ADDING INTELLIGENCE TO DINING: RESTAURANT BUSINESS Next Case Study

Client Organization:

One of the largest dining restaurant chains in the USA.

Project Owner:

VP Marketing & Strategy

The Problem:

With increase in unstructured data (text, multimedia etc.,) across all business, there is a need for an efficient, scalable way to store & retrieve data. SQL architecture proves to be expensive and not scalable, whereas NoSQL scales well and is efficient to retrieve and manipulate unstructured data. Consequently, a noticeable drift from SQL to NoSQLis evident. Likewise, our customer used MongoDB (a widely usedNoSQL document database) to store large amounts of restaurants data. The customer wanted to analyze data, the way they did with SQL databases but NoSQL databases support only a limited type of queries and not all useful for building efficient predictive models.

The Solution:

Most predictive algorithms accept data in the form of relational tables as input. Although NoSQL is efficient to store and retrieve data, building predictive on NoSQL data has always been a challenging job. Hence, an essential first step to build predictive models on NoSQL data is to transform / export just the required data from NoSQL to a tabular format. We designed a generic, flexible plug-in to convert NoSQL to SQL (Our solution works with any NoSQL data at any level of complexity, independent of the domain and schema). Then we built a few predictive algorithms to work seamlessly with the data. The complete design and modeling helped the client in segmenting their customers and the design was flexible to accommodate new updates as per business requirement.

Tools & Technologies:

Python, R, MongoDB, HBase, Hadoop, Map-Reduce.