Legion relies on PredictHQ to help retailers eliminate labor inefficiencies
Legion reduced time spent on manually cleaning and organizing events by 20%.
Use Case
Workforce Optimization
Industry
Retail, Software
Products & Features
Learn more
Introducing Legion
More than 57% of the workforce is hourly in the United States, yet most companies that use hourly workers are using little to no data-driven insight for workforce optimization. There are high rates of turnover, expensive costs to replace workers and rapidly evolving expectations of the new generation of the workforce. The biggest issue is that most businesses aren’t optimizing their labor—leaving significant profit on the table and increasing business risk.
Legion Technologies is revolutionizing the way hourly workers interact with their employers. By using artificial intelligence to help employers reduce costs, Legion eliminates labor inefficiencies and boosts employee engagement.
Companies like Philz Coffee and Barry’s Bootcamp rely on Legion to ensure they are prepared for changes in demand in advance. Legion’s goal is to provide businesses with the best real-time data, automation tools and insights to enable labor optimization and employee engagement.
Legion spent hours cleaning and categorizing event data
Legion’s core product is a machine learning-based solution for demand forecasting, automated scheduling, labor optimization and time/attendance management. Legion computes expected labor demand based on a forecast of sales or traffic at each location, down to 30-minute increments.
Legion produces highly accurate forecasts because its engine incorporates data from multiple sources that influence demand, such as local events, weather, historical sales and promotions. Because it is a machine learning solution, the quality of data is critical. Thomas Joseph, Head of Data Science at Legion puts it this way: “The quality of forecasts is directly related to high-quality data. If data quality is poor, forecasts will be off, which means schedules—fed by forecasts—will be sub-optimal, costing companies a lot of money.”
The Legion team realized the challenge with event data quickly. Other data sets they were working with, such as weather data were clean, time-stamped and had granular details such as temperature, rainfall, and wind speed.
Event data sets were the opposite. They were inconsistent, incomplete and they had a significant number of duplicates that resulted in fake demand signals. Additionally, there were very few attributes in the data on which machine learning algorithms could be applied.
The challenges with event data meant the Legion team was gathering events from a variety of sources, cleaning them and then trying to accurately categorize the events. It was frustrating and arduous work for data scientists, as well as a time suck. Joseph created a homegrown de-duplicator, which pieced varying logic together to solve for these inefficiencies. It improved the data, but there was no way this process would scale as the company needed it to.
With customers across hundreds of cities in the US, scalable, high-quality data is critical to Legion’s business to fuel highly-accurate labor forecasts. Joseph and the Legion team needed to get event data to a suitable form for machine learning in a scalable way to free up data scientists time to work on actual data science.
Legion leverages demand intelligence
Joseph found PredictHQ in late 2016. The Legion team’s criteria was simple: can this company give us useful event data better than we can and ultimately improve our workforce optimization efforts? They made the easy decision to invest in PredictHQ.
PredictHQ’s easy-to-use event API gave Legion direct access to real-world event data, verified from billions of data points every day. Every event is cleansed, filtered, enriched, verified, and then ranked based on predicted impact with proprietary ranking technology.
PredictHQ enabled the Legion team and their data scientists to save significant time by eliminating most manual work so they can focus on the bread and butter of their business—labor efficiency.
PredictHQ has 18 different event categories, along with hundreds of different labels that provide detail about events on a level deeper than Legion had been working with before. Having insight into the event category coupled with more detailed attributes unlocks information about events that Legion never had access to in a scalable fashion.
Results
Legion has been able to reduce the time spent on manually culling events by 20 percent since working with PredictHQ.
Legion reduced time spent on manually cleaning and organizing events by 20%.
Joseph and the entire Legion team have been able to improve their demand forecasting solution and have been able to free up valuable data scientists time to work on more strategic projects.
Legion’s goal to provide accurate labor forecasts to businesses with hourly workers is improved with PredictHQ. Thomas says that Legion labor forecasts are 98 percent accurate and, thanks to machine learning, the forecasts continually get up to date data, enabling organizations to make the best staffing and revenue decisions.
