Optimize efficiency, productivity and customer service with highly accurate in-store traffic forecasts
How a large home furnishings retailer used ML to drive high efficiency across their unique customer service model, while simultaneously driving ROI and streamlining operations

INDUSTRY TECHNOLOGY

The client

The client is a fast growing home furnishings retail chain with 20+ stores across the US. In today’s increasingly competitive retail industry, they are committed to making shopping experience easy and convenient, by delivering an unmatched selection, guaranteed low prices, and same day furniture delivery.

Overview

This client selected Google Cloud and Pluto7 to help optimize their store operations by dramatically improving their in-store traffic forecast accuracy. With the right blend of Google Cloud solutions and Pluto7’s expertise, this enterprise successfully leveraged ML, resulting in significant operational efficiencies and increasing profitability.

With over 20 stores spread across multiple states in the US, this furniture retailer relies on a unique customer service model to achieve key growth targets and exceptional customer loyalty. This approach demands flexible staffing across stores, which relies on accurate traffic forecasts.

Challenges

In-store traffic forecasts are currently developed manually by one person, managed on spreadsheets, and require a month of preparation.

The process is time-consuming, challenging to scale, cumbersome to update, and unresponsive to real-time business changes

As a result, stores can be over or under staffed, resulting in a serious impact on customer service productivity and unnecessary overhead costs.

Strategy

Identify, collect, analyze and validate relevant internal and external data sets.

Leverage ML and AI on the selected datasets and produce a highly accurate 4 week (28 day) forecast of foot traffic for each individual store location.

Utilize the results to optimize staff and stock planning to maximize each store’s operating budget and launch effective marketing promotions.

Results

Forecast accuracy increased to +84% from previous max of ~65%

Improved operational productivity and customer service by scheduling staff more aligned with realistic demand (store foot traffic)

Reduced costs, optimized staff utilization and superior customer satisfaction

“Forecasting our store traffic closer to real count serves as a critical element in extending our store operations productivity and minimizing our costs.”

Key Stakeholder,
Large Furniture Retailer

This enterprise has grown by designing and delivering high quality furniture at affordable prices by targeting a diverse group of quality conscious customers. Their business has thrived due to its well-knit, family-like operating culture and ability to deliver exceptional customer experience. Employees take pride in their company’s extensive product offerings and the brand image it has maintained. The client’s core value is about showcasing high end product design, offering a wide range of selection and providing best in class customer care. A key element to this is being able to detect customer buying patterns.

The key to their growth is to maintain their differentiated customer service model at scale. To do this, they need accurate in-store foot traffic predictions so they can allocate their resources effectively, stock their individual stores appropriately, and ensure high levels of operational efficiency. However, their current manual processes have been reliant on one person and one spreadsheet, limiting accuracy, growth and responsiveness.

What does it take to have a highly accurate store forecast?

Accurately forecasting store traffic is a primary step to driving success in many business areas. Highly effective store planning, optimizing customer experience, and maximizing profitability all rely on understanding who is most likely to come into the store, how many and when.

There is a wide range of data sources that feed our desired predictions. Historical data provides insight into customer behaviors - such as past orders and traffic patterns across the year. Marketing campaigns and how they correlate to store traffic is another helpful comparison. In a multi-store environment, regional differences bring complexities but also result in a wider data set. In order to leverage the potential of this internal data, it must be accessible and consumable.

However, internal data is only a portion of the whole picture - a more comprehensive forecast will take into account external factors. For example - does the weather have an impact on store traffic, if so - it should be included in the analysis. What about customer clicks to Google Maps and their correlation to store proximity? A truly effective ML model takes into account both internal and external data, simultaneously processing, analyzing and correlating it in real time.

The perfect problem for Google Cloud to solve

This involves leveraging advanced predictive analytics capabilities like those provided by Google Cloud Artificial Intelligence and Machine Learning offerings. This is a perfect problem for the Google Cloud Platform to resolve - to handle complexities with data storage, scale of analytics and associated machine learning and artificial intelligence.

Currently, it takes one person about 15-20 days to come up with a traffic forecast for one store, which isn’t scalable. We need to eliminate this dependency and develop the ability to accurately forecast store traffic more granularly at an individual store level. This enables our store operations team to perform at high levels and succeed.

Key Stakeholder,

Large Furniture Retailer

Improved predictions enhance decision making

With these strategic objectives in mind, this large furniture retailer partnered with Pluto7 to leverage Google Cloud Platform to help create a dynamic store traffic forecast model with high accuracy.

With Pluto7’s expertise in smart analytics, machine learning and artificial intelligence, we were able to identify the appropriate Google Cloud solutions and quickly develop a prototype that demonstrated how blending both internal and external data produces a high accuracy, multi-store traffic forecast that achieved our purpose.

This retailer now accesses highly accurate 28-day forecasts, enabling them to plan more effectively. They have since experienced significant operational efficiency and productivity gains, while continuing to delight customers with their unique and differentiated model for customer service.

Rapid growth through innovation

After a comprehensive analysis of available tools and technologies, the customer engaged Pluto7 to leverage Google Cloud's smart analytics and machine learning platform. Given the results achieved, they’re now accelerating business transformation by embedding Google Cloud’s exponential technologies across multiple use cases and business functions.

Why Google

This client chose Google Cloud Platform because it:

  • Enabled an easy way to collect multiple data sources and co-relate them all together in the most cost effective and productive manner
  • Analyzed the combined data elements to generate a high accuracy forecast by leveraging machine learning capabilities of the Google Cloud Platform
  • Allowed the client to focus on solving their business problem without getting bogged down with infrastructure issues

Products used

Google Cloud Platform

Google Cloud Dataprep

Google Cloud Dataflow

Google BigQuery

Google Cloud Storage

Google Machine Learning

Google Tensorflow