Four layers of analytics that data-driven companies are using

Companies with a data-driven mindset are always looking for a new way to build tools, abilities, and more importantly, a culture to perform timely action on data. They have more data than ever before:  traditional and non-traditional, relational and non-relational, simple or complex.

Usually, they start their digital journey with the most critical data, i.e. customer data, residing in their backyard.  Digital transformation leaders always advise to look first into the company’s current digital platforms: websites, apps, and kiosks, all documents and excel sheets, shared file systems, shared services, and legacy applications. And so on. Then depending on its industry and capability, companies start looking into the non-traditional sources of the massive pile of data. Examples:

• Customer and prospect interactions via phone, email, online chat.
• Social media, direct messaging, tweets, etc.
• Sensor data from sensors, RFID tags
• Physical movements logs, records from smartphone apps, etc.
• Web analytics logs

But without actionable knowledge, these data are not very helpful. That’s why companies are deploying their big data ecosystem to ingest the right data from all these non-traditional sources, clean the dirt, and make it ready for their data scientists to run analytics and get actionable insights. Common industry experience says data scientists spend almost 80% of their time obtaining, cleaning, and preparing data. They spend the rest 20% of the time on analytics. However, the analytics part determines success or failure.

A company can profit from its data-driven process only if they do their analytics right. That’s why business leaders in almost all forward-thinking companies are spending a massive amount of time and money in this process to solve their challenges, beat their competitors, or to stay relevant in the market.

What is analytics?

It is about getting the knowledge out of the data by finding a meaningful pattern. It can start from the most basic one and can be extremely complex. In its core, analytics try to answer the following four core questions.

• What happened?
• How or why did it happen?
• What is likely to happen?
• How can we make it happen?

Gartner has developed a useful framework for classifying application areas of analytics (see picture below).

Picture Source: Gartner

Gartner’s Analytics Acsendary Model divides the analytic effort into four categories: descriptive, diagnostic, predictive, and prescriptive. I feel this model is a very effective way to discuss analytics process.

Descriptive analytics

Descriptive analytics helps companies to understand what happened. It provides data-driven factual information. Often this is the starting point of companies analytics journey.

Consider the questions like

• What is the lifetime value of our customer?
• What is our cost of marketing?
• What is a marketing campaign’s success rate?
• Which demographic creates most business for us?

Getting answers to these questions is all about collecting, cleaning, and processing data that we already have.  The outcome of this process usually signals what is wrong or right, without explaining why. For this reason, data-driven companies use descriptive analytics as their first step and use more advances layers as mentioned below.

Diagnostic analytics

At this stage, data is measured against other data and several parameters to answer why something happened. Diagnostic analytics is problem-solving efforts that drills down, find out dependencies, and often done through ad-hoc queries and basic statistics.

Data scientists use both art and science to discover insights, patterns, uncover dependencies and communicate their findings using visuals to appropriate stakeholders and decision makers.

Predictive Analytics

Predictive analytics tells what likely to happen. It uses the findings of descriptive and diagnostic analytics to identify tendencies, clusters, and exceptions to predict future trends.

Predictive analytics helps to understand the likelihood of future events. Examples are supply and demand, income forecast, system failure, or how a specific financial product will perform in near to long term.

It uses more advanced tools and techniques as compared to diagnostic analytics. This process has a significant impact on companies top and bottom line.

Prescriptive analytics

Prescriptive analytics tells what should be done. This is the final layer and the most advanced and sophisticated application area of analytics. As its name suggests, this process prescribes what action to take to eliminate a future problem or take full advantage of a trend.

Examples can be the right pricing model, product recommendation, correct asset allocation, proper recommendation to increase sell, fraud detection and so on.

Use Case

Companies often use these four analytics layers in succession to reach their goal.  Let’s consider a financial service company manages the portfolio of customers. The company offers products suitable for their customer’s objectives and risk appetite. In recent times, they have experienced a loss of customers. They are willing to use all these four analytics application areas to get actionable knowledge.  Some sample findings in each of these areas are as follows.

1. Descriptive analytics flags customer attrition from a specific demographics (say: married, millennials with X income group) during the last one year.
2. Diagnostic analytics reveals that the loss is because of the poor performance of a recently launched financial product for those customers.
3. Predictive analytics can forecast market movement and better risk vs. reward trade-off for medium to long-term time frame analyzing stock price movement, market returns from past years. It also determines the group of customers that are likely to move out near to long term.
4. Prescriptive analytics will recommend a new product with that asset allocation model targeting those customer demographics.

Takeaways

Taking data-driven action is now a key differentiator. By applying the analytics, companies with a data-driven mindset are getting tremendous values.

Analytics are often divided into four layers of increasing complexity. Based on capability and process maturity, companies are free to choose how deep they should dive in their analytics journey with the end goal of extracting actionable knowledge out of their data.

In upcoming posts, we will discuss artificial intelligence and machine learning.