Excellent piece yesterday by Rob Collie at PowerPivotPro about seeing yourself as a Michelangelo of data.
Rob points out that back in the day painting was only for the elites.
But once the cost of paint fell, art became possible for the masses.
The same is happening with data. And data is the analyst’s paint.
It used to take massive computing power and technical know-how to analyze data. Now computing power is nearly free.
A corollary is that if computing power is cheap, then the cost of making mistakes is cheap.
In a way, making mistakes is the cost of creativity. Which brings me to the confidence interval.
We live in the confidence interval economy
One of the amazing things I learned in my first statistics class is that manufacturers don’t aim to make every product perfect.
Instead, they agree on an acceptable error rate and confidence interval. They accept that for greater things in the business, nothing can be perfect.
This is a mindset that spreadsheet reporters (and their managers) need to adopt.
Bean-counting is not bean-predicting
There is the “bean-counting” euphemism. Everything has to balance, or it is wrong.
What if we aren’t bean-counting, instead “bean-predicting” or “bean-analyzing?” Different exercise.
Some analysts want reductive or predictive models to have the same accuracy of full-blown financial reports. But these reports are not the same thing. They actually lose relevance and usefulness the more complex they become.
So what does this mean for your career?
A rambling post, but this idea of data as a medium of expression with low cost of making mistakes should shape one’s career path.
1 See your job as a creator
You are a Michelangelo of data. Stop with the “I’m an analytic type. I am not creative.” You are a designer whose medium of expression is the spreadsheet.
2 Use rapid computing to your advantage
Don’t spend too much time building the perfect solution. Use rapid prototyping to your advantage. I don’t wait to write the perfect blog post. Instead I release the “minimum viable product,” and test to see how it does.
If I see positive trends, I expand on them. If not, I pitch the result. Don’t get too hung up on duds or mistakes — it’s part of the process.
3 Find a boss who thinks in confidence intervals, not equilibria
The tricky part. Get a boss who understands how statistics works and the role of the error term. Bean-counting is quite different than bean-predicting.