Machine learning vs. rules-based systems, explained

Q&A with TechTarget Search Enterprise AI Joseph Carew sits down with Catalytic's VP of Product Jeff Grisenthwaite

The term machine learning gets thrown around quite a bit, but so often companies looking to streamline their operations with data-driven intelligence and better decision-making are unclear as to what machine learning entails, how it differs from rules-based systems, and what’s best for their company.

In this Q&A with TechTarget, VP of Product at Catalytic Jeff Grisenthwaite talks through and compares different rules-based and machine learning applications.

Grisenthwaite answers questions about what rules-based systems are, how they've been used, how machine learning differs, and key advantages.

 

Will you walk me through your background?

As VP of Product at Catalytic, I focus on the current and future needs of our customers and ensure we’re delivering a product that will enable them to achieve their goals to transform their operations. This includes the automation of processes with both rules-based systems and machine learning. Prior to Catalytic, I’ve been in the enterprise software space for 15 years at CEB (now Gartner) and at KnowledgeAdvisors, an HR analytics provider (acquired by CEB). Germaine to this topic, I studied machine learning under Andrew Ng in a Stanford course.

 

What defines machine learning? What is the definition in your words?

All computer programs have goals, but traditionally, people have been required to explicitly code the rules to achieve those goals. With machine learning, the computer programs can figure out for themselves how to best achieve those goals and can self-sufficiently improve as they intake more data and experience the results of differing scenarios.

 

What does a rules-based approach look like? How does this differ from machine learning?

With rules-based systems, people define the logic for how the programs make decisions. For example, a rules-based approach in recruiting could have a set of rules, such as “If the job candidate has less than 5 years experience, then disqualify the job candidate for this role.” If approaching the same situation with machine learning, no person would explicitly define that rule. Instead, a person could enable a program to learn for itself, by feeding the program a large set of training data that includes the circumstances when job candidates were qualified vs. disqualified. The program would identify patterns and apply its judgment to new data that comes in, determining a priority ranking of the incoming job candidates.

 

In what circumstances does one make more sense than the other? What are some applications of both?

Machine learning requires a lot of data, typically thousands of records, in order to make accurate predictions, so it’s only applicable for high volume use cases, such as sales lead qualification or customer support auto-responses. Machine learning is also better suited to situations that have a large number of factors (columns in the data set), since it is better equipped to identify patterns in the data than asking people to both find the patterns and manually develop rules for each of them. An example of this would be predicting real estate prices, in which an algorithm can review historical sales prices and evaluate all the different factors, such as location, square footage, amenities, etc. Finally, machine learning beats out rules-based systems in rapidly changing environments, such as e-commerce recommendations and sales forecasting.

Rules-based systems are best suited to situations in which there are lower volumes of data and the rules are relatively simple. Many companies use rules-based systems for expense approvals, defining the dollar thresholds that require management approvals at various levels. Another example of a rules-based system would be email routing that uses a list of keywords to determine the destination.

 

What are the differences and advantages of both?

Both machine learning and rules-based systems aim to automate decisions with high degrees of accuracy. Rules-based systems consistently output exactly what has been programmed into them, which is an advantage in situations where the incoming data is relatively stable, but a disadvantage in a variable environment. Machine learning algorithms often have much lower accuracy than rules-based systems until they learn from enough data.

This does not need to be an either/or choice, though. The most accurate solutions often combine rules-based systems with machine learning. As an example, one of our customers in the advertising sector combines both to automatically draft responses to Requests for Proposals (RFPs). A rules-based system is leveraged to filter a library of answers to prior RFP questions to those that are relevant to the given RFP (based on country, industry, and other factors). Then a machine learning algorithm is used to predict the best answer to each question from within that filtered library.

Combining rules-based systems with machine learning enables each approach to make up for the shortcomings of the other. Rules-based systems can make the obvious decisions, provide guardrails to avoid undesirable outcomes, and can deliver accurate decisions before any data is gathered. Machine learning can augment these systems by improving accuracy over time and responding to changes in the environment.

 

About Catalytic

Catalytic is the only technology built from the ground up with all of these capabilities in one no-code, cloud platform—giving you the most seamless way to create smart workflows using the latest digital, automation, and AI technologies. Using Catalytic’s user-friendly framework, citizen developers can easily build enterprise-grade automated processes that connect systems, data, and people. The result is a faster, leaner, and fully digitized business that can execute higher volumes of work with less time, cost, and risk.

To discover what the true value of automation can be, try our new business impact assessment to get an instant estimate of the benefits of improving your workflows with Catalytic.

 

Learn more

Are you ready to take the first step in your automation journey? Start your 14-day free trial to experience how when using Catalytic for automation, the possibilities are endless—and scalable.

 

YOUR NEXT READ: 

SIX OBSTACLES TO SCALING AUTOMATION 

Learn how to identify and overcome the obstacles to scaling your automation program. 

Read the eBook

Written by Catalytic