Nike - Missing Sept 13

By:  Erick Nelson
Last Updated:  August 28, 2004


We are announcing that we will not be able to make the Sept 13 target date for Nike Migration.  The primary reason is that interface testing and structured scenario testing were not completed in a timely manner, and we are thus too far behind to be ready on the target date.


Here are straight answers to anticipated questions.

How was the Sept 13 date set?

The original date was July 31, and the software was not complete by that time, and therefore testing had not even started.  When it was clear that this date would be missed,  I was asked to provide the "earliest achievable date possible":  (a) achievable, because the 7/31 had clearly been too aggressive and a realistic date was desired; and (b) earliest, because of the burden of U.S. Customs work-arounds, promises to Nike, and the dependence of the Nike GPO project on this migration.

This new schedule called for 1 additional week of XML mapping (parallel with user testing); 1 week of interface map testing; 2 weeks of interface testing; then 2 weeks of system testing concurrent with Nike UAT.  I also published a schedule that was to be used if it was determined that a less aggressive approach was possible - that was Oct 11. 

What were the key assumptions for the schedule?

1.  XML/AI mapping would be finished, in stages, by Aug 2-6.  It was completed Aug 9-12, about a week late.  Thus, interface testing started one week late.  Explicit - part of the schedule.

2.  Certain key resources would work full-time as the main testers.  The lead tester became unavailable, and the worksheet testing was not done.  Explicit, written and published.

3.  There would be no significant contention with other migration projects.  DuPont, IBM, and Rockport continued in development/testing for the duration; GPO, another Nike initiative, was added and contended for resources.  General multi-tasking also made interface testing more difficult.  Explicit, written and published.

4.  The interface map testing was fairly straightforward and could be achieved in about a week, with the remaining time used for business rule testing and fixes.  Implicit.

What slowed interface testing?

I believe there were multiple factors all contributing in various ways to excessively slow interface testing:  (1) XML errors which needed correction and test re-cycling, (2) contention and multi-tasking, with occasional bottlenecks, (3) mistakes and confusion due to hurrying, fatigue, and stress, (4) lack of proper follow-up on hand-offs, (5) dependence on developers to coordinate their activities rather than over-arching coordination, (6) multiple layers with groups unused to working together (Allogis, AI, CLS/Atlas).  

Where are we now?

Completed since July 31:

Not Completed:

New Requirements:

What do we need to do in order to wrap this up?

1.  Secure a day-to-day business representative, for approving UAT and making business-related decisions immediately as they arise.

2.  Day by Day Plan - Process

2.  Day by Day Plan - Recommendations