Review Strategy

During software development it is possible to perform technical reviews for every piece of software, whether it is a document (e.g. specification or design) or any program unit. Thus, we have to decide how many reviews should be conducted for a particular project. Before you can answer this question and determine a reasonable strategy for your project, you have to make observations about the effort and benefit of reviews.
After this, some criteria are given to determine those results for which a review should be performed.

Effort for Technical Reviews

Documents
(except code)

Code
Maximum Document/Code Size for one Review 50 pages 20 pages
Number of Reviewers
Review Preparation (relative)
Effort Review Preparation
Effort Review Session (absolute)
Total Review Effort
5
10 pages/hrs.
25 hrs.
14 hrs.
5 person days
3
5 pages/hrs.
12 hrs.
10 hrs.
3 person days
Effort for Document/Code Creation (relative)
Effort for Document/Code Creation (absolute)
2 pages/day
25 person days
1 page/day
20 person days
Review Effort vs. Effort for Document/Code Creation 20 % 15 %


Benefit of Technical Reviews

Errors Found/Document (relative)

60-70 %
Reduction of Error Costs 75 %
Total Savings/Complete Development 14-25 %
Netto Improvement/Complete Development 20 %
Netto Improvement/Maintenance 30 %


Reviews Determination Checklist

The following criteria may help you to decide for which results a review should be performed.

Reviews should be performed for

  1. results that are visible to the users or purchaser. At least, this includes the requirements document, acceptance test plan, and user documentation.

  2. results that define essential interfaces, e.g. system design.

  3. results that are afflicted with high risk, e.g. complex parts, user critical functionality.

  4. central parts, used by several components.

  5. time and memory critical parts of the software.

Altogether about 50 % of all results should be reviewed. Thus, about 10 % of the total development time will be spent on reviews. But

The smaller the project, the more results should be reviewed!