[paraschopra]

Plus - GWO won't let you play around with the data. In the end, what you would just know is that one variation performed better than other one. You wouldn't know if one variation resulted in more time spent, less advert clicks, more user interaction (in terms of comments), more purchases. And you won't know how it affected your overall funnel behavior. GWO would give you just one goal tracking, that too a simple convert/no convert.

In short, as Patrick pointed, GWO is developed from a perspective of ease-of-use and simplicity. You cannot do advanced stuff with it.

Even though I agree with Patrick with the impact of load time with site performance, the case against using JavaScript must be seen from a broader perspective. Google, Amazon and other companies stress for millisecond-level optimization because their site time is already hyper optimized. But if you see an average e-commerce (or some other) website, a couple of 100 milliseconds due to JavaScript look like a good bargain when it comes to keeping view layer separate from code layer.

But, overall, I think it should be seen from business objectives. If one cares about milliseconds, best to do testing at code level. If you care more about doing tests quickly and not touching your code every now and then, JavaScript based testing is the way to go.