The proper way to run Ruby migrations is with the change definition. This allows rollbacks to remove any column or change that is run inside that block and reduces the need to have an up and down method to reverse the changes. But, sometimes you need to run some updates on a new table or column that is added to the database. In order, to do this you have to only run that sql on the up state when the column exists.

Example:

class AddCountToThingsTable < ActiveRecord::Migration
  def change
    change_table :things do |t|
      t.integer :items_count, default: 0
    end

    reversible do |dir|
      dir.up { data }
    end
  end

  def data
    execute <<-SQL.squish
        UPDATE things
           SET items_count = (SELECT count(1)
                                   FROM items
                                  WHERE items.thing_id = things.id)
    SQL
  end
end

In the above example, the reversible block is used within change to run a reversible change to the database. The up is provided a closure which will run the data method to update the column that is added. This will not be run with down, since there was no closure provided with it.

Hope this helps, let me know some other techniques you can think of.