On a recent Parker Avery “Talk Retail to Me” podcast episode, Chief Analytics Officer Sam Iosevich discussed Parker Avery’s advanced analytics solution development efforts with the firm’s Director of Solutions Architecture, Leo Greyz. Leo is enhancing Parker Avery’s “Results as a Service” (RaaS) offering and bringing it to next level as a “Software as a Service” (SaaS) product for retailers and CPG companies.

In this blog post, we provide highlights of their conversation, including Leo’s recommendations for retailers and CPGs who are moving forward with their advanced analytics initiatives.

Clean Data & External Sources

As analytics grows and initiatives gain momentum, companies will continue to realize that clean data is critical to successful outcomes. Cleansing the data and making sure the hierarchies make sense for the business are also key focus points. Parker Avery has been helping clients achieve this data cleanliness—including data governance—for quite some time.

Bringing in external data sources by leveraging partnerships is also key. This entails combining clean, internal data with external, more predictive data across all retail and consumer brand segments. Knowing the proper data to bring in is highly important to creating meaningful results for analytics initiatives.

Data Sharing Methods

There are many ways of sharing data, but this could be done easier and more standardized. Some companies are creating different data sharing approaches, but in the future, it will be more standardized. As a result, it will be lot easier to share both internally—between different business functions—and externally to partners, consultants, etc.

Right now, there are differing methods, for example, SSTP and Azure, that use data formats such as Parquet and CSV. As such, Parker Avery is approaching development and architecture by having a place to stage the data. But we are also designing the solution to be flexible enough to enable different connectors for different services. As an example, if one client uses Amazon’s S3 buckets, we built the connector that other clients can leverage. If it’s something new, we’ll build a new connector. Without a standardized approach, multiple connectors are necessary today. If the industry gets to a standardized way of sharing data, we will only need to use specialized connectors for any legacy systems still in use.

GPU-Powered Analytics

One of the most exciting elements on the advanced analytics solution horizon is graphic processing unit (GPU)-powered analytics. GPU has the potential to make analytics processing much faster and much closer to real time. Parker Avery has partnered with professors from the Math and Data Science department at the University of Rochester on our AI/neural net demand analytics offering. As we continue to collaboratively develop that offering, we will increasingly move into the GPU realm as well.

To put GPU in simple terms, it’s the number of different math operations processed simultaneously. GPU can do thousands at the same time. Originally GPU was made to quickly determine what the different colors of pixels on the screen should be for 3D-rendered objects. Therefore, it has to do a lot of work in parallel. Now, with a lot of data, a neural network or any other GPU-friendly algorithm, you can do so many computations in parallel—it’s going to be a game-changer.

GPU enables this by doing simple mathematical operations, like multiplication, incredibly quickly. If you look at the neural net or an AI architecture, in very simple terms, it’s a whole lot of multiplications. This is something that GPU is set up to do very quickly, and it’s the way of the future. In the past, in Parker Avery’s demand analytics platform we’ve thought about how to process hundreds of millions, if not billions, of geography/product/customer combinations—and how to do this efficiently, possibly with shortcuts. With the world of GPU computing, these calculations can be incredibly faster, without shortcuts.

Advanced Analytics Solution Recommendations

Focus on the Business.

Too often we see companies focusing on the actual algorithm, like neural nets or decision-tree-based ensemble models. Instead, the focus should be on how the data is going to be transformed before the model. They also need to comprehend how results of advanced analytics are going to be surfaced and used to the business users. Further, it’s how to make those results actionable—it’s the actionability that needs to be emphasized because that is the key to delivering real business results. If you gain insights on data but it’s not actionable, then you’ve spent time and resources on a problem that you haven’t solved.

Invest in a Business Intelligence Tool.

As your data grows, it will be a lot easier to visualize what the data looks like if you have a business intelligence (BI) tool. As opposed to extracting data from a database into Excel and making graphs from there, using a BI tool is much more effective.

Parker Avery is moving very quickly to expand offerings that move the needle on how retailers and CPG companies approach and use advanced analytics. Leo’s development efforts, combined with the strength and depth of the firm’s retail and CPG experience, will position our clients far ahead of their peers in their advanced analytics initiatives.

To listen to the podcast or read additional Parker Avery insights, please visit our insights page.

Learn More
Contact Parker Avery