Ease of use driving the analytic revolution


Having used alteryx and tableau for quite some time, it’s made me reflect on why it’s made me enjoy using those products and the answer is really quite simple.
It’s all to do with the ease of use and simplicity whereby anyone not just the technical guy can easily combine data sources together and try and answer a question.
Ease of use trumps features

Most business people are actually very time poor,juggling numerous priorities. As such they require tools that are simply ‘easy’ to use and understand.

Take a tool like Google for example, if you type in a google maps how do I get to the Sydney opera house, Google essentially delivers you an answer within 1-2 seconds.

The cost of making an incorrect question is simply reduced to a few seconds. If I decide to go to manly beach instead of the Sydney opera house, I simply type it in and an answer pops out almost immediately . The tool ( in this case google) simply disappears as I can follow my thoughts and ask questions of Google.

The ability to continually iterate within seconds fundamentally changes the way we work.

I can continue to follow that thought process and then see what shows are on at the opera house and the proceed to try and book tickets as it opens up in another browser.

From the customerperspective, this is showcases a fantastic user experience.

Customers and users should not need to care or worry about the underlying architectures, they simply get the answer to their questions immediately.

Simplicity is the best form of elegance

Looking at the field of analytics, it is certainly far from easy.

Many companies struggle with numerous separate systems built from years of acquisition. The whole notion of single source of the truth then becomes difficult due to systems not talking to each other, different definitions etc etc.

Smaller companies struggle with resource levels, funding and also system complexity as well. So how can we solve this problem in the place of continual technological changes?

Make sure you use the right tools then!

Alteryx and Tableau enable me to flow with my thoughts enabling me to deliver answers as quickly as I can think of questions. This enables me to focus on asking the right question.

Just like a carpenter requires the right kind of chisel and saw, you should be using tools like alteryx to easily combine the data , making business rules fully visible and audit able, run spatial or predictive and then visual information in tableau. The benefits are enormous !
Data preparation the key to unlocking predictive analytics

Many companies are still embarking on that journey of predictive analytics hiring data scientists that can code in R or Python and trying to build their models from the ground up. This turns out to be an extremely expensive and time consuming exercise. Worse still , the majority of the effort is spent trying to acquire and structure the data.
Turns out that to do predictive analytics one must spend nearly 60-70% of the effort acquiring and cleaning the data while model building only takes 10-20% of the time. Alteryx excels in this area of data prep as its a completely visual workflow that enables anyone to do data prep not just the SQL guy or data scientist.
Alteryx enables anyone to do predictive analytics ( PHD not required)

predictive model

Also from a predictive modeling perspective , Alteryx also has 20-30 R predictive models (from regression to decision trees and neural networks) builtin which anyone can just drag and drop into the workflow saving analysts hundreds of hours and enabling them to focus on what the model that best fits ie: See lift charts below

Lift models

Don’t try to reinvent the wheel, these results are often good enough for most companies ( unless you are Google , Yahoo or LinkedIn. :-).

Don’t just take my word for it, go download the products yourself and try them out.


Some good materials i have also found regarding the forecasting & predictive models

Getting started with Additive models in R – Difference between simple regression (aka fit a straight line), Polynomial and additive models

Estimation, Out-of-sample Validation, and Forecasting – Helps me understand the rationale behind the create sample tool in Alteryx

create sample tool

Mean percentage error

Mean absolute scaled error

Forecasting – ARIMA


Box Cox transformations – Basically shifting a skewed dataset to become normally distributed