Blogs

Here we go, this is the last question! Thank you for following, sending DMs, and all the great conversations over coffee :). I learned a great deal myself and hope that whoever reads these posts finds value in what I have shared.

I also have to send my gratitude and appreciation to the great folks who run everything SAP. These posts are inspired by the power of goodness that SAP has to offer on so many levels. Now off to the last question:

Question 6: What is the success criteria?

As you may remember from the previous posts, you must feed the algo with data then the algo optimizes the training factors. So who ever leads the ML project must constantly keep an eye on the process to ensure that it supports the business objectives. Therefore, before kicking off the project you must first define the business objective and ensure to be crystal clear about the evaluation criteria. The evaluation criteria is the input that the algo functions on and is being trained on.

Here is an example from Airbnb to show you how this can be applied in real life. The article is from Inc.com and reviews Airbnb's success with machine learning. I have also linked to the original article in the end of the next paragraph.

"....one of the primary success metrics is the platform's conversion rate -- how many people make a booking. Airbnb also examines how long it takes a user to choose where he or she wants to stay, optimizing for quicker decisions. .... introducing a deep neural net to the search-ranking system boosted Airbnb's recent conversion rate by 1 percent. One percent may not sound like a lot, but as you can imagine, a 1 percent increase in conversion rate compounds over time... Curtis doesn't just see Airbnb's use of cutting-edge technology in terms of ROI, but also as a hopeful sign regarding innovation's impact on the future. He pointed out that Airbnb provides a nontrivial chunk of income to many of its hosts." Source: Inc.com

On that note you have all the 6 questions that you need to qualify a business problem as a machine learning problem. Here is a list of all six questions with links to my posts on each question for your reference:

Question 1: Do you really need machine learning?

Question 2: Can you clearly formulate the problem?

Question 3: Do you have data? I mean enough data?

Question 4: Does your problem have a pattern?

Question 5: Can you find meaningful representations of the data?

Question 6: What is the success criteria?

If you ever have any questions, don't hesitate to reach out, I either help or know someone who can help you.

Building an ML practice is not for the faint of heart. I would like to leave you with a lesson that I learned on my vacation this week. When I saw the dude in the photo (AKA the statue of the McGill Student) on the right, I quickly handed my cell phone over to my friend and asked for a photo. The sidewalk was busy and people were passing by quickly without paying attention to this statue. But as soon as I hopped on the bench and started posing with it, a few folks stopped to look, then I could hear them saying let's take a photo here and soon before we knew it there was a line-up (which I wish I had a photo of)! Although I knew that I already lost the opportunity to take a few more shots, (because now there was a line up of people who wanted to take photos) I was secretly enjoying the fact that I wasn't the only one with a photo of this statue that afternoon; I knew that there will be more photos, with more people, and with more social media presence for this piece of work from our beautiful Montreal. Even days after this incident, I was still pondering about what happened that day. I made people stop, look, and follow what I did. Then I asked myself, how we could all change the world around us if our actions would make others pause, look, and repeat!

These six blog entries may have equipped you with the questions that you need to qualify a business problem as an ML problem. But you maybe the only one in your organization, company, department, or team who is aware of the value of ML. What can you do to make your team to pay attention, think, and want to see more of what machine learning could possibly offer? And if you are successful to steal their attention, don't forget to fade into the background and let your team shine:)

Thank you again for reading. Now it really is time to cook something yummy so that you have the energy to start building your ML practice. Let's ask Chef Watson what you can cook today: https://www.ibmchefwatson.com/community

NY Times once mentioned that "Watson makes suggestions that no human would ever make, like adding milk chocolate to a clam linguine or mayonnaise to a Bloody Mary."

Here is what I cooked. Watson said a big orange instead of the regular stuffing would make it tasty. And oh boy was Watson ever right! :)