Single source of truth for client data: a leader's guide

Definition: A single source of truth is one agreed, trusted record for each piece of data, which every system and team works from. For client data, it's one authoritative version of each client, often called a golden record.

In short: You have a single source of truth when two teams asking for the same client's details get the same answer. To build one, find where your data debt sits, agree which record wins and who owns it, and consolidate step by step, not in one big project. AI needs it: if nobody trusts the data, nobody will trust what AI does with it.

In manufacturing, there's one bill of materials. Everyone building the product works from it. If two teams had two versions, you'd have a problem before lunch. Most service businesses run on two, three or four versions of the same client record, and have stopped noticing.

What is the two-answers test?

It's the quickest check I know. Ask two people in different teams for the same client's current details, and compare what you get.

• Same answer: you're closer to a single source of truth than you think.
• Different answers: you have data debt, and your AI will inherit it.

Try it on your ten biggest clients. [Add what you've seen when leaders run this test, and only what you've actually seen.]

Why does AI make this urgent?

Because AI can't tell which record is right. If your client data lives in several systems with several versions, an AI will pick one and act on it, faster than a person would and with more confidence. Scattered data is bearable while people work around it. It gets expensive the moment AI has to rely on it. In regulated operations, I'd add that if you can't tell which record is right, you can't prove it to a regulator or a client.

Why do most attempts stall?

In my view, two reasons.

1. It's run as a big-bang IT project. It's owned by one team, needs months of migration before anyone sees value, and the business loses patience.
2. Nobody has agreed whose record wins. Each system usually has its own owner. You can't fix that with a migration plan. It takes a conversation with the people who own the systems.

What does it look like in practice?

A corporate services and fund administration provider, operating across multiple time zones.

• Before: client records split across four systems, each owned separately. No master data, so no agreed version of the core records. No shared view of how the key systems connected.
• What it did: ran a full tech stack analysis and mapped every integration gap. Ran a data debt assessment and mapped the data across all four systems. Began building towards one golden source, a single record everything else works from.
• After: four systems consolidating into one, a single source of truth underway, and data debt reducing over time.

It didn't happen in one go. It happened in steps, with the owners of each system involved.

The ownership map

Before you touch any technology, fill in one row for each core record (client, account, contact and so on):

• Record: what it is.
• Where it lives: every system that holds a version.
• Owner: who's accountable for it, by name.
• Which wins: the system whose version is right when the others disagree.
• Who's affected: the teams that rely on it.

If you can't fill in a row, you've found a problem worth solving first. The conversation it forces, about whose record wins, is the hardest and most valuable part.

How do you build one?

1. Find where the data debt sits. Run a data debt assessment: where records are duplicated, out of date or in conflict.
2. Map the data across every system, and how it flows.
3. Agree which record wins, and who owns it, using the ownership map.
4. Build towards a golden source in steps, and connect the systems around it.
5. Track the debt. It should go down over time. If it isn't, something upstream is still creating it.

The ten-client test: try it this week

1. Pick your ten biggest clients.
2. Ask two teams for each one's current details, without telling them why.
3. Record every mismatch, in name, address, contacts, status or anything else that matters.
4. Count them. Then trace where each difference came from.

Ten clients takes a morning. It tells you more about your data than a report will, and it shows you where to start. Client data is one of the five problems we help with, and it starts with a free discovery call.

Key takeaways

• A single source of truth is one agreed record everything works from, not necessarily one database.
• Use the two-answers test to see where you stand, and the ownership map to fix it.
• Don't build it in one go. Agree which record wins, then consolidate in steps.

FAQs

What is a single source of truth?

A single source of truth is one authoritative, agreed version of a piece of data that all systems and teams refer to. For client data, it's one trusted record for each client.

What is a golden record?

A golden record is the single, agreed version of a record, such as a client, that all other systems draw on. It's built by resolving duplicates and conflicts between sources.

What is master data?

Master data is the core information a business depends on, such as clients and accounts, shared across many systems and processes.

What is the difference between a system of record and a source of truth?

A system of record is where a piece of data is created and held. A source of truth is the agreed answer to which version is right. They're often the same, but not always.

Does a single source of truth mean one database?

Not necessarily. It means one agreed answer to which record is right. The data can still be used by many systems as long as they all work from the agreed record.

What is the two-answers test?

It's a quick check: ask two people in different teams for the same client's details, and see whether the answers match.

Sources: Results from the fund administration provider as described in its case study. [Add link.]