Scattered data across spreadsheets, your CRM and your ERP: building a single source of truth

Plenty of leadership meetings start the same way: someone presents a revenue figure, someone else says their system shows a different one, and the next twenty minutes go on arguing over which number is right. It isn't a spreadsheet problem or a people problem: it's that the business never agreed on which data rules. Until that's settled, decisions get delayed and trust in the numbers erodes. This piece is about getting out of that loop: what "a single source of truth" means as an operating rule — not a promise — and how to get there without an endless technical project.

Having several systems isn't the problem

Spreadsheets, a CRM, an ERP, loose files and the odd reporting tool coexist in plenty of organisations, and there's nothing wrong with that: each one does its job. The problem shows up when each system tells a different version of the business. Sales don't match between the CRM and the ERP, for example, because one counts won deals and the other counts issued invoices, or because each uses a different date. Neither is "wrong" — they measure different things, and nobody agreed which one counts for deciding. Having several systems is normal. Having several truths isn't.

Why the same data doesn't add up

The mismatches usually come from a handful of familiar causes:

• Data gets entered in different places, with different criteria, at different times. What's a closed sale in one system is still an opportunity in another.
• Nobody agreed on the definition. What counts as an "active customer": one who bought this year, one with a live contract, or one who opened a support ticket? When does a sale count: on signature, on invoice, or on payment? Which date rules: order date, invoice date, or payment date? Without those definitions, two honest people pull two different numbers — and both are right.
• Manual copies that drift out of sync. Every time data jumps from a system into a spreadsheet by hand, a new version is born that starts ageing on its own.

The underlying pattern is always the same: it isn't a technical failure, it's a missing agreement.

What "a single source of truth" means (as a rule, not a promise)

A single source of truth doesn't mean perfect data or one single system. It means something more modest and more useful: for every piece of data that matters, there's one source that rules, one agreed definition, and someone who owns it. It's a business agreement before it's a technical solution. Once it exists, meetings change subject: instead of arguing over who has the right number, people argue over what to do with it. That's the tell that it's working — the conversation moves from the data to the decision.

What it is not

Worth clearing up the usual misunderstanding. A single source of truth is not:

• One piece of software that does everything, and it doesn't mean dropping spreadsheets: the tools can stay plural.
• Perfect, error-free data. It's data that's reliable and owned, not infallible.
• A months-long technical project. You start with the handful of numbers that actually drive decisions, not with everything.
• A pretty dashboard. If the panel draws from sources that don't agree, it lies with better design.
• Freezing the data. It stays alive and changing, just with a definition and an owner behind it.

In one line: a single source of truth is an agreement about what to believe, not a tool you buy.

How to get there (without a technical project)

You don't need to redesign your systems to start. Five steps, in plain business language:

1. List the handful of numbers that drive decisions. Customers, sales, orders, stock, margins… almost always fewer than it feels like.
2. Agree the definition of each one. What counts and what doesn't, with examples: what's an active customer, when a sale counts, which date rules.
3. Pick the source that rules for each one. If the CRM rules on opportunities and the ERP rules on invoicing, write that down.
4. Assign an owner. Someone who maintains that data, decides which definition holds, and answers when two numbers don't match. It isn't a new role — it's an explicit responsibility.
5. Decide how it reaches the point of decision. Reporting should draw from the agreed sources, ideally without a weekly manual reconciliation.

And revisit it: it's a living agreement, because the business changes. That work of ordering and connecting the data is usually where we start.

Why this comes before automation or AI

Agreeing a single source of truth isn't a luxury for big companies: it's the base almost everything else rests on. Automating a process that leans on data that doesn't add up just spreads the error faster — it's the same idea as getting things in order before you automate. And applied AI needs enough reliable data to be worth anything: on top of contradictory data, it inherits the contradiction. Before you "put AI" on something or automate it, you need numbers you can trust.

Three questions to see where you stand

About your most important decisions:

1. Do you know, for each one, which data rules?
2. Is there an agreed definition of that data, or does each team use its own?
3. Is there someone who answers when two numbers don't match?

If the answer to any of those is "no," that's the work — and it usually pays off more than any new tool.

The goal isn't the perfect system; it's to stop arguing about numbers and start deciding with them. That — agreeing what to believe and who owns it — is our approach.