"Single source of truth" is one of the most appealing phrases in enterprise software, because it names something every leader genuinely wants: one place to look, one number everyone trusts, an end to the meeting where two teams present different figures for the same thing and spend the first twenty minutes arguing about whose spreadsheet is right. Vendors know this, which is why the phrase appears in nearly every pitch, attached to nearly every platform, as though it were a feature you could purchase and install.

It is not a feature. It is a discipline, and the distinction is the reason so many organizations buy the platform, complete the implementation, and still sit in that same meeting a year later, now with an expensive system and two teams still presenting different numbers. The technology was never the hard part. The hard part is a human agreement that the technology cannot make for you, and mistaking one for the other is why these projects fail at the rate they do.

They fail, and they fail on the same thing

This is not a rhetorical claim. The failure is measured, and it is severe.

Gartner has found that 75% of master data management programs fail to meet their business objectives, a figure that has worsened nearly every year since 2015, and has separately projected that 80% of data governance initiatives will fail by 2027. These are not marginal disappointment rates. They describe a discipline where failure is the normal outcome, despite enormous spending, in a market Gartner sizes in the tens of billions.

And the diagnosis is consistent across the research. The programs do not fail because the technology could not unify the data. They fail on ownership, decision rights, and accountability, the human questions the technology was never able to answer. Academic work on master data management reaches the same conclusion: success requires strictly defined data owners with explicit authority to make binding decisions, and organizations struggle precisely because that authority is rarely established. Even Gartner's own definition of the field describes it as a technology-enabled business discipline in which business and IT work together, not a product that produces truth on installation. The word "discipline" is doing the work in that sentence, and it is the word buyers skip.

The question the platform cannot answer

Here is the moment where a single source of truth is actually won or lost, and no software attends it.

Two of your teams produce a number for the same thing, and the numbers differ, and both are defensible. Finance and operations calculate occupancy differently, one excluding down units, the other including them, each for a reason that makes sense from where they sit. Two departments define a "customer" or an "active lease" or "revenue" with slightly different boundaries, and both definitions are correct for their own purposes. This is not a data quality problem. Both numbers are right. It is a definitional problem, and it has no technical answer, because the technology cannot decide which team's legitimate definition should be the authoritative one for the enterprise.

That decision is a governance decision, and it is often politically uncomfortable, because choosing finance's definition of occupancy over operations' definition is choosing one team's framing over another's, and the team whose definition loses experiences it as a loss. Which is exactly why organizations avoid making it, and why they reach for a technology project instead. Buying a platform feels like progress and requires no one to lose an argument. Deciding whose definition wins feels like conflict and requires someone with authority to impose an answer that not everyone will like. So the platform gets bought, the decision gets deferred, and the two numbers survive inside the new system as comfortably as they did in the old spreadsheets.

There is a clarifying test buried in the research, and it is worth adopting directly: when two parts of your organization produce different figures for the same thing from the same governed platform, whose number is correct? If the honest answer is "it depends on who you ask," you do not have a single source of truth, regardless of what you installed. The project was never actually started.

What the discipline actually requires

A real single source of truth rests on four things, only one of which is technology, and it is not the hard one.

• Agreed definitions - For every number that matters, a single authoritative definition that the whole organization accepts, even the teams whose preferred version did not win. This is the foundation, it is achieved through decision rather than software, and it is where most of the discomfort lives.

• Named owners with authority - Each critical data domain needs an owner empowered to make binding decisions about it, and empowered is the operative word. An owner who can be overruled by any team that dislikes the ruling is not an owner, they are a suggestion, and the data will fracture again the moment a powerful function disagrees. The research is blunt that authority without teeth is the most common failure.

• Enforced discipline over time - A single source of truth is not achieved once, it is maintained continuously, because entropy is constant: new systems get added, definitions drift, exceptions get made, and a truth that was single last year fractures quietly this year if nobody is maintaining it. This is why it is a discipline rather than a project. Projects end. This does not.

• And, yes, the technology - A platform that can actually hold the agreed-upon single version and serve it consistently to everyone. This is real and necessary, and it is the part vendors sell, and it is genuinely the easy part, in the specific sense that it is the part money can buy directly. The other three cannot be purchased.

The reason projects fail is that organizations buy the fourth thing and skip the first three, because the fourth is purchasable and the otYer three require someone to make uncomfortable decisions and sustain them against constant pressure.

The honest part

Several qualifications matter, because the strong version of this argument overreaches.

