Wikipedia claims (although with no citation) that in order to make the space shuttle land, an initial powered delta-v of 322 km/h was applied in orbit, retrograde to the shuttle's orbit. 322 km/h is equal to 89.4 m/s. This caused the orbit to be lowered into the atmosphere, ultimately causing the shuttle to come to a full stop intact on the ground (or that was the general idea; as we know, it didn't work out perfectly every time).
Organic Marble points out the Shuttle Crew Operations Manual, which states that
The deorbit burn usually decreases the vehicle's orbital velocity anywhere from 200 to 550 fps, depending on orbital altitude.
where 200 fps is about 61 m/s, and 550 fps is about 168 m/s. Given this data and the space shuttle's operational range, 90 m/s seems a reasonable figure to use as a median-mission deorbit-burn delta-v.
What I fail to understand is how this relatively small (about 1% velocity change: in the case of Wikipedia's figure, 90 m/s out of the on the order of 7 km/s orbital velocity in a low Earth orbit) could be enough to sufficiently lower the orbit to commit the orbiter to landing, rather than being just a small change in the shuttle's orbit.
Why was such a small delta-v applied under power in orbit sufficient to commit the orbiter to landing?
I expect that good answers will draw on orbital mechanics and atmospheric gas density (aerobraking) to show why the small change was sufficient.