When a business case unravels six months after approval, the instinct is to blame events. The market moved, a vendor failed, priorities shifted. Sometimes that is genuinely it. More often, the seed of the failure was sitting in the case on the day it was signed, and the approval process was not built to catch it.
The first culprit is the single-point estimate. A case that shows one cost, one date and one benefit number is hiding its own uncertainty. Real delivery has ranges. When a case presents the most optimistic point in each range as the plan, it is not wrong by accident. It is wrong by design, because the optimistic version is the one that gets approved.
The second is benefits that no one will own. Costs are concrete and land on a specific budget. Benefits are often spread across the organisation in a way that means nobody is personally accountable for delivering them. A case approved on benefits that have no owner will deliver the costs and quietly lose the benefits, and the gap only becomes visible long after the decision.
The third is the unexamined dependency. Almost every case rests on something outside its own control. Another programme finishing, a platform being ready, a regulatory position holding. If those dependencies are listed but never validated with the people who own them, the case is resting on assumptions dressed up as facts.
What I look for, before approval rather than after, is whether the case can survive its own worst honest version. Show me the downside numbers, the benefits with a named owner, and the dependencies that have actually been confirmed. A case that holds up under that scrutiny is rare, and worth backing. A case that only works in its best version is a problem you are choosing to discover later.
The uncomfortable truth is that the approval moment is where most of this is fixable, and where there is least appetite to look. Everyone in the room wants the answer to be yes. The job of good governance is to make sure yes is earned.