Most folks believe more time on site is better -- it means customers are viewing more products, reading descriptions and reviews word-for-word and "engaging" with your content.
But time on site is a poor indicator that your site is doing well. Perhaps you do offer fantastic product information, videos, product guides, customer reviews etc. and visitors love the content you provide - it helps them decide what they want to buy offline, from someone else. Or, customers comparison shop and leave your site open in its browser along with 6 other tabs.
Maybe your site is so difficult to navigate and find product, a lower time on site would actually indicate better customer experience and smoother checkout.
Don't forget that average time on site is just an average. It can be segmented out by visitor type, referring site/campaign or by keyword, for example.
You would expect time on site to be lower when a visitor lands on a specific product page (rather than your home page) from a shopping engine (having compared the product and price against competing offers) or by a specific keyword (like "Marvin the Martian basketball"). If a customer doesn't need to search or browse to find a product they want, shouldn't a short time on site be better?
Not necessarily -- a short time on site could indicate the customer never bothered to complete the purchase, or bailed when she hit your checkout page with required registration. It could also mean customers are ignoring your cross-sell suggestions and don't care about your free-shipping-over-$100 offer.
And don't forget, time on site cannot account for the visitors who view only one page. Unless a visitor loads a second page on your site, your analytics program will not calculate the time on the first page. That's why you find many pages have an average time of 0.00.
Though time on site is usually explained in any Web Analytics 101 material, I believe time on site is a broken metric that is not important for online retailers. Longer or shorter time on site does not indicate site effectiveness.