Q&A with Lucas Gorganchian (Prisma): “AI bridges the gap between analysis and decision making”
How can retailers use AI to improve their day-to-day operations? In our conversation with Lucas Gorganchian, CEO of Prisma, we explored the opportunities that AI presents to streamline operations and spoke about the challenges of applying data science in the c-store market.
Lucas Gorganchian defines himself as a Serial Tech Entrepreneur, having founded six companies over the course of his 22+ years career, mostly dedicated to providing software to large enterprises. After earning a master's degree in Innovation and Global Leadership at MIT, he founded Prisma alongside Damián Barletta, Co-Founder and Chief Technology Officer, by the end of 2016. The company offers retailers a unified planning platform that analyzes, predicts, and suggests actions, covering key variables of business optimization such as pricing, promotions, assortment, exhibition, and supply.
Q. What benefits does AI predictive analytics bring to retail data science?
A. There are many transactional, data analysis, visualization, and business intelligence tools, but there are very few that effectively bridge the gap between information analysis and decision-making. This function comprises two key components, the first of which involves aligning with the marketing strategies of a company, often governed by specific business rules or algorithms. In these cases, even when AI suggests a price change, the decision is rooted in a pricing strategy that may vary across brands. The second component entails integrating these business strategies with machine learning models or more sophisticated solutions. One single solution on its own isn’t enough, it may be sexy from a technological standpoint, but it alone cannot achieve overarching business goals. A convenience store chain can leverage this type of tool even with little data, without the need to spend three years improving a data set. The reality is that revenue is generally higher during the first few months, it is a curve of diminishing returns where there is a lot to do with little data. On the other end, with more sophisticated models and lots of data, the line generally gets thinner.
Q. How important is it to have access to vast amounts of data for decision-making?
At Prisma, our mission is to bridge the gap between information analysis and decision-making processes. In a recent case, a client approached us seeking the optimal price for a Coke. When confronted with such a scenario, the initial step is to analyze a historical dataset to train a model. Upon reviewing the historical data, it became evident that the price for a Coke had remained relatively stable over the years, with minimal variation aside from inflation adjustments. Consequently, the efficacy of machine learning models heavily relies on the quality and diversity of available data. Without sufficient and varied historical data, even the most sophisticated models are bound to yield low accuracy levels, thereby diminishing confidence in decision-making processes. AI use cases, whether in retail, convenience stores, or fuel pricing, are not without their limitations. These limitations stem from the availability and diversity of data, which directly impact the effectiveness of AI-driven solutions in making informed decisions.
Q. How did Prisma originate based on this context?
A. Prisma’s inception traces back almost 20 years ago, coinciding with the advent of companies adopting management modules like SAP or CRMs to streamline their operations. During this transition, despite the abundance of data at their disposal, executives continued to rely on intuition rather than data-driven decision-making. The primary challenge of the era lay in fostering a cultural shift and providing adequate training to empower professionals to base their decisions on data. In collaboration with Claudio Reboredo, who served as the Director of Retail for ExxonMobil Latin America at the time, we devised a business simulation game aimed at training these professionals. The simulation involved five individuals assuming control of a virtual convenience store chain, competing against other teams by simulating the day-to-day decision-making processes of such outlets. These decisions encompassed product selection, inventory management, promotional strategies, pricing, store layout, product display, staffing levels, and salary considerations. Executives from diverse retail backgrounds worldwide participated and recognized the necessity for such tools in their daily operations.
It was during this period that I noticed the inadequacy of existing CRM and business intelligence software in addressing this specific need. I researched alternative platforms and found out that, while some solutions were emerging to fulfill certain functions, they were predominantly patchwork solutions, lacking integration, and coherence. Recognizing this gap, I seized the opportunity to create Prisma, a holistic solution tailored to address the intricate processes and challenges faced by retailers in their day-to-day operations. Our goal is to make the volume of data transparent to the user. This implies capturing extensive data from various sources, processing it, aligning it with predefined business strategies, and consistently providing actionable suggestions to the user.
Q. Are there any issues if a retailer relies too heavily on data for store management and customer experience?
A. Firstly, it is essential to pose several questions for each use case. One critical inquiry is whether we are measuring a valuable insight. If not, we must evaluate the cost associated with initiating measurement. As part of our product suite, we offer an app designed to capture data on pricing, display, service levels, and other relevant attributes. The subsequent step involves pondering what actions can be derived from this captured information and how it influences pricing and demand. If the analysis reveals that the cost of capturing the data outweighs the potential return on investment (ROI) or the benefits derived from it, then it is often not justifiable to proceed with data capture.
Q. What about store automation?
A. I often cite an example of a technology I saw in San Francisco several years ago. It involved a company offering robots-as-a-service that traversed store aisles, capturing images to identify stockouts. The reality is that there are far less expensive solutions than hiring a robot to do that. If a robot not only identifies stockouts but also assists in replenishing missing products, it could potentially replace human labor. Yet, if its function is solely to detect stockouts, there are significantly more cost-effective methods to achieve the same outcome. This pragmatic approach guides our mindset. While we operate as a tech company, we prioritize the application of common sense to deliver straightforward and cost-efficient solutions leveraging technology. The most important thing for operators and c-store chains is to understand why their customers choose them. If the primary consideration for customers is quick in-and-out shopping experiences, it may be worthwhile to explore smart shopping solutions. If that is not the answer, investing in such technology may currently outweigh the returns it can provide.
Q. How does automation influence the customer experience?
A. The first time I visited an Amazon Go store over five years ago in New York, there were two people assisting customers entering, I had to update my app, and when I entered there were many products in breakage. The feasibility and return on investment (ROI) of implementing technologies like 20 cameras and sensors must be carefully weighed against potential labor savings. And the result of this is different in the U.S. than in Europe, Latin America, or Asia. In addition, companies like Amazon often prioritize pushing technology and selling it rather than genuinely embracing the underlying concept, driven by innovation trends and hype.
Retail chains may adopt different formats and offers tailored to specific geographies and customer segments. The evolution of smart store infrastructure costs will also play a crucial role in shaping future strategies. Ultimately, each chain must ascertain why customers choose them and adapt accordingly. The issue of charging times is another key element in these strategies when considering the transition to electric vehicles (EV). Currently, charging an EV can take from 30 minutes to an hour. While technology is expected to improve charging times, there remains a need for amenities to occupy customers during this period. What will they need? A smart store, a meeting room, a restaurant, or a gym?
Q. What are some of the limitations and challenges for data science and AI within the retail sector?
A. This is directly tied to data complexity. While a tool may suggest an optimal price for a product, there are numerous variables to consider: How does customer service quality impact sales? What about checkout speed? Availability of stock? Store display quality? These factors, both measured and unmeasured, influence demand, highlighting the value of a holistic approach to enhance the accuracy of predictive models. Our strategy is to guide clients toward maturity in their data science endeavors. I often use the metaphor of aiming for the Moon – achieving our goals requires step-by-step progress. We must first hire the astronaut, purchase their suit, assemble the rocket, and fuel it. Similarly, AI processes require training and maturation. Some users expect instant solutions at the push of a button, but the reality is that AI development involves iterative refinement and construction. It is essential to identify specific use cases where AI can genuinely enhance productivity or efficiency.
Related contents
Discuss