Anheuser-Busch InBev
We worked in tandem with ZX Ventures, the innovation branch of Anheuser-Busch InBev, the world's largest brewery, on this project for their portfolio of bars and restaurants in Latin America.
Our mission was to grow sales at Anheuser-Busch InBev bars and restaurants across Latin America by giving their managers immediately actionable data driven insights in real-time.
We created a mobile app to convey meaningful information to managers for them to use to incentivize their team through friendly competition. These goal-oriented, competitive sales recommendations are powered by machine learning and predictive analytics derived from millions and millions of sales transactions.
Defining Success
We were brought on early, during the initial idea phase to create simple prototypes and to help validate preliminary prototypes. Once validated through user testing and to ensure the business result, we created a roadmap to build and launch a full scale architecture, focusing on a narrowly defined MVP to maximize speed to market.
While we love building sophisticated, modern software platforms, we can’t take pride in a delivery until it achieves the business result for the customer. This only happens when we create a positive experience for the people using our product that ensures a real-world business outcome. In other words, we don’t consider this project a success just based on the quality of the software, but also on its ability to increase the staff members’ enjoyment of their work and the resulting increase in revenue for the bars.
An App for the People
Bar managers and waitstaff are on their feet all day long, so if you want to “meet them where they’re at,” you need to do so on-the-go. That’s why we chose to develop the user-facing portions of this project on iOS and Android devices. In addition, we spent time to truly understand the life and routines of our users and this is reflected in the product design.
At the beginning of their shift, the bar manager is given exactly the information needed to plan his or her day. This includes a forecast for the day’s sales and recommendations for what the waitstaff should focus on selling if they want to over-achieve versus that sales forecast. We deliver recommendations based on machine learning algorithms (more on that later) that focus on creating friendly competition amongst the team to drive motivation and engagement and ultimately increase sales.
Having access to both current and historical performance data is critical for managers to provide in-the-moment guidance to their staff. To achieve this, sales data throughout the app is updated from Point of Sale systems in near real-time, with aggregated daily and monthly performance information available instantly in an intuitive display to support the conversation.
Data. Lots of Data.
To power both the app’s real-time data visualizations and the machine learning models, we ingest millions of transactions from a variety of Point of Sale (POS) systems and create a normalized data set. Each POS vendor provides a different mechanism for retrieving data, so a multi-step ETL process is used to retrieve, store, and transform the data into usable form. A serverless, microservice model ensures utmost scalability so users never experience a delay. As new customers are brought on, the turnkey system automatically imports all available historical sales data and provides it to the forecasting model.
Another feature of the cloud-based architecture we designed to store the incoming sales data is the ability for enterprise customers to easily access it. The granular sales data is readily available for business and marketing analytics as well as Business Intelligence dashboards and reports through any data visualization tool the client wants to use.
"The team at Beta Acid provided both incredible technical solutions as well as an insatiable desire for solving complex domain problems, all while keeping the experience of our users as their top priority"
Brian Reece, Global Director Retail Digital
Machine Learning
For this product, a great user experience is useless without actionable data and insights to power it. Luckily, we have tens of millions of transactions we can use to deliver forecasts and recommendations with a high level of accuracy.
Using line-item level granular sales data, we developed algorithms to cluster bars with similar behavior and look for patterns to determine what products provide the most value on specific days and times. We then relate those insights to an individual bar’s menu to create daily recommendations for what management should focus their team on in order to maximize revenue for that day and even for specific shifts.
AWS
React.js
Node.js
Prophet
Lambda
Measuring Results
Our startup-like approach included releasing an MVP version of the product as soon as possible and with just a few customers in Argentina. With positive feedback, and a few iterations, we were able to roll out to dozens of customers and hundreds of locations across Latin America.
Within 2 months of coming onboard, bars see a clear 20% increase in top line revenue driven by increases in average order value. In addition, staff (managers and waitstaff) are more engaged and excited to perform their duties.
Get in touch
Whether your plans are big or small, together, we'll get it done.
Let's get a conversation going. Shoot an email over to projects@betaacid.co, or do things the old fashioned way and fill out the handy dandy form below.
Get Buzzed
Visit our privacy policy to learn how we collect, keep, and process your private information in accordance with these laws.