Certified and promoted datasets: how people know which data to trust

When an organization first adopts Power BI, there are a handful of datasets and everyone knows which is which. As it grows, that changes: datasets accumulate, some official and carefully built, some thrown together for a one-off need, some outdated, and a person building a new report faces a question that quietly matters a lot: which of these datasets should I actually trust and build on? Power BI's answer is endorsement, the ability to promote and certify datasets, marking the trustworthy ones so reporting builds on the right foundation. Here is what it is and why it matters for a growing estate.

The problem: which data do I trust?

As an organization accumulates datasets, a real problem emerges: people need a way to know which ones are trustworthy to build on. When there are many datasets, some are the official, well-built, authoritative sources, and others are ad hoc, incomplete, or outdated, left over from a one-time need or built outside any governance. To someone creating a new report, they can all look similar in the list, and without a signal, there is no way to tell the authoritative dataset from the unofficial one.

This matters because the choice has consequences. Build a report on the right dataset, and it inherits trustworthy, consistent data. Build it on the wrong one, an outdated copy, an unofficial version with its own definitions, and the report spreads inconsistent or incorrect data, often without anyone realizing. In a growing estate, people build on whatever they find, and if the trustworthy datasets are not marked, they will sometimes find and build on the wrong ones, which quietly propagates bad data through the organization. The problem is not that bad datasets exist; it is that nobody can tell which is which.

Promoting and certifying

Endorsement solves this by letting datasets be marked according to how trustworthy they are, at two levels. Promoting a dataset flags it as ready and recommended for others to use, a signal from its owners that it is good to build on. Certifying a dataset goes further: it marks the dataset as reviewed and authoritative, the official source for its data, usually after a defined review by whoever governs data in the organization.

The distinction is roughly recommended (promoted) versus officially vetted and authoritative (certified). Both put a visible mark on the dataset, so someone browsing for data to build on can see at a glance which datasets are endorsed and at what level. That visible signal is the whole point: it turns an undifferentiated list of datasets into one where the trustworthy, official sources stand out, so people can find and build on them instead of guessing. A certified dataset says, in effect, "this is the authoritative source, vetted, build on this," and that message reaches the person who needs it at exactly the moment they are choosing what to use.

Why certification has to be controlled

For a certified mark to mean anything, the ability to certify has to be controlled. Certification is typically granted not by anyone who happens to own a dataset but by whoever governs data in the organization, a data governance function or designated owners, and only after the dataset meets the organization's standards for a reviewed, trustworthy source.

This control is essential, because the value of certification is entirely in its credibility. If anyone could certify anything, the mark would signal nothing, it would just be a label people apply to their own work, and the trusted-data problem would return. By restricting certification to a governed process, the organization ensures a certified mark actually means the dataset has been vetted and can be relied on as official. The restriction is what gives the mark its weight. So endorsement is not just a label feature; it is a governance mechanism, where the authority to certify is deliberately limited so that certification remains a trustworthy signal rather than a meaningless badge. Promotion is lighter-weight, owners recommending their own, but certification carries the organization's authority precisely because it is controlled.

How it fits the bigger picture

Endorsement works alongside the other foundations of a well-governed estate. It pairs naturally with shared datasets: when an organization builds its reporting on a few authoritative shared models, those are exactly the datasets to certify, so endorsement marks the shared, governed models as the official sources and steers people toward them. It complements workspaces and apps, which give the estate structure, by adding a trust signal on top of that structure.

Together these make a growing estate navigable and trustworthy. The organization ends up with a clear set of certified, authoritative datasets that people can find and build on, and the signal keeps reporting resting on the right foundation as the estate grows. Without endorsement, datasets multiply with no indication of which to trust, and the estate drifts toward inconsistency as people build on whatever turns up. With it, the authoritative data is visible, the unofficial data is clearly not endorsed, and the sprawl is contained because people are guided to the right sources. A good partner uses endorsement as part of a broader governance approach, so that as the estate grows, the trusted data stays easy to find and build on, rather than getting lost in a pile of datasets no one can sort through.

The point in short

Endorsement is how a growing estate tells people which data to trust:

• As datasets accumulate, people need to know which to build on, and without a signal they build on whatever they find.
• Promoting recommends a dataset; certifying marks it as reviewed and authoritative.
• Certification is controlled by data governance, which is what makes the mark credible.
• It works with shared datasets and workspaces to keep the estate's trusted data visible and the sprawl contained.

The throughline is that trustworthy data is only useful if people can find it and tell it apart from the rest, and as an estate grows that stops being obvious on its own. Endorsement, marking the reliable datasets as promoted and the authoritative ones as certified, is how an organization keeps its trusted data visible and keeps reporting building on the right foundation, which is one of the quiet but important parts of governing a Power BI estate so it stays trustworthy as it scales rather than dissolving into a sprawl of datasets no one is sure about.