Spec QC, Extreme Inspection
The New Inspection and Review Process.
Extreme Inspection, a simple and powerful version of inspection (Specification Quality Control – SQC) that your Company can install immediately at low cost.
Extreme Inspection Outcome The outcome of this type of inspection is to give a fair measure of Major defect density.
Intents of Outcome: The intents of the Major defect density measure are: Clean: to make sure that polluted specifications do not enter the next working processes. Learn: to motivate specification writers to learn and follow Company best practice specification rules.
1. Inspection Outcome Justification
The outcome of this variation on conventional Inspection processes is to determine ‘specification exit’ by measuring and estimating Major defect density. The outcome is NOT (as with conventional inspection) to ‘clean up’ bad work.
The result of this outcome limitation is that many of the time honored conventions of Inspections (as in Gilb & Graham: Software Inspection) are NOT necessary or desirable. We only need to do whatever gives a reasonable measure of defect density. We only need to focus on determining that the specification is exit-able or NOT .
In simple terms, if we find (if the checker detects) one or more Majors in a page, it is NOT exit-able, because the real estimated quantity of majors actually there, exceeds the Exit limit of ‘one per page’. If we find less than one major defect on 4 pages, it probably is economic to exit the spec.
Economic is the key word. We are trying to determine if it pays off to exit now, or to rewrite the spec to a cleaner level now.
All costs for the spec writer, the spec checker and a possible process guide, will be charged to the project the writer is working on, and to the QC process costs specifically.
Rationale: so we can track the true costs of doing this and the degree to which it is done.
3. Auditing this process:
The Inspection (Spec QC) process must be regularly (monthly) audited to make sure it is really conducted according to intent and is not corrupted or misunderstood.
This includes double checks on audits to see if the conclusions of the check and the audit are reasonably consistent. Frequent audits are necessary in the beginning and with newcomers. Auditing will be done by the process owners.
4. Process Improvement
The process needs to be continuously updated (mainly in the 'tool kit' which defines and supports it: the checklists, the process definitions, the computer data collection support) by the official process owner.
5. Process Ownership
There must be an official process owner to champion (and to manage ‘local’ champions), spread, audit, and improve the process, as experience and insight dictates. This can be a group.
6. Process Sponsorship
The executive sponsor of this process should be official and visible.
The checker shall never reveal the numeric result of an Inspection to anyone else except the writer. The writer may reveal the results if they want to, but they are not obliged to do so even to their direct manager (who should not even ask!). The results of an inspection, as recorded in the Specification Quality Control Database, are never to be released, revealed or reported with the name of the writer or information (such as document ID) that can lead to their identification.
Rationale: to prevent fear of defamation leading to false reporting of results. To emphasize that the process is there to help the writer reach the corporate quality level required. It is not, in any way or time, to be used for personal job performance evaluation. Evaluation should be based on EXITED specifications, and their timeliness only. Managers need to be informed and reminded of this cultural paradigm by the process owners.
8. Expected Effectiveness
We expect that the Major defect finding effectiveness of the checking process will be in the range of 10% to 35% of the actual real Majors present in a specification. This if consistent and known level, is quite sufficient to estimate the actual total number of majors actually present. We can then estimate with sufficient accuracy (say ±20%), to determine levels of Majors in entire spec; and also in the spec after correction of listed defects (listed in their notes by checkers) defects.
9. True Measure of Inspection Progress.
The correct and relevant measure of how effective the Inspection process is working, is NOT as many would assume the quantity of Major defects found and fixed by an Inspection. In fact we strongly recommend that this measure is well hidden from public view! (It has its uses!).
The true measure is the average level of Major defects/Page which we can consistently release. We need to move from about 100 Majors/Page down towards about less than one per page. This cannot be achieved by finding and fixing defects (because we cannot find a large percentage at all)! It can only be achieved in practice by motivating writers to reduce defects actually injected in their work, from a typical 100, and move them down towards one maximum injected/page. This is the ‘individual defect injection learning rate’. Individuals seem capable of reducing their own defect injection by about half (50% fewer defects injected for each cycle of learning (write, inspect and rewrite).
The measure of real progress is the released defect density, and it is this measure that will most closely correlate with later statistics on quality and productivity of projects.
 So we do not need to get maximum effectiveness by having a large team or by using one hour per page or by looking at all pages (we can sample in 10-40 minutes and use one or 2 people).
The outcome of this type of inspection is to give a fair measure of Major defect density.
Intent of Outcome
The intents of the Major defect density measure are:
Clean: to make sure that polluted specifications do not enter the next working processes.
Learn: to motivate specification writers to learn and follow Company best practice specification rules.
Internal Extreme Inspection Goals
“The expected effects of rigorously carrying out this process are:”
Scale: Estimated remaining Major defect density per logical page (300 Non Commentary words)
Past [Jan. this year] 50-100 Majors/Page <- Multiple sample inspections
Goal [Jan. next year] less than 10 Majors/Page
Goal [Jan. two years from this year, or sooner if feasible!] less than 1 Major/Page
External Extreme Inspection Goals
Scale: Total project time to successfully complete a project
Past [Jan. this year] ???
Goal [Jan. next year] = 70% of Past [Jan. this year]
Goal [Jan. two years from this year] = 50% of Past [Jan. this year]
Comment: This will be accomplished by less back and forth, and reviewing of requirement documents, and by shortend coding and test times, and by less effort when work is contracted out of country or to sub-suppliers. More time at the requirement stage is expected.