Recommendations and User Journey in the Age of Digital News

21/Mar/2017 Posted By admin no comments.
Tags , , ,

“Data are just summaries of thousands of stories – tell a few of those stories to help make the data meaningful”. Chip & Dan Heath, Authors of Made to Stick, Switch

Analytics dashboard are a great way to understand how users are finding the content, how they are interacting with it, how often do they comeback, how much time do they spend on the site so on and so forth. All this is invaluable information that helps define the content strategy, find new avenues/segments for growth and make other business decisions.

But, what if the data that we gather for analytics dashboards can also drive user engagement, make them spend more time on your site and make them keep coming back for more. Say hello to piStats!

piStats goal is to not only provide great insights into your business, but also provide data driven services that can directly add value.

One such service is piStats content recommendations.

Content Recommendations

Understanding your audience’s mindset is a slippery slope. It can accelerate your page views, user base, or it can be disastrous! Well, in the simplest of the tones, we can explain it as an art of story-telling.

To communicate effectively with data, you need to tell a story with it. While data relies on logic and reasoning, decisions are often made based on emotion. Merging logic and emotion can be a powerful combination to drive action from your insights.

One of the strategies that can improve user engagement, is to make it easier for users to find what they are looking for. Bubble up the information they would be interested in and information that is relevant to them. All the readers are not looking for same content all the time, even each user’s preference changes overtime.

The key then is to personalize the experience for each user, so that, what they see is tailored for them and is data driven.

At a very high level we can say that piStats recommendations is a pluggable module that learns each user’s preferences and then finds the best content to recommend to each user.

How does it work

In recommendation engine parlance, piStats is a hybrid recommendation system. What this means is that, it is a finely balanced mix of different algorithms. Instead of relying on a single dimension like content similarity, piStats takes a few different attributes to find the relevancy. Some of the most important aspects are:
  1. Context – In the current user journey, which stories have been read by the user in order and which is the most probable story user is going to read next.
  2. Recency/ Trending – Breaking news is something that most users like to keep an eye on. So, the contemporary the news, the more relevancy score it has.
  3. Popularity – Which stories are the most read stories on the site. The more popular the story, the more weightage it gets.
  4. Content Similarity – Stories which provide similar enumerations along the current story are also considered.

So, think of it as finding a relevancy score for each of these factors against all the content that can be recommended. Once the relevancy score is calculated, the most relevant content gets served to the users.

Now, let’s up the game. Think of doing it in ‘real-time’ for a million users and finding relevant content for them, on each page they visit.
This is exactly what piStats recommendation engine does!

Key features:

  • Configurable per client asset – All the machine learning algorithms that are used by piStats are home grown and they can be tuned for each client. Which means it takes a very short period for the recommendation engine to learn about your users and start dishing out great recommendations.
  • Real time – This, of course, is the basic requirement for any recommendation system, because it needs to react to the ever-changing nature of news publishing. For example, on election’s result day, all the users might be looking for election results, but the very next day or even by evening that day, the biggest news might be related to movies. Multiple hats to juggle, piStats does it well.
  • Multi lingual – Most of the recommendation engines work well with English, but we are proud of the fact that piStats works seamlessly with any language, be it a regional or global language. We already have it running for multiple Indian regional languages as well as English, for a leading 24-hour daily media giant.
  • Metrics – piStats module also provides the metrics on how the recommendations are doing. How many people are seeing them, how many are clicking on them. How is recommendations effecting the user engagement?

As they say, if you can’t measure it, you can’t manage it!

Hope you liked our little take on what power does recommendations impart to your user’s digital journey.
If this made you look for deeper answers and suggestions, here’s a jackpot for you! Its not only the customers who get benefitted by personalized recommendation, recommendations can help your editorial team too. Download our free eBook on ‘The Art of Mastering Tag Recommendations’ to see how automating tags can make the content more manageable, searchable and uniform.

Also, if you have any feedback to share or want your brain ninjas to help you overcome any related issues, please feel free to write us at

Till then, Happy Recommending!

This blog has been written by Mr. Gaurav Batra, one of our core Leadership Team’s Member

Gaurav Batra, Head, Products

A technology enthusiast, Gaurav has worn multiple hats all along – from designing and developing applications and databases to taking care of everything from requirement analysis to client delivery, he’s been there, done that! In his short stint at BluePi so far, he’s worked across the entire tech stack – Javascript, Meteor, ElasticSearch, MongoDB and the likes.

With renewed focus on creating a product roadmap and figuring out the go-to-market strategies at BluePi; Gaurav has his task cut out, for sure!

Category : Analytics CMS