How long to review requirements documents?

[ccallen]

Unless the requirements were very simple and there wasn't a lot of information on the page (e.g., a lot of whitespace), I don't see how you can do more than a cursory review in three or four minutes per page.

Nothing like being called a bad reviewer based on nothing more than a time statement. I would challenge that assumption.

Perhaps we need to clarify the initial question. Are we looking for the time associated with an initial review or a subsequent review? Are these necessarily different? If they are, how and why? If they are not, why not?

My review was perhaps shorter than you would expect because I have read this document many times over and am very familiar with the requirements I was reviewing. This particular review was focused on: 1) identifying the requirements that are existing functionality 2) finding the duplicates in the requirements. I was not reviewing for missing requirements with this review.

Other characteristics that should have influenced your decision to call it a "cursory review": The 16 pages have 85 requirements and 11 use cases. Does that change how long you think it should take? If so, why?

Maybe we need a time per use case or a time per feature or a time per requirement. But then it would depend on the author. I imagine scenarios when a short document could take HOURS because there are so many gaps and so little clarity but a very long document could take no time at all because it is well done and easy to read.

Based on this short discussion, I would say it's impossible to come up with an average review time per document or per page. It's like asking for an average time to complete a book: depends on the number of pages, the content (technical, fiction, comic book?), the font size (large print books are fast reads), how many pictures, etc.

So maybe the question should be reconsidered: what criteria affect the time it takes to review a requirements document?

[kwiegers]

My review was perhaps shorter than you would expect because I have read this document many times over and am very familiar with the requirements I was reviewing. This particular review was focused on: 1) identifying the requirements that are existing functionality 2) finding the duplicates in the requirements. I was not reviewing for missing requirements with this review.

Of course, familiarity makes a huge difference. Often reviewers aren't very familiar with a work product, so they're going to need to go more slowly to understand it and drill down below the surface to look for problems.

Other characteristics that should have influenced your decision to call it a "cursory review": The 16 pages have 85 requirements and 11 use cases. Does that change how long you think it should take? If so, why?

I think it depends more on how many words are on each page, how well organized the information is, how clearly written they are, how complex they are, and how much risk is associated with leaving a defect undiscovered.

[AlanAJ]

I'm having this debate at the moment. I work on the basis of 1500 words per hour, wich typically equates to 4 or 5 pages an hour. I'm advised that the "recommended rate" in our organization is 22.4 pages per hour. No-one is owning up to determining the recommended rate...

My response suggests applying the Capture-Recapture Method to indicate the likely level of defects remaining following reviews conducted at different rates, as a proxy for the review effectiveness. Does anyone have any other suggestions for objective evaluation?

[achen]

I'm having this debate at the moment. I work on the basis of 1500 words per hour, wich typically equates to 4 or 5 pages an hour. I'm advised that the "recommended rate" in our organization is 22.4 pages per hour. No-one is owning up to determining the recommended rate...

My response suggests applying the Capture-Recapture Method to indicate the likely level of defects remaining following reviews conducted at different rates, as a proxy for the review effectiveness. Does anyone have any other suggestions for objective evaluation?

22.4 pages per hour is ridiculous assuming that no one has purposely lowered the density of information

[MGoyal]

22.4 pages per hour is ridiculous assuming that no one has purposely lowered the density of information

yeah, that doesn't seem realistic. Averaging a page in less than 3 minutes?