Talk to the people affected

The people who live with a river, a road or a plan know things no dataset holds: where the water came up in 2019, which exit people really use, why the bus is late. They are also the ones the decision lands on. Talking to them first makes the model better and makes the result usable, and skipping them is the most common way a good model gets ignored.

Who

• The people who will act on the answer. You met them in the first guide.
• The people the decision affects: residents, businesses, the ones who cannot easily leave or move.
• The people who run the thing being modelled: the operators, the drivers, the field crews.
• The people who hold the data: the utility, the planning office, the agency.

What to ask

• What happened last time? Where, when, how bad? Ask for photos.
• What would you want to know? Their question may be sharper than yours.
• What would you do differently if you knew? This is the decision, from their side.
• What do you already measure or write down? Many a model has been fed by a logbook.
• What should we not get wrong? Each answer is a test the model must pass.

How

Small groups, on site where possible. Walk the place. Keep a record of what was said and by whom, with their permission, and send it back to them to correct. Say what you will do with it and when they will hear from you again.

What it changes

Expect the question to change, the data to grow, and the list of scenarios to become the list people actually argue about. Expect also to find that some of the question is already answered by someone in the room.

The same people, later

They will check the model against what they remember, and they are the users when you put it in front of its users. Keep their contact details and the promise you made.

Next

Find the data.