One explanation, six copies: author once instead
Count the places in your company where the same explanation exists.
Take one feature. The tooltip next to it in the product. The onboarding step that introduces it. The help centre article. The paragraph in the sales deck. The reply your support team sends about it three times a week. The section in the API docs.
Six versions of one explanation, written by five people, at four different times. None of them are wrong on the day they are written. They drift apart afterwards, one edit at a time, and nobody is responsible for the gap because nobody can see all six at once.
Why copies are worse than they look
The obvious cost is maintenance. The real cost is that your users notice.
Someone reads the tooltip, then the help article, and the two say different things about the same limit. Now they do not know which one is true, so they write to support, which is the outcome all six pieces of text were supposed to prevent.
It also decides who bothers. When updating an explanation means finding and fixing six copies, the rational move is to fix the one in front of you and leave the rest. Everyone does that, which is exactly how the drift happens.
What one source actually means
Not one document. One piece of content with an address, requested by whoever needs it.
The tooltip asks for it. The onboarding step asks for it. The support screen searches for it and shows it. Your published help page renders it. Each surface decides how it looks and when it appears; none of them keep a copy.
The test is simple. Change the sentence once. If it is now correct everywhere it appears, you have one source. If you have to go and update the other places, you have six documents and a naming convention.
What stays with each surface
This is where the idea is usually oversold, so let us be precise about what does not move.
Presentation stays with the surface. A tooltip has forty words and no headings. A help page has structure. The same content can feed both, and each decides what to render.
Timing stays with your product. Which step appears when, after which event, for which user. That is logic about your product and it belongs in your product.
The decision to show anything at all stays with you. A content system that decides when to interrupt a user is a content system that will interrupt them at the wrong moment.
What moves is the words, and the fact that there is one of them.
The second effect, which is larger
Once all the explanations are in one place with addresses, you can read them together. That is when someone notices the three screens describing the same concept differently, the tooltip that explains a feature removed last year, and the four different names your product uses for the same button.
That reading is impossible when the text is spread over six systems. It takes an afternoon when it is not.
Where to start
Pick the feature that generates the most support questions. Find every place it is explained. Put one version of that explanation at one address, point the tooltip and the help page at it, and delete the copies.
Then change a word in it and watch both surfaces change. That is the whole demonstration, and it takes twenty minutes.