First, a genuinely uncomfortable truth the research itself concedes: a single source of truth may not always be the right goal. Multiple legitimate truths can coexist across functions, and forcing finance and operations to use an identical occupancy figure for every purpose can destroy information that each definition was capturing for a real reason. The sophisticated version of this is not one number imposed everywhere. It is agreed definitions with an explicit, authoritative enterprise version for cross-functional decisions, while allowing functions their own views where those views serve a real purpose. Insisting on literal singularity where plurality is legitimate is its own failure.

Second, technology is not nothing, and the reverse error is as real as the one this piece warns against. An organization that gets the governance right but runs it on disconnected systems and manual reconciliation has made agreement harder to sustain, because the discipline has to fight the infrastructure. Good technology makes the discipline maintainable; it just cannot substitute for it. The argument is about sequence and emphasis, not about dismissing the platform.

And third, the governance work is genuinely hard, and saying so is not a reason to avoid it, but it is a reason to be honest that "just decide whose definition wins" is easy to write and difficult to do. It requires an executive sponsor with authority across functions, the willingness to make some teams unhappy, and the persistence to hold the line for years. Underestimating that difficulty is how organizations end up reaching for the platform as an escape from it.

Do the hard part, then buy the easy part

The discipline reduces to correct sequence. Before buying a platform to create a single source of truth, do the work only you can do: identify the numbers that matter, surface where definitions genuinely conflict, and make the authoritative decisions about which version wins, with an owner empowered to enforce each one. That work is uncomfortable and unglamorous and it is the actual project. The platform comes after, to hold and serve the agreement you have already reached, and connected systems make that agreement far easier to maintain than a fragmented estate would, which is the legitimate role technology plays here.

An organization that does this in the right order gets a single source of truth that survives, because the humans agreed first and the system enforces what they agreed. An organization that buys the platform first gets an expensive place to store the same disagreement it always had. The phrase describes something worth wanting. The mistake is believing you can acquire it rather than build it, and the tell is always the same question: when two teams disagree about the same number, does your organization have a way to decide who is right? If not, no platform will supply one.

FAQs

Q1. Isn't a single source of truth something we can buy?
You can buy the technology that holds and serves it, but not the thing itself. A single source of truth requires agreed definitions, empowered owners, and sustained enforcement, none of which a platform provides. This is why so many organizations complete an expensive implementation and still have teams presenting conflicting numbers: they bought the storage and skipped the agreement it was supposed to store.

Q2. Why do these projects fail so often?
Gartner finds roughly 75% of master data management programs fail to meet business objectives and projects 80% of data governance initiatives will fail by 2027. The consistent reason is not technical, it is human: unresolved ownership, undefined decision rights, and a lack of accountability. The technology usually works. The organization never made the definitional decisions the technology needed to enforce.

Q3. What is the definitional problem you describe?
It is when two teams produce different numbers for the same thing and both are legitimately correct, because they define the thing differently for defensible reasons, such as finance and operations calculating occupancy with different treatment of down units. This is not a data quality issue, since both numbers are right. It has no technical answer, because software cannot decide which valid definition should be authoritative for the enterprise.

Q4. Why don't organizations just decide whose definition wins?
Because it is politically uncomfortable. Choosing one team's definition over another's is imposing a loss on the team whose version does not win, and that requires authority and the willingness to make people unhappy. Buying a platform feels like progress without conflict, so organizations reach for the purchase and defer the decision, which is exactly why the disagreement survives inside the new system.

Q5. What actually creates a single source of truth?
Four things: agreed authoritative definitions for the numbers that matter, named owners empowered to make binding decisions, continuous enforcement because definitions drift over time, and a platform to hold and serve the result. The first three are governance decisions that only the organization can make. The fourth is the purchasable part, and it is the easy one precisely because money can buy it directly.

Q6. Is a single source of truth always the right goal?
Not necessarily. Multiple legitimate truths can coexist across functions, and forcing an identical figure everywhere can destroy information each definition was capturing for a real reason. The mature version is agreed definitions with an explicit authoritative enterprise version for cross-functional decisions, while allowing functional views where they serve a genuine purpose. Literal singularity imposed where plurality is valid is its own mistake.

Q7. Does this mean technology doesn't matter?
No. Governance run on disconnected systems and manual reconciliation is far harder to sustain, because the discipline has to fight the infrastructure continuously. Good technology makes the agreement maintainable and is genuinely necessary. The point is one of sequence and emphasis: the platform cannot substitute for the governance decisions, but it makes holding them enormously easier once they are made.

Q8. What is the single question that reveals whether we have it?
When two parts of your organization produce different numbers for the same thing from the same system, is there a way to decide who is correct? If the answer is "it depends on who you ask," you do not have a single source of truth no matter what you installed, because the authoritative definition and the empowered owner do not exist. That question exposes the gap faster than any technical audit.