The 5 Stages of Data Science Adoption

Some companies are built for data science. Others need to make some serious infrastructure changes in order for data science to work effectively for them. In this article, I outline the 5 stages of data fluency that companies go through on their way to leveraging data science. This is based on my experience of around 6 years post adaptation of Analytics in to my organisation and the progress there of

If you’re a data scientist, the stage of data fluency that your company is at will greatly impact the type of work that you do. If you are a data science manager or someone in charge of the data strategy for your company, this article can act as a road map to take your organization to the next level.

Stage 1: Data Collection

This is the foundation of the data science adoption life cycle. If your company doesn’t have quality data, and don’t have mechanism to collect the same, you simply cannot perform high quality data science. Organisations at this stage recognise that data may be important down the road so they begin to collect it.

This is the first stage of our analytical journey @Reliance Entertainment. Identifying the data points, building a mechanism to collect the data at application level, course correction to get relevant data points, building infrastructure to collect the data and storing the data.

This is critical stage considering, you need to make data strategy and you need to build this at your won at application level. Partners will come in to picture building infra to collect and store. Our system now collects 65 Million gaming events per day from mobile devices from across world.

Most companies are already past this stage, but it never hurts to reassess your data collection protocols and best practices.

Stage 2: Data Aggregation

Collecting data isn’t enough. Organisations realise that the collected data is useless unless it is cleaned and organised. Data scientists and analysts can work with messy data, but usually, it needs at least a little work.

Companies at this stage usually hire a data architect and start moving their data into some database infrastructure.

This stage can have more overhead than people expect. Data can be collected in a multitude of ways by many different systems. Data engineers need to take all of this data and format it so that it is cohesive.

Usually this process takes time, so it does at our organisation too…However my view is, this is never ending process, as your data grow you keep re-looking at your Data strategy.

Stage 3: Analytic Insight

Companies recognise that data can help them improve their decision making. At this stage, the insights are usually descriptive and are based off of trends found in pivot tables. Human decision making is being complemented with analytic insight. This work can generally be done by a team of analysts.

Examples of this type of insight are, let’s start with us : We see k lots of insights from data, like Game Mode Analysis, Churn Analysis, Spend Analysis, Effect of Design Changes, Player Inventory Analysis & Game Economy Analysis are some of them.

In general, You can see this insight exercise in finance or marketing divisions of companies. Companies segment customers and sell different products in different markets based on this type of analysis.

Stage 4: Advanced Analytics

You begin using more advanced machine learning models to improve decision making. This stage is characterised by models making decisions with human oversight. Organisations need to have buy-in and trust in their data science team to get to this stage of fluency. At this level, data scientists are a must.

Advanced analytics treats data science like a product. Data scientists produce research and models that are “used” by themselves or other people. An example of this would be if a trucking company built a model that predicted when new tires would be needed for the vehicles. This model, combined with a human assessment, would dictate when the tires would be changed.

This step can be seen once you Analytical ecosystem is mature. For us it is, considering the long journey and harvesting we have done to platform. The world call it AI but it’s actually ML exercise behind which can make your products more automatic, and can give you more insights. The data volume also play important role. The formula on implanting AI is ‘More the meaningful data, more effective your AI is’. We are building Churn prediction, and built a propensity to convert model algorithms around the ecosystem. Here it looks like ‘we are at start’ no matter how long you travelled to make data learn patterns. So again never ending task for data team and strategy team.

Stage 5: Analytic Integration

The final stage of the data fluency hierarchy is total integration of data science into some areas of the business. Here, machine learning models are put into production and are making decisions without immediate human oversight. At this stage, your data scientists (or machine learning engineers) are working with your engineering team to make data science a service.

Analytic integration treats data science like a service. An example of this would be a recommender system for an online website that showed customers different products based on their preferences and past purchases.

Believe me implementing Analytics from scratch on any digital business is a challenge however fun at the same time. I have implemented this, step wise and done a required course correction time to time from scratch. I enjoyed this a lot, considering one, your Engineering and logical knowledge in designing system will be in peak, and the results of this step on business are wonderful. When you contribute in Data driven decision making for the company and at the same time increase revenue because of your effective implementation of ecosystem, the pain, hard work and challenge all looks worth and you end up in a successful Data and business strategist.

Building a matured analytical setup is big and critical in your organisations Digital Transformation Journey.

#DataScience #Analytics #BigData #AI #ML #DigitalTransformation