.png)
I am not trying to sell you on this idea in the sense of converting you to it, I want you to play with it. I want you to think of its possibilities, I am not trying to prove it. I am just putting it forward as a possibility of life to think about.
— Alan Watts
Shall we play then and go on a journey of a beautiful possibility?
Music… how do we discover them today? We interact with Apps like YouTube, Spotify, etc and they recommend new possibilities for us. They predict what would I would like to listen next.
Today’s technology of recommendation engines has made it possible to predict based on user-song interactions without necessarily knowing the song or user details. Through this method, each song and user is a vector that encapsulates a lot more information than descriptive information like genre, music tempo, lyrics, etc. It can for example capture culture which is hard to capture as an explicit item.
Entertain the possibility will you, that we can do the same for products. From a playlist of recommended songs to a playlist of product features. A road-map for the product.
A product is not a user so we will have to tweak the approach a bit. Imagine you predict the next song for every user and then figure out which song is top of the list… what do you get? You get the “Future Billboard Top 10” and not the “Current”. This indicates the prediction of the future of music as a whole and not a specific user’s future. Let us model a product this way and find the “Future Billboard Top 10”.
Today there are multiple websites where products get feedback. From commerce sites like Amazon to developer sites like Stack Overflow. We will consider a developer site since it is closer to the product development process.
(Attribution: Tag data is from Stack Overflow)
For illustration let’s consider Android. Android is a great platform/product that has heavy interaction on Stack Overflow. There are a whopping 1.2 million questions asked by 10s of 1000s of users. 450,000 questions are not answered or have an answer accepted by the questioner. Therein lies a lot of intelligence of what developers want but have struggled to solve to help us define our “product feature playlist”.
Like in music… Vectorize… recommend… across all users… prioritize = our product feature playlist… the “Future Billboard List”
Below are the top 12 features to consider for the road-map. For each recommendation, we can figure and rank what are the other aspects/ tags people are asking around it to extract more meaning.
Like any product road-map, there is a lot in there. Here is one example zoomed out. The bars indicate context around which people have asked questions and can be checked for exact questions to get deeper context.
These are recommendations and can be processed in multiple ways. Like songs, the one that you get recommended is the one you may most like based on your personality and not necessarily the most popular. A popularity check on top would add finesse. Similarly one can check how popular (and engaged) is the user community on these recommendations and device a product build strategy. What are new use cases versus where can developer community be leveraged?
One can also understand these recommendations and translate the road-map to internal teams. Below are the key aspects people talk about around a feature. One can see aspects around android integration, android development modules, language-specific, new features like text-to-speech and new technology developments like TensorFlow. They typically may feed to different departments within the company. A typical output of a road-map.
This gives one view you may wonder. So I ran an ensemble of algorithms… leveraging deep learning to traditional machine learning and found what machine recommends most. One can add to the robustness of the road-map playlist by crunching various other relevant user feedback… Google Play App feedback, Android mobile phone user comments, etc.
In today’s high pace world, the road-maps can be generated every quarter and not restricted to a monolithic yearly exercise. Product managers… think of this possibility and introspect… can you crunch this volume and compete with a machine at this velocity?
It’s a possibility to entertain… a playlist of product features. Increasingly everything is a vector… time to do thought experiments on what other playlists are possible… strategy playlists, market playlists, insight playlists… enjoy the new life of AI possibilities. Intelligence redefined.