Helping a global plastics manufacturer reduce its use of a pricey, acidic ingredient by 20%
Helping a global plastics manufacturer reduce its use of a pricey, acidic ingredient by 20%
A global plastics manufacturer was looking to reduce its consumption of a high-cost ingredient called acetic acid. In just six weeks, Bain helped them implement a new, machine learning-powered process to optimize yields. Bain then handed over that solution to ChemicalsCo so its teams could continue to use it.
These teams helped the client optimize its yield
Consultants conducted the initial due diligence and confirmed this project was largely a question of data science—if the Bain team could gather the right data from the client’s locations, they could create an algorithm to optimize yields.
These teams helped the client optimize its yield
Consultants conducted the initial due diligence and confirmed this project was largely a question of data science—if the Bain team could gather the right data from the client’s locations, they could create an algorithm to optimize yields.
-
Advanced Analytics Group (AAG)
AAG’s data scientists used a combination of machine learning and engineering knowledge to identify the key factors in acetic acid loss. This allowed Bain to train a simple, transparent algorithm to identify the factors behind periods of low and high consumption, and thus resolve the problem.
-
Expert Consulting
Bain chemicals experts advised on chemical reactions at the manufacturing plants. They also provided guidance on connecting the Bain team’s analyses to the actual model output.
-
General Consulting
Consultants conducted the initial diligence to scope the project, coordinated with the client’s leadership, assessed the reliability of the product, and more.
-
Product, Practice, and Knowledge (PPK)
The knowledge team helped scope the engagement and shared examples from Bain’s prior chemicals work. They provided fresh market trends, topical insights, and unique perspectives to help shape the approach. After the case, they captured the full breadth of work and codified it for future case teams.
Background
In the years leading up to this engagement, ChemicalsCo’s consumption of acetic acid had risen. This ingredient has become one of their largest variable costs and its unpredictability was hurting the company’s profitability. As leadership looked to grow margins, they hired Bain, which assembled a case team of consultants, data scientists, chemicals experts, and process engineers.
ChemicalsCo’s engineers were already aware that whenever the company used more acetic acid, the quality of its chemical products tended to be lower. Meaning, these two factors were inversely correlated. The more they could reduce acetic acid use, the better. But beyond that relationship, understanding the specific parameters that caused the acid use fluctuations was a challenge.
The plan
Bain addressed the acetic acid question as part of what's known as a “full potential” initiative—analyzing the client’s entire company to determine what it ought to be able to achieve, and the levers it could pull to get there. This required looking at ChemicalsCo’s operations which spanned the US, Europe, the Middle East, Africa, and Asia-Pacific, and searching for key acid use variables and potential predictors. If Bain’s data scientists and experts could isolate those variables, they could build a case for developing a machine learning model to turn the problem into an equation.
Bain teams learned that one manufacturing site was using much more acetic acid than others, and the spike in usage had occurred right after the site had expanded. They visited to understand why.
-
What was ChemicalsCo’s starting point?
-
What was the full potential the company could achieve if the problem was fixed?
-
What benchmarks and historical data already existed?
-
What variables and predictors were correlated with increased acid use?
-
How could the client scale a local ML approach across many sites?
-
What was the most efficient path forward?
The approach
Bain’s investigation into that high-usage site supplied the variables the data science team needed. They applied a combination of machine learning and engineering knowledge to train a simple, transparent algorithm to identify the factors that caused periods of low and high acid consumption. This, however, proved more complicated than it at first appeared.
The team had to clean up and standardize the site’s data, build a baseline model with all variables, construct a causal model, and run t-tests to ensure those measures were statistically significant. Once the model was working, they prepared it for handoff so the client teams could use it themselves.
The insights from that model were enough for ChemicalsCo to immediately act upon. Usage at the problem site fell and the company was able to apply that formula to sites around the world.
The results
Within just four weeks, Bain teams solved the mysterious, persistent, and expensive use of acetic acid. The client was able to reduce its global use of that ingredient by 20%, for significant cost savings and the higher margins leadership had hoped for.
consumption of the costly ingredient, acetic acid
additional machine learning use cases identified at the same site