What a source of truth means
Every business ends up with several systems describing the same activity. The ad platform reports its conversions, the analytics property reports its events, the CRM reports qualified enquiries and the accounts report money received. None of them is wrong, but they will never agree. A source of truth is the agreement about which of them settles a particular question, made before the question is asked in anger.
The important word is particular. There is rarely one dataset that should decide everything. Revenue and refunds belong in your own records. Whether a campaign is worth more budget than another usually belongs to the ad platform, because only it can compare its own auctions consistently. Website behaviour belongs to analytics. Naming a single winner across all three is what causes the trouble.
Why a source of truth matters
Without one, reporting turns into negotiation. Someone brings the flattering figure, someone else brings the unflattering one, and the meeting is spent deciding whose screen counts rather than what to do next. Decisions get postponed, and the person who shouts loudest wins by default.
It also makes faults visible. When one dataset is designated authoritative and the others are compared against it on a schedule, a sudden change in the difference is a signal that something broke. If nothing is authoritative, a broken tag simply looks like another disagreement.
Common mistakes with a source of truth
The first is choosing by convenience: whichever tool is easiest to open, or whichever shows the better result. The second is declaring one and never writing down what it actually measures, so the same dashboard means different things to different people within months.
A third is expecting the designation to close the gap. Naming your CRM authoritative does not make the ad platform agree with it; it only decides which figure goes in the board pack. The gap still needs to be measured through regular reconciliation.
How to act on it
Write a short table listing your recurring questions and, beside each, the system that answers it and how that system defines the measure. Include the filters, the time zone and whether repeat actions count. This is dull work that pays for itself the first time two people disagree.
Then make the choice visible where people read the numbers. Label each figure on the dashboard with its source, keep the ad platform’s own view for optimisation decisions, and reserve your business records for anything involving money. Consistent definitions built into reporting and dashboards mean the answer stays the same whoever opens the report, and the discussion moves on to what the number should be next quarter.