All Meat, No Fillers: Tableau Conference 2019 Session Wrap Up
Here at Keyence Analytics, we had a unique opportunity to not only be at this year’s Tableau Conference, but to get to release Ki to the data world. It was a ton of fun letting everyone there interact with the system we’ve all grown to love. As fun as Vegas and these types of conferences can be, (catch me on my break laying out in the data ball pit pool) it was a great experience to meet the #DataFam and be able to learn from each other on all things data, data viz, and Tableau.
Director of Artificial Intelligence, Rob Ortiz, lead one of our sessions, All Meat, No Fillers:
Sometimes the hardest part of analytics isn’t the dashboard. It is figuring out what to put in it. Too much data leads to bloat. Slowing down decision-making and processing time. So put your data to work better, leaner even. Come see how Keyence’s Ki AI trims the fat on data. Add a touch of salt and you have the making of a real treat!
If you’d prefer to watch his session, check out the full video here. Otherwise, let’s dive in and break it down.
Asking the Right Questions
“The job of the data scientist is to ask the right questions” – Hilary Mason, GM of Machine Learning, Cloudera.
Asking the right questions is core and critical to any business setting. Of course, if you had the ability to ask the perfect question in every situation, you’d be off somewhere consulting, making millions. But for the rest of us down here, we often get asked the one question by our boss that we tend to dread…
”Hey…….. quick question.”
Ah, yes, the one question that can derail your week and send you down a rabbit hole of a new or extended project you weren’t expecting. Nevertheless, that’s business and we often need to dig deeper into our dashboards to find those answers.
These quick questions usually lead to a discussion or strategy meeting to dive in deeper. In that meeting, your boss asks something such as; “Why sales have dipped below the threshold set on the dashboard?” Ideas start to fly around the room and you begin the analytics process. You start with the obvious answers then move to the wilder and far reaching ones. Some of this analysis you have the ability to do with the dashboards and sheets you’ve already made. Others you will need to build new ad-hoc reports and still others you may need to reach out to others for help on.
Fast forward a few hours to those completed dashboards and it doesn’t exactly show what we were looking for. As you go back to make adjustments, create sub groups, and more dashboards we have fast-forwarded again, and again. By the time we think we are done building the dashboard or analysis that answers our question the days have flown by and we have wasted a week or more on the problem. We are being reactive, and no longer proactive.
Why Do We Get Stuck Into This Cycle?
As business and marketing analysts, our job is to build deliverables and provide actionable insights to our teams. But the risk of being stuck in that question cycle can be all too easy.
As analysts, we use business intelligence tools to become this all-powerful ‘Genie’. Someone that can answer all the questions and make those wishes come true. After all, you are the domain expert, the Zen Master, or someone that leverages AI tools to get where you need to be.
Now as that ‘Genie’, you build the most beautiful ‘lamp’, your dashboards. Often times, your dashboards send you alerts, show problem areas, help you develop insights and more. And of course, they look beautiful!
But at the end of the day, it’s the skill of the ‘Wisher’ that makes the biggest impact. The individual who says the “Well how does this…..” or who comes up with the analysis that should be done, they are the one who is the master of your destiny. Sometimes that person is you, sometimes it is your ‘customer’. Regardless who it is, you/they must ask the right questions to lead to the right results.
For example, if your dashboard alerts you to a problem area, in Rob’s example he discusses seeing an issue with a sales line’s repeat rates. You’ve now been able to see this glaring issue and are able to move forward figuring out how to fix it. You ask yourself and your team questions like ‘maybe it’s by region’, ‘by sales amount?’, ‘Let’s look at it by industry’, ‘Or is it by actual product?’ Now that you’re twenty dashboards deep, many strategy sessions and some adhoc reporting later, it’s been three weeks and you finally get to a solution (hopefully).
What if you were able to cut through all that in the beginning and find out how ‘this’ impacts ‘that’ or how ‘this’ causes ‘that’. Well what you’re doing with or without realizing it is ‘Feature Engineering’.
What is Feature Engineering?
Feature Engineering is a term typically discussed by data scientists around the water cooler, but it isn’t limited to them. The art of Feature Engineering is the process of taking data and through the use of your own expertise, reorganizing the data to pull out the most important pieces of information.
Feature engineering is the art of looking at something like ‘cost’ and ‘sales price’ and being able to build ‘profit’. Or to understand that your industry has historically annual buying cycles (like retail). This ability to make the data speak more clearly to your business language is all part of feature engineering. And is something that comes with lots of experience as a subject matter expert.
But what happens when you don’t have lots of experience as a subject matter expert? This was what Keyence tried to find a way to solve. How do we help an entire generation of users who didn’t have years of subject matter knowledge? How do we help them to ask the right questions? Well it started with us asking ourselves the right questions.
What Are The Right Questions?
The First Question: What is the most important information to show?
With an infinite number of angles we could use or look at, what is the piece of information that we have to look at that explain the situation. Too much data causes bloat and slows down the decision making process (Analysis Paralysis).
The Second Question: What Can We Change?
It may be interesting to find out that we have higher sales when there’s a full moon but what kind of change can we actually impact? When we look at the “question to ask” we have to insure that whatever answer we get, it is something we can impact. Without impact, analysis is useless (yes…I said it….send all hatemail this way).
The Third Question: What has the largest impact, for the least amount of effort?
We’ve all heard the adage of working smarter not harder. What can we find that has the largest impact to our goal but doesn’t require a vast amount of resources? We all have limited resources, be they financial or manpower resources. Every day you will have 100 different ways to spend your time, if you HAD to choose just one. Which would you choose that had the highest impact?
Making the Magic of Ki
These questions lead Keyence to finding a need to figure out how to not only automate this process but also make sure it was usable by your average business user as a no code solution. One of the key features that makes this no code solution possible, and we touched on this briefly before, is the automatic feature engineering (AFE).
AFE automatically combines different data sources together into one table without the need of doing any manual joins, aggregations, or data blending. It does this by looking at a specific target that you select, for example, the total amount of sales per month for each sales person. Once this target is selected, all of the data is aggregated down to that level without any effort by the user.
Ki uses this information to build any aggregations, or calculations. Then it selects the most important information that answers the question or target of your investigation. It even simulates business strategies to correct the issue and allows you to export them in a format ready for import into Tableau to make your strategy dashboard.
All of this sounds really sci-fi to those who have never seen a machine do in minutes what takes people with years of experience, weeks to do. But that is why we made the system. It’s been a massive help to our own people internally, and now we are excited to see others interacting with one of our favorite teammates.
LATEST POST : What Is Prescriptive Analytics?