For years we have implemented the concepts of sense and respond, in the context of a learning organization. For new web services, we reduced the time of market trials from months to weeks. We dramatically lowered the actual costs of experiments to the point of immateriality. And we improved forecasting accuracy.
In short, we were better, cheaper, and faster.
But we were still a bit too slow. By stringing together waves of testing, it might take many months to extract major learnings. And being slow is very expensive.
We've recently begun to solve this problem through massive parallelization of sense and respond techniques. Rather than waiting for the results of a wave of testing, we continually launch new, overlapping waves. While we are unable to incorporate learning from one wave into the next sequentially, we have multiplied the speed of individual insights. We then reconcile and apply the learnings "on-the-fly".
Our new process is less scientific and a little more costly. But it is faster, creates far more value, and leads to higher quality. It is consistent with Kaizen, helping us to better meet the challenges of the acceleratron.
Update: There seems to be confusion about how we utilize this approach. As I wrote, we use it to test new features on our websites before we formally introduce them. We do not use this approach to test software applications, such as vBulletin.