First off, I tend to be of the opinion that you can't negotiate with a terrorist group once it's taken that final step. Most terrorist groups, if they last past their initial stages, evolve (or devolve, depending on how you look at it) into rather amorphous entities motivated by the cycle of violence and revenge more than any measurable (or attainable) goals. They may pay lip service to goals, but they are generally so fuzzy and indistinct that they're window dressing by that point. And once a group crosses the line into terrorism of that type, you can only eliminate them.

Punitive expeditions can work, provided you have a good understanding of the actual centers of gravity of your targets. Most of the actions in the Indian Wars could actually be classed as punitive expeditions, and they were successful (or not) based on the understanding particular commanders had of their opponents. Mackenzie's efforts during the Red River War (culminating with the Battle of Palo Duro Canyon) were successful because he understood that the key to breaking the power of the Comanche and Kiowa on the Plains could be found in their logistical systems (actual camp supplies and horse herds). If you took out those systems while minimizing actual casualties (he also seemed to understand the revenge requirement in most Plains cultures that could drive them into conflict) you could force the tribes onto reservations. Custer, on the other hand, never seemed to understand that. Crook also failed in that regard, although he was successful against other tribal groups with different environmental considerations. We have a mixed history with these kind of expeditions (Pershing's foray into Mexico is just one example), and I think it's mostly because we don't develop commanders who understand the situation they're going into and fail to adapt. We have perhaps been most successful when military force is used as an adjunct to State Department goals (the so-called Banana Wars), although even then our longer-term legacy is very uneven.

Metrics in my view are a convenient crutch for this lack of understanding. It's easier to construct a fancy Excel spreadsheet than it is to identify points of vulnerability in an organization that doesn't look like a fielded army. We saw this during the development of air target lists during the Vietnam War. Air Force planners looked at maps of North Vietnam and for some reason saw WW2 Germany and listed their targets as if NVN was a fully industrialized society with the same needs and dependencies as a European opponent. So the war became all about sorties flown and tons dropped and less about the actual value of the targets.

What we have likely done with AQ is a larger-scale version of what happened with the IRA and the PLO. We've caused it to multiply and divide, with smaller pieces motivated more by revenge and general bloodlust than actual measurable goals and objectives. Those pieces, IMO, can only be killed or otherwise neutralized, and even then it's a very long haul (David can speak better than I regarding the issues in NI, but even there it continues to rumble on at a very low simmer...peace talks and settlements aside).