AI/ML at scale: The next horizon for PPG’s data strategy

On balancing standardization and flexibility:

Jeff Lipniskis: [A]s you look at a transformation built around acquisition, you have a lot of infrastructure diversity, different ERP platforms, a more complex application portfolio.  And most importantly, a lot of variation in business process, as you bring these organizations together.  And it is at that business process level where data intersects, where our data is generated, where it’s managed.  So we, as an organization, are spending a lot of time focusing on standardization.

And if I continue on that journey, to think about how do we optimize that supply base as we bring organizations together?  How do we optimize our manufacturing and lab footprint and consolidate that and have it at the right size?  How do we create a customer experience that never feels like you’re doing business with sixty companies that came through acquisition, but you deal with one PPG.  And data is a key part of that, that drives that experience.

But, at the end of the day, we need to be flexible on the IT side, to be sure we’re hitting the mark on these business outcomes.

On good governance:

The core for us starts around governance and governance globally, having a good master data management and data enrichment program and process standardization and continuing to evolve that. 

Then we looked at the next pillar of that strategy, which is around that data architecture development, getting to a common view of data and definition, while you have sources across disparate systems.  How do you tie that together, and across multiple lines of business? 

On the right tool(s) for the job:

We, as a Microsoft client within Azure, we’ve worked with Microsoft toolsets around reporting and analytics and so forth.  But then, as you move to the AI/ML world, what you will see, from our perspective anyway, is you’ll start to see some variation, because it becomes more fit for purpose than one size fits all.  I mean it’s about what is the most ready model or most ready tool.  And that can get you into a multitude of different suppliers and sometimes you’re connecting multiple clouds here, multiple solutions to build out that model or capability. 

On investing in data architecture:

I would be the first to say we came from a very low level of maturity, and we arrived at the current architecture by bringing in outside expertise to assist us.  Our architecture continues to evolve.  We are building internal talents and capabilities today, so we’re continuing to keep that architecture current and grow it.  I will say if we could turn back the clock, investing more up front in architecture would have helped us in the long term.

On data readiness:

We’re really focusing on all new system implementations, to be sure we’re not creating more data and more legacy that’s not AI ready.  And you need to create data quality metrics, you need to do audits, and validate from day one, that even if you aren’t putting the data in a model, it will get you to where you want to go, in one year, two years, three years, so you don’t get a bad surprise down the road. 

On what’s next for PPG’s data strategy For us it’s going to be one of continuing to mature the foundation that we have in place, and I think we have a good base to build upon.  And building upon this by learning and adapting and continuing to be flexible.  But if I look on that horizon, we will definitely increase the focus and see more impact from AI and machine learning, and we’ll see that continuing to grow rapidly, if not, I’ll even use the word “exponentially,” as we have more data readiness.  I think really that’s the next horizon is AI/ML at scale.