Single source of truth (SSOT)

What is a single source of truth?

A single source of truth (SSOT) is the agreement that, for any given fact, there is one place your organisation treats as authoritative and everyone reads from it. Ask how many active customers you have or what last quarter's revenue was, and the answer comes from one agreed place, so every report and dashboard traces back to the same figure.

The problem it removes is familiar. Sales quotes one revenue number in the monthly meeting, finance quotes another, and the next twenty minutes go to whose spreadsheet is right instead of to the business itself. An SSOT ends that argument by settling in advance where the trusted answer lives.

It is a state you reach, not a product you buy. And it does not mean cramming everything into one system. Your CRM stays the home of sales contacts, your accounting package the home of invoices, your HR tool the home of employees. What you agree is which place is authoritative for each fact, so nobody keeps a private version beside it.

How a single source of truth is built

An SSOT rests on three moving parts, and only one of them is technology.

First, each fact is traced to its system of record, the application where that data is born and owned. Invoices come from the ERP, sales contacts from the CRM, and when copies disagree the system of record wins, so there is always a clear origin to point back to.

Second, the meaning is pinned down and computed once. A business glossary records what "active customer" or "net revenue" actually means, and a semantic model or metrics layer turns that agreed meaning into logic every tool runs. Define the metric once and it is served the same way everywhere: dbt's Semantic Layer refreshes a metric across every report and app the moment its definition changes, and a certified Power BI semantic model gives report builders one model the organisation has marked as reliable and authoritative.

Third, the data is consolidated so reports query one place instead of ten. A data warehouse or lakehouse pulls copies of your source data into one central repository, so reports read from that rather than reaching into ten live systems. Where the same customer or product appears across systems under different spellings, master data management resolves it to one shared version first.

Single source of truth, system of record, and golden record

These three get tangled because each one sounds like "the real one," but they answer different questions.

A system of record is about authority to write: the application that owns a fact and where a correction has to be made to count. A golden record is about resolution: the single best-value version of one entity, such as a customer, that master data management assembles from every system holding a copy. A single source of truth is about consumption: the one agreed place people go to read a value, usually a warehouse or a certified model that harmonises and republishes what the systems of record produce.

They stack rather than compete. Take a customer whose billing address differs between the CRM and the ERP. The ERP is the system of record for billing, so its value wins; master data management merges the duplicates into one golden record; the warehouse republishes that resolved address, and the figure on every invoice report is now the single source of truth for reporting. The invoice is still born in the ERP. The warehouse only agrees to be the one place everyone reads it.

One source of truth per domain

A literal single source of truth for everything, one place that holds every fact for the whole company, is neither achievable nor desirable. Real organisations are federated: finance, sales, operations and logistics each run their own systems and know their own data best. Pushing all of it through one central model creates a bottleneck and a single point of failure, and it slows the decisions the SSOT was meant to speed up. This is the reasoning behind data mesh, where each domain owns and publishes its own data as a data product instead of shipping everything to a central team.

So the useful version of the idea is scoped: one source of truth per data domain. Customer data has one agreed source, product data another, financial reporting a third, each with a named owner who settles definitions and approves changes. Holding that together across domains is what data governance is for.

What to watch out for with a single source of truth

• Settle the definitions before the technology. Building a warehouse while sales and finance still disagree on what a customer is only moves the argument downstream. Agree the meaning and the owner first, then implement it.

• It decays without maintenance. Source systems change, connections break, a new tool arrives. An SSOT that nobody tends stops matching reality within a year and loses the trust it was built for.

• Shadow spreadsheets creep back. If the official report is slow or hard to find, someone exports a private copy and shares it around. Treat every shadow spreadsheet as a signal that the agreed source is falling short, and fix the source rather than scold the person.