Nobody sets out to build a bad brand architecture. Bad architecture is what you get when a series of individually sensible decisions are taken years apart by different people with no view of the whole.
A product launch needs a name, so it gets one. An acquisition comes with a name customers know, so it keeps it, for now. A regional team builds something local and calls it something local. A sub-brand is created for a campaign and then does not die, because nothing ever dies. Ten years later you have forty names, no logic, and a new marketing director asking why the portfolio is incomprehensible.
Nobody decided that. It accumulated. And the reason it accumulates is that every individual decision is small enough to be made locally, while the cost of the whole is felt centrally and only much later.
The costs are real and mostly invisible in the year they are incurred. You are paying for every name: trademark, maintenance, design, media to build awareness, and the sales time spent explaining the relationship between things. You are splitting equity that could compound in one place across several that never reach escape velocity. And you are making buying harder, because customers have to learn your internal structure to purchase from you, which is an unreasonable thing to ask of someone spending money.
The decision underneath all of it is simpler than the frameworks suggest. For any given thing, you are choosing between three options.
It carries the parent name, and therefore borrows the parent's position and contributes back to it. Cheapest, fastest, and correct far more often than it is chosen. The cost is that it must fit the parent's position, and anything that does not will damage it.
It gets its own name, endorsed by the parent. Useful when the thing serves a different buyer or sits at a different price point, but still benefits from the parent's credibility. Middle cost, and the endorsement has to be real rather than decorative.
It stands alone. Correct when the position genuinely conflicts with the parent, which happens most often at the top and bottom of a price ladder, or where the parent's category associations would actively harm it. Most expensive by a distance. Justified far less often than it is used.
The test for which one, and it is a positioning test rather than a marketing one: does this thing strengthen the parent's position, sit neutrally beside it, or contradict it. Strengthen means carry the name. Contradict means separate it. Neutral is where judgement lives and where most of the argument happens.
Two rules I would apply regardless. Decide the architecture before you name the next thing, not after, because names are much harder to remove than to withhold. And schedule a portfolio review, because the default state of any architecture is accretion and nothing in the normal run of business will ever prompt you to kill a name.