I have been and remain a great fan of ITA. However, as a free service it cannot afford to spend unlimited computer time on a query and must cut it off. As a result you might miss the proverbial needle in the haystack. I had an illustration of this today. (A) The following search gave $1807 as the lowest price ORD:: AA (to LGA, 4/29) JFK:: AA NRT AA (to BKK, 5/2) BKK:: AA NRT AA (to ORD, 5/6). (B) However, specifying exact flights got down to $1397 ORD:: AA398 (to LGA, 4/29) JFK:: AA167 NRT AA5834 (to BKK, 5/2) BKK:: AA5835 NRT AA154 (to ORD, 5/6). The difference is better than any FD most of us ever dream of. Of course, B is no "search", it is pricing a specific itinerary. Search A is about as economical as it gets without specifying actual flights: we are specifying one airline (no aa,jl,cx), exact dates (no date range), a specific connecting airport. Still it misses the low combination B. Why? We get some idea by considering - (C) The following search gave $1397 as well ORD:: AA398 (to LGA, 4/29) JFK:: AA NRT AA (to BKK, 5/2) BKK:: AA NRT AA (to ORD, 5/6). So, one possible culprit is the number of ORD-LGA flights: the number of possible combinations, which involve not only different flights but also fares available in those flights, could have overwhelmed the search. I am not sure, even surprised, but it does seem like a reasonable guess. A general lesson is to be cautious of "powerful features" (in this case multi-destination search). Without matching computer time, a powerful script is like a great business plan with no budget. You obviously can't begin by guessing the right flights, else why would you be using ITA in the first place? However, you must build and narrow down your search carefully and slowly. Tie down whatever you can as you go along. The art and science here are not unlike those involved in gold- or oil-exploration: you have to scan broadly, analyze the results, select a narrow area to dig deeply, analyze the results, and keep repeating.