Google announced time-based bidding as a new feature, which we consider a good thing for savvy advertisers.
For those clients for which it makes sense, we've been adjusting bids on time-of-day, day-of-week, and season trends explicitly for a long time now. The RKG bid platform uses a dynamic schedule for updating bids -- ads for which it matters get bid very frequently; ads for which it matters less get bid less frequently.
We don't use the Google feature for this -- our optimization platform does it explicitly, automatically computing the right time to raise or lower bids on ads for which it matters, then communicating with Google (and Yahoo and MSN et. al, for ads on the other engines) at the right times to instantiate those bid changes through the APIs.
As more folks experiment with time-of-day effects, it is important not to confuse time-of-click with time-of-order.
That is, to determine the efficiency of clicks on a certain ad between (for example) midnight and 1a Sunday night in your timezone, don't compare the sales in that time window with the ad cost in that window.
Because there's a time lag between paid click and the resulting order (with larger orders coming in slower), to compute (say) the efficiency of the 12-to-1a time block, you need to compare the cost of the advertising bought in that block with the resulting sales from that advertising, even if those sales occurred in later hours. That is, you need to match cost to sales at the ad level by user (via cookies), and not just by aggregate time bins, comparing what you bought (paid clicks) with the resulting benefit (sales from those particular clicks) .
A seemingly small point, but really important.
Check out our study on click-to-order interval from last summer's SES show for more info on this.