70% – 80% of data projects fail because they don’t start with a question or hypothesis – they start with data.

Here’s a scenario you’re probably familiar with

You see an issue and ask your team to “do some digging” to better understand what’s driving it. They come back with data, you ask questions, they go back and do some more digging. You can repeat the cycle until

  1. you get fed up
  2. the team get too frustrated
  3. you run out of money to bribe your team in chocolate

All jokes aside, the challenge here is that you will never find what you’ve been looking for due to one simple reason: you haven’t defined what it is you are looking to find. Or in analytics terms, you have forgotten to do the most basic of tasks: to define a question or hypothesis you want to test.

6 Simple steps

Senior executives often ask broad and strategic questions that can create the temptation to simply ‘start digging’. However, in order to drive meaningful learnings and answer broad questions accurately, breaking down the question into specific hypotheses to test will always bring more clarity. If you’re already driving a test & learn culture in your team, then this is the same 6 step process that applies.

  1. Ask a broad question
  2. Conduct background research to deepen your understanding
  3. Create one or multiple hypothesis/hypotheses
  4. Create experiments to test the hypotheses
  5. Analyse the results and formulate your learning
  6. Share the results

Data Science / Analyst Briefing templates

For those of us fortunate enough to work directly with data scientists or business analysts, we’ve created a downloadable briefing template you can use to bring more rigour to your processes.