Personalized retail marketing using in store location

Personalized Retail Marketing Promotions

Using In-Store Location Analytics

And Customer Preferences

OPIM 5671 Data Mining and Business Intelligence

MSBAPM, University of Connecticut

Submitted By: Team 5

Arun Velmurugan, Divya Aggarwal, Monica Ashokan, Neha Agarwal

1. Executive Summary

Retailers are using analytics for a broad variety of purposes. This report showcases a framework that utilizes data from various sources in the retail store and predicts the next probable purchase for the customer. Based on this output, retailers can send personalized marketing promotion offers to customers which will have a better conversion rate as opposed to generalized marketing promotions.

This business study is intriguing because targeting customers with the right promotion message at the right time is important in retail industry. Current analytics techniques focus on analyzing customers’ past shopping patterns to send promotional offers. We believe that this strategy could be enhanced if location data of customers' past travel paths inside the store and that of current visit are used in conjunction. The solution details also talks about how we can further augment this approach by including customers’ online browsing data to better capture their interests.

One major issue with the implementation of any analytical solution is its hefty cost. It can make the retailers want to adopt those analytical applications first that have saved money for others. However, there is little doubt that aggressive adoption of analytics leads to competitive advantage.

1. Introduction

Analytics is fast becoming the answer to retailers’ questions. From store site selection to product shelf placement, analytics can tell everything. Leading retailers like Bloomingdales, American Apparel, Walmart, and Nordstrom have discovered the transformative power of analytics and are utilizing it to make decisions in store operations, merchandising, loss prevention, customer insight analysis and supply chain management. This has made the retail environment highly competitive. New channels like online and mobile are exerting additional pressure. This calls for a need to stay relevant, optimize operations, maximize returns and provide better customer service in order to realize profits.

In this report, we set out to examine how we can use predictive analytics and location analytics in context of in-store retail environment to send personalized marketing messages to customers for product promotions.

1. Problem Statement

The breadth and depth of data is increasing exponentially with every passing day. Retailers have to sift through this data to discern which information will be the most valuable in directing their marketing efforts. Unfortunately, retailers are not as prepared as they should be when it comes to personalizing their offering for every customer. In some cases, they are not even as sophisticated as their customers are.

There also appears a divide between the retailers who know how they can leverage in-store data for sending customized marketing messages vs those that are still struggling with the concept. Having the right products and marketing promotions is challenging enough, but it’s the personalization of these promotion messages that will help the retailers in realizing steady stream of cash inflows.

1. High Level Solution

The below diagram showcases the strategy framework.[pic 1]

1. Solution Details

This section details the data sources and analytical modules that yield actionable insights along with inputs and outputs for each module.

[pic 2]

Data sources:

• In store purchase history data and Current Basket items (captured through the scanner): Customer comes to the store for shopping and scans items that he purchases through the scanner given to him. This data provides information about products relevant to individual customers.
• Web logs: Customer visits the retailer’s website and browses products/items and can add products to his online cart which he may purchase. Data is gathered from the web that includes variables like frequency of product view, time spent on the product, product added to the cart or not etc. This data reveals the interest level of the customer for the product and his preferences.
• Sensor data: Customer enters the retail physical carrying his smart phone and uses free Wi-Fi service. The signals are captured by the sensors positioned at different locations in the store (zones/ aisles). This data not only tells the location of the customer but also the in and out time of the customer at that location. This is used to capture the shopping paths of the customers as well as the dwell times [1]. This data is further analyzed to estimate the next most probable location/path that a customer may follow.

Analytics Modules:

[pic 3]

• Analytics Module 1 - Shopping Path Analysis and Dwell time analysis:

For the path analysis module, we advocate Markov Chain Model in order to statistically predict the probability of visiting a zone/ aisle of product in the future given the previous data of shopping path. Following Markov Model approach, we intend to model the sequences of shopper’s in-store states and movements, which is then used to calculate the transition probability matrix [2] for in-store shopping paths. Based upon the current location of the customer, using this matrix we can predict the most probable path that a customer can take. Dwell time analysis gives us the mean dwell time for each product at customer level which can be exploited while creating the promotion messages.

...