Limits on Connections

Jared Walker talks about connections here. He's established a 'floor' for connections, demarcating when it becomes useful to use connections and avoid the spaghetti overlay. What about the ceiling? How often is is reasonable to ask people to make a connection, in terms of a) Connections per trip, and b) distance between connections?

Regarding a): Personally, if a transit trip involves more that one connection, I either find an alternate that involves a greater than half mile walk, or don't make the trip.

Regarding b): The urban* heuristic is: "Is it faster to walk"? is always a good metric. Ie, if the wait time+transit travel time > walk time, walk. Assume urban Salt Lake City conditions (15 minute headway, 8 blocks/mile). Assume 'random' wait times, so that connections have not been timed, any connection imparts a 7.5 minute wait time. At 3 mph walking pace, you can walk 3/8ths of a mile in that time--3 SLC blocks. At an average 12 mph bus speed (time spent at stops included). Thus, a bus will carry you 12 blocks in the time it takes you to walk 3, once you get on it.

This implies that it is worthwhile for the average person to make a connection, if it will save three blocks of walking. But this analysis presumes a time-only comparison, where real connections have other factors to be considered.

*At the suburban scale, it's almost always faster to wait, simply because the distances you are taking transit is measured in miles, not blocks.