During the last 240 days (from April 16 to December 12, 2008), I pegged the price to visitor statistics. This had a positive impact on revenues.
Every visit counts toward a day's total number of unique visitors, and the current numbers can be seen at shufflesome.com/mint/. Whenever the 60-day moving average was beat by the day's total number of unique visitors, prices dropped. Conversely, prices moved up again, when traffic was lower than the benchmark 60-day moving average. The details I documented from day-to-day on the /play page.
The idea, as explained at /play and in related blog posts, was to let prices be user-driven. The results, briefly, are as follows: On 140 of the 240 days, the stickers were priced with the upper price level. On the remaining 100 days, stickers were available at the lower price level. This pattern emerged in response to how many times visitors beat the benchmark.
On average, there were 213 unique visitors per day, with an all-time high of 771 unique visitors per day. The majority of all orders during the period of 240 days were made on days with the lower price level (51%). Since there were 100 days with the lower price level, it follows that there were 44% more orders on days with the lower price level, compared to days with the upper price level.
Without going into statistical significance, the difference between the two price levels seems to have had an impact on customers' decisions to buy. Even though my implementation of the idea wasn't technologically as good as I would like to have done it, it was still sufficiently effective to encourage customers who cared to notice the offer. Without it, revenues would have been slightly lower.