One potential action is to do nothing, be happy, and concentrate on producing more content. I'm not being facetious in saying this but, to open that old chestnut again (or recent chestnut in the case of my discussions with EFV), I genuinely believe it may sometimes (but not always) be misleading to worry about a "declining eCPM".
For example, suppose the world is completely static, advertisers, bids, smartprice, nos of visitors, etc.. However, the one thing that changes is your ISP upgrades their server to be much faster. As a result, by definition, traffic goes up and eCPM goes down.
This is because, in the "slow" era, a lot of people gave up waiting for your page to load. And those that were patient exited your site, perhaps via an Adsense ad, never to return. In the new "fast" era, however, the pages load much more quickly, so you get more people seeing your pages, and they are more likely to stick around your site before exiting, or return to it because it is a nice, quick site.
The earnings on our site this month are looking as if they will be roughly 5 times what they were in the first month (Dec 04). But eCPM is running about 10% lower. One reason (amongst others) for this is that we have speeded up our site and improved site navigation. As a result, the average number of pageviews per visitor has increased as they explore the site. This increases 'traffic' as measured by pageviews, which by definition depresses eCPM, but it also increases earnings because it is easier for them to find more relevant ads and they come back to the site more often.
So, the first action I suggest when observing a declining eCPM is to ask the question whether the decline is an artifact, a statistical quirk that results from a change that is actually increasing your earnings, which is where Earnings Per Visitor can help.
Mentat's post suggests that his problem is not an artifact - less income with more visitors. In that situation, we would look in detail at our channel statistics to see where any variation in earnings is coming from. Mentat, I hope this may be helpful: we use a recent 'good month' as a benchmark and make a detailed comparison of channel performance; often any decline is due to a significant change to one or two particularly important channels, not a site-wide problem.
The next question to ask is whether this is a temporary change or a long term trend. If it is a short term decline, then unless there is something obvious about the ads on the website, it may be best to leave it a while and see what happens. As a rule (with some exceptions, perhaps), the Adsense robot will try and maximise your earnings, but it can take a couple of weeks for it to do so after a change.
Even if it is a long term change, there may be nothing one can do about it. It may be a decline in bidding rates, because of a shift in the demand/supply equation (could this be happening to you, Mentat?). I'm afraid this type of change is just a tough fact of life and can only be counteracted by adding more content.
Only then, if there seems to be a long term decline not explained by other factors, is it worth trying some of the other measures. But, as I've said before, trying to manipulate smartprice is like trying to untangle a plate of spaghetti with a blindfold and gardening gloves on. It's not straigthforward.