Hey CEO an advice to take advantage of machine learning as Netflix does in your startup.
Netflix is well known for using machine learning algorithms in their platform, but sometimes, we don't know how "simple" it can be to have it in our company.
In my last article, many tech-guys were very upset with me because I simplify too much the topics. But that's the idea; Simple information for CEOs to short the tech-adoption and avoid some tech-cofounders or CTOS driven by passion (with good intentions) take the company into a quest of developing algorithms that you can rent before reaching the market.
Before you or your CTO start posting vacancies for Python developers with machine-learning experience or, even worst, start developing an algorithm in-house, think with the pockets and not with the hart. Yes, we all want to be the one who creates the next-gen of personalizing algorithms, but unless you have a barrel full of money to invest in research and development, take a look at the commercial offers.
I am pretty sure Amazon, Google, and IBM can give you a solution as well. But the one I know is Microsoft Personalizer, included in Azure cognitive services. With personalizer, you can have the same accuracy as Netflix by a fraction of the development cost and time.
Yes, you will need great developers (as always), but you and your team can focus on the User experience by renting the algorithm.
But let's see three things that Netflix is doing to make you watch more movies and series.
1- Similarity on your topics based on your ranks.
The most basic but efficient. Netflix lets you rank the content you like, and all the content is previously categorized. So, it will recommend you more content from that category. It's a good start when you don't have enough data to feed your model. (Even if you don't give the thumbs up, but you see the full content, in their system they give it a "reward" to that content in your preferences)
2- Recommendation by behavior.
My favorite by far. They (Netflix) will look at the historical preferences of people who behave exactly like you on their platform. For example: If you started watching "The Expanse," but you don't finish seeing the first chapter, you move to "Away" and complete all the season in one day. The algorithm will find the next content that all the people did the same as you liked the most.
The recommendation by behavior is excellent when you have data from many users but not for the new ones.
3- Attention catcher.
When Netflix know (or assume) they have content for you, but you don't click it, they will start changing the picture of the cover to catch your attention and make you give a try for it. Once you click on the content, they save the kind of image or colors that attract you, and it goes to your profile.
There you have three simple algorithms that you can rent from Azure in the same Personalizer API to improve your product. I guarantee you will be happier than ever when your team estimates costs and time by using this.
Your team must have a good understanding of M.L. without being experts and be skillful API's consumers, no matters what technology are you using in your development.
Share this article with your tech team if you think they are taking the long road, and if they are open-minded enough, I'll be super happy to have a call with them to share some experiences without any cost.
Or if you need to hire people with this kind of skill, consider my company to shelter them.