Single source of truth (SSOT): what it means and how to actually build one

Guide · 6 min read · Updated July 2026

A single source of truth (SSOT) is the one agreed place where a given fact is authoritative — every other copy of that fact is derived from it, references it, or is wrong. It is not one giant system that holds everything. It is a design rule: one fact, one home. The renewal date has a home. The headcount has a home. The refund policy has a home. Everything else points there.

Most articles on SSOT stop at the definition. This one is written from running a company on the principle: what the term actually means, what it is often confused with, how to establish it fact class by fact class, and why the AI era has quietly turned it from good hygiene into a prerequisite.

One fact, one home: why this is the rule everything depends on

Every data problem a company complains about — conflicting reports, "which spreadsheet is right?", the CRM disagreeing with finance — is downstream of the same root cause: the same fact living in two places with nobody nominated as authoritative. Once two homes exist, they will diverge; the only question is when, and who gets misled first.

The fix is unglamorous and organisational before it is technical: for each class of fact, name the one place it lives, write that decision down somewhere everyone can find it, and treat every other appearance of the fact as a copy that must trace back. The write-it-down part matters — an SSOT that lives in people's heads is just a shared assumption, and shared assumptions are where drift starts.

SSOT vs system of record vs data warehouse

Three terms get blurred together, and the blur causes real design mistakes:

The practical takeaway: you cannot buy an SSOT. You can only decide one, document it, and hold your systems to it.

Establishing SSOT per fact class

The realistic version of SSOT is per fact class, and different classes have naturally different homes:

That last class is where we spend our working lives: turning signed contracts into structured data and holding the CRM, ERP and billing systems to it is what contract-to-cash intelligence means, and it exists precisely because "the contract is the SSOT for commercial facts" is easy to say and hard to operate manually.

The drift problem

Naming a source of truth is the start, not the finish, because facts do not stay put. The renewal date gets copied into the CRM, a forecast spreadsheet, a QBR deck and a billing schedule — and then the contract is amended. Now four copies are stale, each one still looking perfectly authoritative to whoever reads it. Drift is not a rare failure; it is the default behaviour of copied data, and it compounds quietly until an invoice, a forecast or a renewal conversation surfaces it at the worst moment.

The counter-measure is reconciliation: routinely comparing the copies against the named source and flagging every mismatch. An SSOT without reconciliation is a label; with it, it becomes an enforced property of your data.

SSOT in the AI era

Here is why this decades-old idea suddenly has teeth. Companies are connecting AI assistants to their systems and documents — building what we'd call a company brain — and an AI answers from whichever "truth" it reads. If two systems disagree, the AI does not adjudicate; it amplifies. One wrong copy used to mislead the person who opened that record. The same wrong copy behind an AI assistant misleads the entire company, fluently and on demand — we've unpacked that failure mode in why AI gives wrong answers about your company.

So the order of operations matters: SSOT first, AI second. An assistant grounded in named, reconciled sources is genuinely useful; the same assistant pointed at disagreeing systems is a fiction generator with your logo on it.

Practical steps

  1. List your fact classes — process, customer, commercial, financial, people — and name one home for each. Write the map down; a page in a shared doc is enough to start.
  2. Mark every other location of each fact as derived, and note what refreshes it.
  3. Give each home an owner and a freshness cadence.
  4. Set up reconciliation for the classes where drift is expensive — commercial facts first, because that is where drift becomes money. If your CRM, billing system and signed contracts have never been compared, that is worth seeing.
  5. Only then connect AI. If you are building towards an assistant that answers company questions, our guide to the AI company knowledge base covers the full structure, and the free company-brain starter template gives you a file layout with the SSOT registry built in.

One fact, one home, written down, reconciled. It is not a platform purchase; it is a decision — and it is the decision every trustworthy answer in your company will trace back to.