With your “top down is best” you state very strongly an absolute position to a question that I believe has very context-dependent answers. I was trying to uncover the kinds of assumptions that would need to be tested before choosing the top-down approach or any of its numerous possible alternatives.

To the 3 assumptions I wrote down:

No, I do not believe that two people will necessarily come to the equivalent decompositions (#1). I’d go as far as saying that this is closer to design than requirements capture than some are prepared to admit. This is not to say that it cannot work, but it certainly speaks in favour of collecting multiple perspectives (e.g. bottom-up, middle-out and other approaches).

I do not believe that the traditional project approach with tightly-defined (let alone complete & accurate) scope, cost and duration (#2) is always the most appropriate approach. Other well-known approaches can be effective in the right circumstances. In other words, context matters (#3). A lot.

I don’t agree with your assumptions. How can a product breakdown be knowable and repeatable? If it were there would be no need for the project. Have I misunderstood you?

Nonetheless, I agree that multiple points of view are essential. This is not unique to the bottom-up approach. That is, unless you incorrectly interpret what I say to mean management style.

I write about the early stages of project planning and specifically the preparation of the project plan. At this time it is usual to plan for the entire project in outline and the next stage in detail. I concede that detailed planning will often benefit from a bottom-up approach. For instance, identifying all the different invoice types would be beneficial when building a system to handle payments.

However, your point about the impact of missing products is very pertinent. It is these omissions that often challenge project costs and delivery schedules.

Thanks for your feedback.

This position rather assumes that:

1) Breakdowns are both knowable and repeatable (two people would arrive at equivalent result)
2) Completeness and accuracy are overriding concern, always
3) Context is never important

If #1 is true then you are home & dry, easy! Often it is not, and other approaches are called for. Bottom-up done with the participation of multiple points of view speeds the discovery of both requirements and risks. And many problems call for styles that are more evolutionary or exploratory.