If you want to let the world know about a project of yours, it is clearly not sufficient to put up a web page and wait for people (and Google) to discover it. However, when it comes to announcements it wasn't too obvious to me which strategies would be most effective. Is it worth to spend time on a press release, or is a post to a focused mailing list as potent? How do traditional journal releases compare to social media coverage when it comes to special-interest projects like this one.
As there can only be one initial announcement, the following data needs to be taken with a grain of salt. However, I made an attempt to spread announcements made through different channels over time in order to get at least some idea of what is most effective.
Three weeks of announcements
Here is the plot that summarizes the effect of various announcements made over the first three weeks. Each significant event is marked on the time axis, and I'll make a few notes about each of them in the following.
On May 27, Nature publishing group launched their new open-access journal Scientific Data with our data descriptor in its inaugural issue. This meant there were only two-clicks or less from nature.com to the article that contained the link to studyforrest.org. What sounds nice so far, turned out to be more or less irrelevant regarding the observed website traffic. Until today only 1.4% of all visitors are coming from any page hosted by nature.com. And that is despite the fact that our article is the most accessed one in that issue. I suspect that a link to the web page in the abstract would have been more effective.
On the day of the launch, we also issued a press release in English and German. This press release was managed by the public relations office of the German Bernstein Network (BCOS), who coordinated with all involved institutions. A BIG thanks to them for making this happen! Unlike the link from nature.com this was a big success: On the day after the release, the biggest European tech news site picked it up (in German). Until today, this post is the origin of almost 20% of the website traffic.
Next up were announcements to more focused news channels that I consume myself. First, the global computational neuroscience mailing list comp-neuro on June 2, and soon afterwards, on June 4, the FSL mailing list. The original plan was to see how much traffic these mailing lists would generate, but we fell victim to a fortunate "accident": @neuroskeptic tweeted about the real-life cognition contest on Twitter. Although, it is not possible to tell what the effect of the mailing posts alone would have been, the combined traffic is very similar to the one from the more conventional media outlets. All data taken together, social media (incl. Twitter) is the origin of about 17% of all traffic to studyforrest.org.
A few days later, I presented dataset and contest at the OHBM hackathon, and, subsequently with a poster, at the OHBM conference itself. While a hackathon is probably the optimal environment for engaging people, a conference poster is often similar to random sampling of people with tangential interest strolling through the poster hall. On the other end of the attention spectrum is a conference keynote. We were extremely lucky that Jim Haxby gave a talk on "A common high-dimensional linear model of representational spaces in human cortex" that was closely connected to natural stimulation paradigms in neuroimaging research, and that he included the contest announcement as the last slide in his talk (Thanks!).
Despite a clear increase in website traffic it wasn't even half of the magnitude of the previous announcements. However, people attending OHBM and seeing the poster or the keynote are probably easiest to get interested in the contest, and may turn out to be the most frequent participants.
The last announcement targeted the SPM mailing list. This is one of the longest running and most read communication channels for neuroimaging research. However, the response to the announcement on Jun 16 yielded a traffic increase that was significantly less than the one in response of the post to comp-neuro -- despite identical wording of the post. I suspect the reason for that is saturation. At this point, we had probably reached most interested people in neuroimaging research. Google analytics counted about 1800 unique visitors at that point.
Based on these observations, I believe that you can get your word out to many people using mailing lists and social media alone. However, I saw no evidence that you can reach anyone outside your own bubble without conventional, more general purpose media channels.