big data product. A mid-sized IFS Cloud implementation can apply it by making each business domain, for example the warehouse or purchasing, accountable for the accuracy of the data it creates, under a short list of enterprise rules.
What is a data contract in an IFS Cloud project?
A data contract in an IFS Cloud project is a versioned description of a shared dataset, such as customer, supplier or part: its fields, their meaning, the mandatory fields, the quality rules and the owner. Consumers rely on the contract, and a change that breaks it is caught in testing.
How do CRIMs relate to data governance in IFS Cloud?
CRIMs, the configurations, reports, integrations and modifications made on top of standard IFS Cloud, change how data is created and read. Governing them means giving each CRIM a business owner, a list of the datasets and consumers it affects, the reason it exists and a review date such as the next release update.
Which IFS Cloud functions support access governance?
IFS Cloud controls access in two layers. Permission sets grant users access to projections and the actions in them. Custom Data Access Control adds access control lists based on the attributes of a projection entity, with separate create, read, update and delete permissions, and is switched on with the system parameter Data Access Control.
8.About the author
Dariusz Myśliwiec: 18 years hands-on with IFS (Apps 7.5 to 10 and IFS Cloud). IFS Certified Associate Consultant. PRINCE2® 7. Independent practice based in Kraków, delivered remotely across Europe and globally. You talk to the consultant who does the work, not a sales layer.
Selected clients from my IFS work: Fugro · LGC · BVI Medical · Betafence (PRÆSIDIAD) · Barlinek · NGK Ceramics · Newag · Oleofarm.
IFS is a registered trademark of IFS AB; this practice is not affiliated with IFS AB.
Check your three most shared datasets
Take customer, supplier and part. For each one, can you name the owner, find the written rules and say which test would fail if someone changed a field? Each missing answer is a place where the next data failure can start.
Book a call about data ownership in your project See services for IFS Cloud
