Using descriptive, predictive, and prescriptive analytics frameworks to ask the right questions in manufacturing
You’ve seen the warnings time and again:
- “Data is Worthless if you Don’t Communicate It”
- “Good Data Won’t Guarantee Good Decisions”
- “Big Data: Too Many Answers, Not Enough Questions”
The list goes on.
These words of caution are justified. According to Chris Brahn in his "Last Mile of Analytics,” 70 percent of enterprises view advanced analytics as a critical strategic priority, but only 10 percent believe they're recognizing the full potential of their business data.
The disconnect between data and business value often is due to a failure to identify and ask the right questions.
So how can you craft meaningful questions that drive business insight? We suggest an approach that frames inquiries around three types of analytics: descriptive, predictive, and prescriptive.
Descriptive analytics answer the question “What has happened so far?”. These queries expose past trends and analyze historical data to help you better understand your business.
Most manufacturers are making use of some type of descriptive analytics. We often see these take the form of KPIs (key performance indicators) and dashboards used to monitor business processes. Descriptive analytics frameworks are the easiest to implement and can be powerful when applied to key areas of your company.
Descriptive analytics can answer questions such as
- What is my inventory turnover over the past year?
- What is my percentage reduction in machine downtime over the past two years?
- What is my order to shipment time over the past six months?
- What are my year to date sales compared to last year?
Descriptive analytics can help you understand your operations and set baselines for improvement.
Predictive analytics answer the question “What could happen?”. These lines of inquiry can help you anticipate likely events, mitigate potential risk, and assess future opportunities. This type of analysis is more difficult to implement than descriptive analytics as it often requires more advanced techniques such as artificial intelligence or deep learning models. However, a successful predictive analytics strategy can give manufacturers a major competitive advantage. Predictive analytics can help manufacturers answer tough questions such as
- How many parts should I order to meet next month’s demand?
- When is my equipment likely to go down? How should I schedule future maintenance?
- How many employees will we need for service calls next quarter?
Predictive analytics can also be applied to your service offering to provide greater value to your customers. Giving your customers insight into their investments, helping them prevent costly product failures, and anticipating their future needs are all smart strategies to address the “Amazon effect” that now drives B2B customer expectations.
Prescriptive analytics answer the question “What should happen?”. The purpose of prescriptive analytics is to determine the best course of action for your business based on insight from predictive models. Prescriptive analytics often answer questions around optimization and can help a company determine which process improvements will yield the biggest return. Prescriptive analytics can be used in manufacturing to answer questions such as
- What mix of products should we buy to meet demand while taking advantage of bulk discounts?
- What will happen to profits if we increase prices? Decrease prices?
- What will happen to our sales if we add an additional salesman to a specific territory?
No company has unlimited resources, so prescriptive analytics models can help you prioritize your investments of time, money, and energy.
Tying It All Together
To utilize these analytics frameworks together, begin with descriptive analytics by deciding on your KPIs. These should be small in scope and answer very specific questions about each key area of your business. After monitoring your KPIs, it will be easier to see the areas of your business that will benefit most from predictive analytics. We often recommend focusing on those areas where you are underperforming based on the information gained from your descriptive analytics. When you have your predictive models, you can then apply prescriptive analytics to optimize the processes.
A Real-Life Example
For example, let’s say your company has a repair shop. You might begin by setting up descriptive analytics to tell you the average turnaround time of repairs and the amount of time it takes to complete each stage of the repair.
Based on this analysis, you realize that your turnaround time is slower than expected because it is taking too much time to order and receive the correct parts for the repair.
Now your predictive analytics strategy has become obvious. Predictive analytics can help you forecast the parts you will need for future repairs so that you can get the parts on the shelf before they are needed for an incoming repair.
After you have a predictive model, you can use prescriptive analytics to optimize this process by finding the optimal number of parts to order to take advantage of bulk discounts and reduced shipping costs while keeping inventory overhead to a minimum.
Ready to put your data to work? Get in touch! Our data scientists can help you build an analytics strategy tailored to your business. We leverage the latest tools including Salesforce’s Einstein Analytics.