we’re currently looking into using DataSift as the source of data for our apps and have run a couple of test during the day.
One of the most important ones was collecting mentions for a relatively high-profile account like @techcrunch. I created a very simple stream with only one filter, namely:
twitter.mentions in “techcrunch”
…and I left it to run for approx. 1.5 hrs. It only caught 28 interactions which struck me as odd, so I went to compare Twitter’s search results with DataSift’s export. Plenty of tweets which were showing up in Twitter’s search weren’t caught by the stream. I can provide the exported JSON and the screenshot of the Twitter search results page for the relevant period if needed - I don’t see an attachment option here on the Discussion forum, but I can send it by email.
Have I missed something (i.e. is there a subtle CSDL quirk I’m not aware of, which could cause the issue), or is the stream simply incomplete?