Recently I have been working on two different kinds of projects. Ones where I can start from a blank canvas (although I re-use as many concepts and as much code as I can in the circumstances), and ones where I am starting from a pre-existing model. These both have their benefits and drawbacks, but people will always say “I wouldn’t start from here mate!”.

New


Nothing is more frustrating in developing a new model or helping to develop a model which doesn’t deliver all the benefits it could, so here are a few things you can do to make sure any investment in modelling and analysis delivers its critical benefits:


• Decide early in the decision making life-cycle that you want to use simulation – don’t leave it to the last moment to engage internal or external experts
• Build a capability within your business – to understand the results of simulation if using external consultants, or go the whole way and build a capability to do everything from collecting data, to building models, to analysing results.
• Make all the key stakeholders of your business decision aware of the project and its benefits - but get a specific champion, someone who is enthusiastic and can back you up
• Target the pieces of the decision making process that are going to get most value from modelling and simulation
• Know what results you need to see to make a decision
• Re-use as many components of existing simulations as you can

Old


On the other hand old models present their own challenges, in particular with features that people forget about and build again, a second time. Also with features that now no longer serve a purpose (I found an interesting comparison from the world of software development and policy reform here: Software Fences). Old models can get additions that make them now too complex or too cumbersome to use and over time skills in finding appropriate data for them are lost. Here are some actions you can take to get better use out of an old model:
• Re-scope 
- Get rid of features no longer used (keep code in case you want it back and modularise features so they can easily be resurrected)
- Focus the model back on the outputs that now make sense to the decisions being made
• Re-focus - do a project to clean up and standardise the model
• Re-train - Get training in the use of the model(s) and make cumbersome features easier to use