Bulletproof Problem Solving - Conn and MacLean

1. Take the time upfront to really understand your problem. Often the initial problem that is posed isn't the right question. Before you conduct analysis, probe the problem carefully with the decision makers. Know the boundaries and test them to maximize creativity. Required accuracy, time available, other forces acting on the problem. Use one day answers to keep refining your problem statement iteratively.
2. Get started with nothing more than the problem statement. Don't wait for the giant dataset or computer model. Just get a large whiteboard and begin sketching the logic tree. When you start looking at the data, then your thinking will evolve from simple component structure to more elaborate hypotheses. You are thinking of what needs to be true in order to support your hypothesis
3. Cleave the problem in different ways. Solve the problem backwards. Herbert Simon : Solving a problem simply means representing it so as to make the solution transparent. Exhaust possibilities.
4. Use a team whenever you can. Diversity in thinking and deep domain experiences are what you are after. Also you want to ward off cognitive biases. You have to watch out for authority bias. Team voting, red/blue team competition, mock trials are your tools.
5. Invest in a good work plan. It takes time, but will save wasted effort. Prune your tree by focusing on the big leaves of impact (use post-it notes on the whiteboard) that you can move. Remember : strong hypotheses are easier to challenge and pressure-test. Be very precise about what a particular output should look like - what hypothesis it questions, who will do it and by when. Use chunky plans that run only 2-4 weeks out so you don't overrun your initial thinking and lean/GANTT study plans to keep your work on track.
6. Start analysis with summary statistics (average, etc), heuristics and rules of thumb to get a feel for the data and solution space. Before you call in the big guns, explore the data - learn its quality and understand the magnitudes and directions of key relations and assess whether you are trying to understand drivers to plan an intervention or predict a state of the world. Big guns have their place, but one day answers, supported by heuristics and good logic are often sufficient to close the book on many problems to move on to the bigger ones.
7. Call in the big guns when needed - school bus routing across a major city, detecting disease in medical images or optimizing production facilities for a major company does require advanced analytics. You can outsource the big-gun analysis if needed. Game theory, machine learning, risk management, strategy staircases, long term theories of change - are all available if needed. A statistics of operations research class makes these accessible through intuitive software.
8. Make synthesis and story-telling as important as the analysis. It's tempting to quit after the analysis phase - to think the problem is solved. You need to be persuasive to convince the stakeholders. Humans are visual learners and love storytelling.
9. Treat the 7-steps process like an accordion. It's an iterative process. Expand and compress steps based on the problem you are solving.
10. Don't be intimidated by any problem you face. If you invest the time in mastering 7-steps, you will be confident in tackling even the wicked problems. You will find insight into any problem of consequence.