In trying to decide whether to use DeviceAtlas or WURLF (or both), I took a sample of our mobile traffic (which includes inevitable robots and desktop visits) and distilled 1,500 unique user agents. I then ran a comparison test using the DA API and WURLF's API, in Java (it may be worth noting that this was testing solely for resolutionWidth, which is one of our most important info requirements) and got the following results for successful matches, i.e. where resolutionWidth > 0 :
DeviceAtlas: Processed in: < 1sec, success rate: 76%
WURFL Strict (no fall-back): Processed in: < 1sec, success rate: 27%
WURFL Loose (does fall-back): Processed in: 96 seconds, success rate: 96%
So, DA compares very well with WURLF in strictly matching incoming UAs, but not that well against WURLF using its "loose" query, i.e. with a fall-back if an exact match is not found.
Given that this is influencing our purchasing decision, could you suggest whether a) there is some (undocumented?) way of getting DA to fall-back to the next-best device, or b) whether you have plans to add some sort of fall-back/fuzzy-match facility?
Cheers.
