That's concerning! I'd really appreciate if you could collect a set of specific examples and report at http://duckduckgo.com/feedback.html

That's how we improve fastest. But generally, the way we get results shouldn't result in a lot of stuff actually missing, which may speak to some bugs in there.

I've been noticing the same pattern -- the results from DDG aren't quite complete enough. Sometimes, it's just missing a few minor things, other times, I'll run a search query and there will be a major search result that I know is missing, that I can get in one of the top 5 results from Google.

Which sucks, because I hands-down full-on no-buts prefer DDG's interface to Google.

I hope I'm not being uncool here, but a while back I privately suggested mining the data from "!g" queries as a way to see where DDG wasn't returning the results that people expected. Is there a reason you can't do that? In my case at least, I usually start with DDG, and if I don't get what I'm expecting, then I use a "!g" -- so you'd have a nice little database of results to perform comparisons from.

Since we can't see vote counts anymore, I just want to chime in and say I agree. As a previous DDG user, I feel much the same way. And mining !g sounds like a great idea, though there might be some privacy issues to deal with (I'm imagining a dedicated DDG user who wants to Google his name).

I feel the same way. I keep trying to switch back to DDG and then run into results that just don't have what I want (and Google does).

I'm going to switch again and start sending DDG results that aren't helpful.

Thank you!

Actually, something else that bothers me, but I'm not sure there's much you can do about it. I really like the Chrome location-bar search completion thing. Most of the time, when I start searching something, I can see if my search terms are good by seeing whether it's been searched before. If it comes up with nothing, I know my search terms are probably not well-worded.

Not sure how you could provide for that, though.

Thx, duly noted. We are working on a suggestions feature though not sure it will address this particular issue.

I used DDG for about a month as my main search engine, I love the bang syntax and lots of other goodies, but for me down in Brazil DDG just responds 10-15x slower than Google -- so my commitment is that I will keep coming back to test for latency improvements until I can switch over completely.

Same here, but in Canada. DDG response time is a major showstopper. From a brief examination the issue appeared to be rooted in DNS response time, which is something that I wasn't able to fix by trying different DNS servers. So, regretfully I had to switch back to Google.

That's weird (on DNS). We have been having an issue with in Canada where one of the major ISPs is dropping packets on route. If you'd email us I'd love to investigate this further.

I don't think that's strongly geography-related. I see the same thing from California with a very fast connection.

If you search for some things (try "stack overflow" or "red bull"), you seem to get a near-instant results for the thing judged most likely/relevant in a red box at the top of the page, then the "ordinary" results are loaded separately, I think in a separate javascript call.

They seem to have a some split architecture that keeps a few bits spooled up for quick retrieval, and say to heck with the rest. My first instinct would be that they're RAM-poor, and also wonder if, not being able to cache sufficiently in RAM, they're also being hamstrung by I/O performance on EC2.