Percolata

Percolata is a Silicon Valley-based predictive analytics company that applies machine learning and deep learning to retail and restaurant operations. Its core offering is a fully managed "Forecast as a Service" product that predicts sales, takeout orders, in-store traffic and occupancy on an hour-by-hour, per-store basis, which customers use to set staffing and delivery levels. Inputs include physical traffic history from security cameras processed with computer vision, point-of-sale data, historical online traffic, marketing calendars, worked-hour data, weather forecasts and other external data such as COVID-19 data. The computer vision layer counts shopper traffic, distinguishes employees from shoppers and identifies individual versus group shoppers before demand forecasts are normalized and converted into staffing level ranges.

A second product line addresses marketing optimization: the company trains predictive models on historical POS and Google Analytics data together with current and planned campaigns to forecast the sales impact of different multi-channel campaign mixes. Outputs include a future campaign simulator and a campaign schedule and budget allocation optimizer, allowing customers to reallocate an unchanged marketing budget toward channels with the greatest measured sales impact. Deliverables are available through analytics visualization, CSV export and API integration with custom applications.

The company states it also applies its predictive analytics technology to power algorithmic trading at Pika.group.

Business model

Percolata sells a managed forecasting service rather than installed hardware or software, taking raw customer data and returning forecasts and staffing recommendations. It states there is no hardware installation, no upfront payment and no package subscription unless desired, with customers paying only for what they use, and it offers a forecast accuracy guarantee under which customers do not pay for forecasts that prove wrong.

Usage-based pricing with no required upfront payment or subscription package, plus an accuracy guarantee whereby inaccurate forecasts are not billed; optional subscription arrangements are available.

Traction

The company reports customer outcomes including 10-20% same-store sales increases, an average return on investment of 150%, a 10% sales uplift for an apparel customer at an unchanged labor budget, $19,000 in monthly labor cost savings for a drugstore customer, and 22% sales increases from marketing budget reallocation, with results visible within one week. Fast Retailing's VP of Finance and Business Development is quoted describing Percolata as its preferred platform for monitoring store traffic, conversion rates and shopper metrics.

Latest developments

As of April 8, 2020 the company ranked in the top 1% of 2,018 contestants in the M5 Forecasting competition on Walmart product unit sales. Its site markets occupancy, staffing and delivery-level forecasting positioned for the COVID-19 period, and notes the extension of its predictive analytics to algorithmic trading at Pika.group.