Reorganising media product analytics teams reveals new decision-making opportunities

By Utkarsh Arora

The Economic Times

Noida, Uttar Pradesh, India


Do you feel your analysts resemble “spreadsheet reporters,” who are just good at repetitive tasks, such as calculating basic ratios in spreadsheets and copying that data onto e-mail reports? Or, do you see them as business consultants and partners, who can guide decision-makers toward new untapped opportunities or optimise existing ones?

Have you seen stakeholders persistently chasing product managers or business leads to get some basic data?

Do some of your non-business teams feel clueless and disempowered because they don’t have sufficient context about your digital platform’s audience profile, feature performance, or reader engagement habits?

Do you feel the need to arouse people’s curiosity and build a culture of innovation wherein all people — regardless of their systems, context, function, or role — are empowered to play with data easily, generate ideas, validate using data to confirm their gut reaction, and take a lead in trying out new initiatives without being asked to do so by the top management?

Four or five teams should focus on different aspects of the value chain.
Four or five teams should focus on different aspects of the value chain.

Are stakeholders able to get into a meeting and walk out with a clear decision? Or does it take numerous meetings to close the decision, and, still, your teams end up with insufficiently reasoned decisions as they run out on time?

Is your data deeply analysed? And does it connect the dots? Or does every stakeholder have a different opinion or interpretation when they look at a common data set?

These problems seem different but can be solved by effectively envisioning and organising your product analytics team. I’d like to share a basic structure you can use to organise your analytics team members to resolve issues like these in a sustainable way.


There are several objectives to such an exercise. First, it democratises access to clean, structured, and cured data across business, editorial, and product teams. Also, it significantly reduces time to discover deeper insights by enabling connected data sets across databases.

Finally, it empowers and enables analysts to take the lead and drive product, editorial, and business managers to discover new opportunities and make effective decisions.

I suggest there should be four or five teams (pods) that focus on different aspects of the data value chain.

Team 1: Data governance pod

The first team’s primary objective is to elevate data quality across databases to enable it for use in business intelligence (BI) and data sciences.

This team’s duties include:

  • Conducting audits: It must conduct manual and automated audits across devices. Any errors in data ingestion across channels in data flow should be reported to the data warehouse.
  • Cleaning and data massaging: It must identify opportunities for cleaning data and enlist corrective measures to the BI team.
  • Operationalising standards for data ingestion and storage: It must define, evangelise, and ensure adherence to standards across product manager teams for data ingestion.
  • Instrumenting intelligent alerts: It must monitor instrument alerting mechanisms to proactively identify data flaws before they impact BI systems.
  • New data ingestion and quality control process: It must systemise sprint planning to new event ingestion and quality control processes with the tech team prior to going live.

Team 2: BI and data product development pod

The second team’s primary objectives are to centralise data; create faster, deeper, and well-rounded insights; and make exploratory analysis delightful.

This team’s duties include:

  • Data cleaning and massaging: It must set up automation for cleaning and massaging data using BI tools or with help from the data engineering team.
  • Data access democratisation: It must create local data lakes and self-query systems. Additionally, it must create a cheat sheet of ready-to-use queries for 90% of the most frequently required information.
  • Developing personalised business intelligence applications: It must link multiple data sources to offer well-rounded cuts at all levels of granularity on BI applications that connect the metrics and custom measurements across different data storage systems.
  • Stakeholder training: It must train and enhance BI toolkits.

Team 3: Exploratory data analytics pod

The third team’s primary objective is to develop and hone a culture of presenting fresh and well-rounded insights on specific problem statements or exciting hypothesis every 10 to 15 days by each analyst on the pod.

This team’s duties include:

  • Identifying hypotheses or high-leverage specific problems for 15-day sprints: It must source semi-structured problems or hypothesis statements by consulting with stakeholders.
  • Identifying or deriving the right set of metrics and uncovering hidden insights by running structured analysis on merged data that is connected across tables.
  • Shifting focus from 5%-10% time invested in hypothesis exploration to 80% or more time on exploratory data analysis.

Team 4: Predictive modeling and personalisation pod

The fourth team’s primary objective is to make decisions proactively using prediction services. It must automate or augment intelligent decision-making in products and processes by building data services that leverage machine learning.

This team’s duties include:

  • Taking matured problems to create predictive data science models for key business problems.
  • Building personalisation algorithms and expose them in the form of APIs (backend-only applications) that can be plugged into multiple systems for deploying personalisation (such as your e-mail sending lists, app push notifications, the home page, and recommendations inside your stories).

Potential use cases for media subscription businesses

There are two primary use cases for media subscription businesses worth considering: subscriptions and content recommendations.

For subscriptions, considerations include:

  • Advertising versus paywall. There should be a revenue maximisation model for dynamic paywalls.
  • A churn propensity model.
  • A personalised plan assortment and discounting.
  • Pitching upgrades and renewal offers to the right audience with and effective plan and pricing.
  • Curated push notifications by persona.

For content recommendations, considerations include:

  • Recommended stories, stocks, and events based on recent behaviours and habits.
  • Adding “x” to your watchlist or an alerting programme.
  • Discovering what other members from your industry, profession, or locality are consuming.
  • Quantifying your learning score on a specific topic.
  • Cross-format recommendations that drive a diversity of habits. For example, if someone like long-form articles, perhaps they should try a podcast.
  • Content hyper-personalisation on the home page.

About Utkarsh Arora

By continuing to browse or by clicking “ACCEPT,” you agree to the storing of cookies on your device to enhance your site experience. To learn more about how we use cookies, please see our privacy policy.