In thinking about the difficulties of cross-boundary leadership and analytics -- for emergency response planning in New Orleans, for example -- I'm reminded of an "aha!" from many years ago.

Just out of graduate school, I was working as an assistant to Mayor John V. Lindsay of New York City. I'd been assigned as liaison to a consulting project with McKinsey & Company. The McKinsey team was analyzing refuse collection and disposal and had developed a series of recommendations based essentially on industry best practices.

As they prepared to present their results to city officials, the McKinsey manager asked the team to fill out a simple grid (PCs and spreadsheet software were not yet invented). The rows represented the recommendations, with columns for each participant invited to the meeting. Into each cell the team had to produce a number estimating the degree of support expected from that participant for that recommendation, along with the basis for their estimate.

This tool made the team probe deeper than "best practices" research. It forced them to explicitly include the problems of negotiating among the "cross-boundary" interests at the meeting (representatives of the Bureau of Cleaning and Collection, the Bureau of Waste Disposal, the Bureau of Vehicle Maintenance, the Mayor's Budget Bureau, etc.). It made them collect a new category of information, and also influenced the design of the presentation agenda. For example, for meetings expecting ready support, the strategy was to jump quickly to recommendations in order to spend time locking in the details of implementation. On the other hand, for meetings where the spreadsheet analysis highlighted serious opposition, other tactics dominated such as pre-meetings to seek support from those in authority, and an agenda that proceeded more slowly to carefully lay out the problem before building the incremental logic needed to maintain agreement so far as possible as the recommendations were rolled out.

My own graduate work had exposed me to statistical and quantitative modeling tools. But I had never been introduced to an analytic spreadsheet that assumed that the core problem was implementation, and that the core analysis was negotiation analysis.

Aha! ... Pretty neat. Unfortunately, almost 40 years later, this kind of analysis is still too often missing. I wonder if the pre-Katrina planners could have made more progress had they defined their problem this way.