This is a sample order_syncer which syncs externals and needs the output of what is being synced in the next method.

order_syncer.rb

class OrderSyncer
  def initialize(order)
    @order = order
  end

  def synchronise_all
    external_order_id = synchronise_order(@order)
    external_payment_id = synchronise_payment(external_order_id)
    external_contract_id = synchronise_contract(external_payment_id)
    synchronise_authorisisation(external_contract_id)
  end

  private

  def synchronise_order(order)
    # If the fetcher does not return an order_id we need to synchronise
    unless order_id = Fetcher::OrderId.new(order.id).fetch
      order_id = Synchroniser::Order.new(order).synchronise
    end
  end

  def synchronise_payment(external_order_id)
    #Same kind of steps as in synchronise order
  end

  def synchronise_contract(external_payment_id)
    #Same kind of steps as in synchronise order
  end

  def synchronise_authorisisation(external_contract_id)
    #last step which always synchronises and returns an id or nil
  end

end

The problems with this setup (imo):

1. The synchronise_all is getting really big for 12 steps.
2. The separate synchronise methods depend on each other being called in succession. (Don't know if I see this as a problem or just something that I should accept).
3. The OrderSynchroniser as it stands is 150 lines. Not a problem per se but thinking about changing behaviour in this class is a pain.

Solutions I thought of but fear will overcomplicate the actual code:

1. Extract the separate synchronise methods to their own classes. So the synchronise_order method will be extract to the class SynchroniseOrder which will Fetch OR Synchronise the order. Although this will make the OrderSynchroniser class smaller now I have 3 separate classes that are being called.
2. Since the synchronise_methods are all alike I thought about creating a more generic method which will be called with the classes it needs to call, making the setup some what more abstract and generic. I'm not sure this would be a step in the right direction since I'm afraid that it will make the code to abstract and hard to change in the future.