What is a single source of truth, and why does it matter?

The phrase sounds like enterprise architecture jargon, and in most SMEs it means something much smaller: for any given fact about your business, one system is the one you believe.

Not one system for everything. One system per thing. Your accounting package is probably already the single source of truth for what has been invoiced, and nobody argues about it. The problem is everything else.

What it looks like when you do not have one

• Two people produce the same report and get different numbers, and the meeting becomes about the numbers rather than the decision.
• Someone updates a client’s details in one place and nobody knows which other places also hold them.
• Every month-end involves an export, a lookup formula and an hour of squinting.
• New starters are told which spreadsheet to trust, verbally, by whoever remembers.

None of that is a data problem exactly. It is an unmade decision, repeated daily by everyone who has to work around it.

How you actually get one

It is less technical than it sounds. Three steps, in order.

1. List the things your business tracks. Clients, jobs, properties, cases, invoices, leads. Usually five or six.

2. For each one, name the system that wins. Not the system that has the most data in it. The one that will be believed when two systems disagree. This is a management decision, not an IT one, and it takes an hour with the right people in the room.

3. Make everything else read from it. Either by connecting the systems so the definitive value flows outward, or by making the other copies clearly secondary. Both work. What does not work is leaving two systems as equals and hoping.

The step people skip is the second one, and it is the only one that cannot be outsourced.

Why automation depends on it

Software cannot resolve a disagreement between two of your systems. It has no way of knowing that the CRM figure is usually stale or that the spreadsheet is the one finance actually uses. If you have not decided which source wins, every automation you build either encodes a guess or has to be told, case by case, and you end up maintaining that logic forever.

Once the decision exists, automations become small. Read from the definitive source, do the thing, write back. That is why the second and third automation in a business usually cost a fraction of the first: the expensive part was never the code.

What it does not require

It does not require replacing your systems, buying a data warehouse, or a transformation programme. Plenty of SMEs get there by keeping every tool they already pay for and adding a thin layer that reconciles them and decides the answer.

It is worth being honest about the other case though. If the definitive version of something important lives in a spreadsheet that one person maintains, no amount of connecting will fix it, and building somewhere proper for that data to live is the cheaper route in the end.