How AutoML Paves the Path to Predictive and Prescriptive Analytics

Over the past five-plus years, “predictive analytics” has been alternately equated with advanced analytics, machine learning, and artificial intelligence (AI). But in all the confusion about the term, not a lot of companies are realizing the benefits by integrating predictive analytics into production systems to help transform their organizations.

Now, the term “prescriptive analytics” is gaining momentum. Is prescriptive analytics just another way of framing predictive analytics, or is it just another meaningless buzzword that software companies and data scientists throw around to impress and confuse?

The short answer is an emphatic “NO!” When done correctly, both predictive and prescriptive analytics provide powerful insights that help any organization make better decisions, increase productivity, reduce costs, and drive revenue. But there are specific differences in the two advanced analytics methodologies and how they are used.

First, let’s talk about the different types of business analytics. This chart shows the Gartner Analytics Ascendency Model as defined by Gartner back in 2012. Although the chart can be a bit misleading – most successful companies are utilizing all of these types of analytics methodologies and not “evolving” from one to the next – it offers some insights into how these different analytics methods are valued based on the difficulty in implementation.

Gartner Analytics Ascendency Model

First, here is a quick description of each type of analytics and how they are used, and then we’ll focus on predictive and prescriptive analytics.

Descriptive Analytics: What Happened?

Descriptive analytics provides a view of the key metrics and measures within a company.

This is the most common type of analytics in use today. A common example of descriptive analytics is a simple profit and loss statement. Another example is a breakdown of a company’s customer base – i.e. 20% of our customers are work-at-home mothers and 30% have part-time jobs. Using business intelligence (BI) visualization tools like Tableau or Qlik helps to bring descriptive analytics to life, especially for a business audience.

Diagnostic Analytics: Why Did It Happen?

Diagnostic analytics are the next step up in complexity for business analytics. Diagnostic analytical tools empower users to drill down into the data from descriptive analytics to isolate the root cause of issues for problem-solving and decision making.

BI visualization systems like Tableau and Qlik enable diagnostic analytics by allowing users to explore data visually by clicking on different pieces of a chart or graph to dig deeper into a specific metric or metrics. For example, a retail chain can drill down into the numbers for an underperforming store to see if there are any issues like weather or employee turnover that cause the drop in sales giving insights into how to correct the issues.

Predictive Analytics: What Will Happen?

Predictive analytics delivers a look into the future. Whether it’s forecasting the likelihood of a customer churning or estimating the total revenue next quarter, predictive models created with machine learning provide the necessary intelligence and speed.

Using historical data and statistical modeling, predictive analytics looks for patterns to create a predictive model that can predict what will happen when new data is applied. For example, a business analyst at a financial institution can build a predictive model from the company’s past loan performance to predict if a new loan applicant will default on a loan, helping loan evaluators decide to write the loan or not. These predictive models augment human decision making with deeper insights, allowing them to make better-informed decisions.

Prescriptive Analytics: How Can We Make It Happen?

Prescriptive analytics builds on the “what will happen” insights from predictive models to help determine why it happened and what action you can take to influence the best outcome.  A prescriptive analysis typically offers a host of possible actions to take based on current conditions.

A very good example of prescriptive analytics is the traffic application on your smartphone that helps you choose the best route home after work. This application looks at the distance of each possible route, the speed you can travel on each segment of each route, and most importantly, the current traffic situation. By analyzing all of this information, the application can prescribe the most efficient route for your trip home from work at that time.

What’s the Hold-Up?

Even though most companies are using descriptive and diagnostic analytics already, many have yet to implement predictive analytics into their operations. And even fewer are using prescriptive analytics to provide actionable next steps to solve issues that are discovered using predictive analytics.

There are very good reasons for the delay in the adoption of predictive and prescriptive analytics, but chief among them is the shortage of data science talent that is hampering all but the largest companies. There are just not enough skilled data scientists for every company to take advantage of predictive and prescriptive analytics, and this shortage will only intensify as demand for data scientists continues to accelerate.

The good news is that automation is being applied to the advanced analytics space. Automated machine learning (AutoML) solutions integrate the expertise and skills of advanced data scientists along with the speed and flexibility of massively scalable software systems, all while delivering features that allow business analysts and business users to develop advanced analytics. In short, AutoML solutions, like the Ople Platform, empower non-technical business analysts to practice advanced analytics.

AutoML and Advanced Analytics

A true AutoML system delivers an end-to-end solution for advanced analytics. From data ingestion to model building to deployment, AutoML platforms help organizations build and deploy AI solutions that span predictive and prescriptive analytics that can be developed by just about anyone within the organization.

With the Ople Platform, the process for creating machine learning models that drive predictive and prescriptive analytics is simple:

The simplicity and power of the Ople Platform allow anyone with knowledge of your business and your data to quickly develop AI solutions that enable both predictive and prescriptive analytics. And by enabling business users – business analysts, BI specialists, and business leaders – to work effectively, companies of any size no longer have to rely solely on data scientists to create a competitive advantage with advanced analytics.

Ople and Predictive Analytics

Users of the Ople Platform create powerful predictive analytics solutions by simply bringing their data into Ople, answering a few questions, and letting the platform build the best predictive model for their analytics challenge. 

For example, an online retailer was having some issues managing email communications with customers and prospects. Over the years, they developed more than 50 email campaigns for different situations, including abandoned shopping carts, not browsing the site for a few months, and different sales offers. But knowing when to send which email to which customer was challenging, and sending the wrong email at the wrong time increased unsubscribe substantially.

With Ople, a junior data scientist and the email marketing team were able to develop a predictive model for all 50 plus email campaigns that predicted the best time for each individual customer or prospect to receive that campaign within just a few short weeks. Having a model for each email campaign would have taken their data science team months to complete using traditional data science methodologies instead of AutoML.

Ople and Prescriptive Analytics

To take the email example further, Ople can enable prescriptive analytics that serves as a meta-layer over the 50 plus email campaigns. In this case, the retailer has a rule that any individual can only receive six emails in a calendar year. 

To make the most efficient use of these limited opportunities to influence customers and prospects, the retailer can set up a prescriptive analytics layer that takes into account the predictions of each email model to prescribe the best communications for each customer.

For example, if a customer abandoned a shopping cart and the model predicted it was because of price, the prescriptive model could suggest sending a discount offer to the customer for the items in the cart. Or, if a customer had just had a bad experience with customer support, the prescriptive model could suggest they either delay any further emails or they can offer the customer another email with a special gift.

Using the Ople Platform, prescriptive models can be developed with ease using the right information and insights from your predictive models, as well as the knowledge and experience of your team. By accelerating the delivery and deployment of its models, the Ople Platform puts you in a position to transform your business with these two advanced analytics methodologies. To start your transformation, contact us today to start your free trial!