Data is useless



The quote goes "without data you're just another person with an opinion".


I'd say "with data you're just another person that mis-interprets it to fit their beliefs". Not as catchy. I hate quotes anyway.

Data is EXTREMELY easy to mis-interpret, either on purpose or unknowingly. It is usually the latter - think about it like a confirmation bias X 10. Not only it can be useless, but it can be damaging.

Common misconceptions 

It's machine learning, dummy

"We have a proprietary machine learning algorithm". That is mostly bullshit.

Machine learning can be extremely simple. Say you want to predict sales tomorrow and you have a few days of data, showing a 10% increase in sales day over day. So 2 days ago you had 100 sales and today you had 110 sales. Machine learning can be as simple as predicting that tomorrow you will have 121 sales. It is all about prediction.

Most machine learning is simply a regression with a human deciding which variables go into the model. IT IS NOT MAGIC.

You have to understand the limitations of the underlying techniques, and there are many. You need to get a good grasp on prediction uncertainty. If I told you tomorrow you are going to get 100 sales but the statistical confidence interval was between 0 and 140, would you take 100 at face value? You shouldn't.

There is more uncertainty, driven by the data itself. Too little data or un-representative data lead to uncertainty. Most times that uncertainty cannot even be quantified but having an understanding of that uncertainty should be part of your decision making process.

My analysts know what they are doing

They usually don't. They are usually inexperienced. Even if they have a solid grasp on statistics , they will likeley not have a full grasp on the business.

The managers know what they are doing

They usually don't. They are usually experienced but they don't really use numbers that much. It's all about the soft skills. Likely to have a great understanding of the business, that can also be a liability as any analysis will have some bias in it. 

You know what you are doing 

Great, I'm done here! You've probably been reading this blog.

Analytics should be a shared responsibility 

Every person in your organization should have an input into how you interpret and use analytics, not just the people that are dedicated to that area. 

No one has the full picture and even experienced analysts and manager will make mistakes that can really hurt the business in the long term. 

Trust your colleagues but question everything.

Why should you read this blog?

You're an entrepreneur

Fantastic. You're probably bootstrapping. Your data is not exactly big. There will be a lot of content here that will be useful for you. How do you project the life time value of a customer? How do you measure the impact of your bootstrapped marketing strategies to make sure you're not wasting money? How can you use analytics tools to run your day to day business operations?

You're a manager

Hopefully you have biggish data. And you're probably not really using it. If you want to know what you can learn from it, you're in the right spot. Also, you want to learn about data pitfalls and how to avoid them. And about dashboard and maybe interactive data apps that your analysts can build using *MACHINE LEARNING*

You're an analyst 

You'll find this interesting too. I'll put links to git for each project I showcase. I usually attempt to solve hard problems with very simple methods. I also know stuff about big data (that is petabytes).

You don't really fit in a box

But you care about data and how to use it to better run businesses and make decisions. It would be nice to have you around. Please subscribe to my mailing list to be the first to know when I publish something new. I don't spam. Promise.


Share:

0 comments