The project applies an analytical approach to identify customer segments and enabled RMS to comment on how specific groups are responding to any change. The complex part of solving the business problem was analyzing 12 Million restaurant customer’s Transaction Level Data (TLD) which contains an itemized list of items that each customer bought. The technical objective of this project is to use Text Mining and Cluster Analysis on the item names of each order and identify customer segments and how they evolve over time. Analyzing and applying our acquired knowledge of machine learning algorithms on big data was our second motivation. We approached our critical business problems by analyzing the requirement, understanding the customer pain, analyzing customer gains, and building our value proposition. This project’s value proposition refers to the value that we promised to deliver to Revenue Management Solutions (RMS) that will solve the business issue that currently they are facing. Our project when applied will add value to the overall business by automating the analytical process which now is done manually and the same model can be applied to massive amounts of big data and various types of data. The project will solve the most challenging question that how treatment effect can be measured on customer clusters. The project will help RMS to automate the process of customer segmentation and clustering based on items the customer ordered. The process is currently done manually costing time and money to RMS’s clients and delaying the delivery of the business. The project model can process and analyze millions of customer TLD data and will suggest the best number of customer clusters computing statistical results which will be then used for treatment effects and behavior analysis of those clusters.
Following are the four important value propositions of our project:
- Analytical Solution Using Machine Learning Algorithms
The main challenge of our business model was to process millions of customer TLD data. The data was provided by RMS and it consisted of 7 different restaurant categories. And each category consisted of 10-15 restaurant’s billing data over a period of a year. There was a total of 12 Million observations of customer orders starting August 2018 and till July 2019. Applying Machine Learning algorithms helped us to process and prepare the big dataset for statistical modeling and analyzing. The application of machine learning in our model also enabled future enhancement of the model for parallel processing and advanced computing when multiple store groups will be taken under consideration. Our Minimum Viable Product will be the statistical model that processes a single store from a store group and does customer segmentation. The generated model then can be used to process multiple stores and compare their results. With multiprocessing, we can use additional processing power to process new incoming data and do computation.
- An Automated Process
Building an automated process is our most important feature of the project and key point for our minimum viable product. The automated process will be replacing the manual task for analyzing customer transaction data and segmentation. The current process requires an understanding of business domain data and various other factors to make the decision of the number of customer clusters. This requires a lot of time for manual processing, cost, manpower and thus delaying the delivery of business insights. This raises the cost of the project and delays in business insight analysis and thus project delivery. The automated process of the project will be processing all the data, analyzing all key variables, and then applying the statistical model for clustering. The model will compute a cluster range and evaluate each and then make a decision by itself for the best cluster segments. This model will achieve the decision making faster and add value to the overall business product reducing cost and labor.
- Informed Business Decision
For insight analysis, the model will be doing data mining and will result in each cluster’s details like the total number of tickets each cluster represents for each month, average sales per ticket each cluster for each month, and per ticket price range analysis. The insight analysis will help to make informed business decisions. It will help to measure the success of treatment analysis for the clusters.
- Visualization of patterns and trends
Insight analysis of customer segmentation when plotted using a visualization tool for each month, it l helped to understand the characteristics and patterns each clusters. This will help to understand the growth and behaviour of that cluster over a period. If any informed decisions where taken for treatment analysis like launching a marketing campaign targeting a particular genre of cluster then with the help of visualization of the patterns it will tell us if the campaign was successful or not for that targeted customer segment. This will help clients to take further decisions to launch more campaigns or stop existing campaigns, target multiple customer segments based on their pattern and demands. With the help of visualizing the analytical results, it will benefit the marketing and sales team to visualize the market results. Similarly, business team can take risks or plan in advance for any new product launch. This also helps to be competitive in the market and do better business.
The solution of our project will be given to the analytical company, RMS who does analytics for Food & Beverage Industry. And the customers in business who will be benefitted by our project model are the 100,000+ restaurant customers that RMS serve. It includes variety of customers like single restaurant owner, restaurant with multiple chains, or even a multinational restaurant franchise. Our developed model will benefit RMS with data processing and statistical decisions which will help restaurant clients with faster business insights and decision making. Our is a B2B2C or business to business to customer model where rather than accessing the end customers (restaurant clients) directly we will be accessing via another business, RMS. We all understand the B2B and B2C business are very straightforward but our business model primarily seems like B2B but in the end the goal will be to become a B2C business. The important logic behind B2B2C model is if a business doesn't have direct access to the end customer then it get the access via a second business. This applies to our scenario as well. We got the customer TLD data from the end customer to work on via RMS. The middle business was looking for an analytical solution and we did provide them which served the end customer.
To understand the end customer pains, we went out and scheduled customer interviews (restaurant owners) which not only helped us to understand the need of our business model but also shaped our model for our MVP. We interviewed multiple restaurants owners, shift managers, branch manager, and RMS and were constantly validating our approach for the ultimate solution. We conducted interviews in different locations asking direct questions related to our value propositions. We wanted to understand the exact need of our model and if applied how it will improve their daily business. Will the results be seen in short term or a minimum 6 months of wait period is required to see the results was asked. We got some direct answers to our questions which the customer was able to answer without looking into sales record or data. We understood that there is a need of our solution and it will help their business to improve. Customers we interview like big franchise mentioned that they have analytical team who does the customer segmentation and run product campaign based on analysis. This confirmed that customer segmentation is a valid solution for treatment analysis. But we also understood that there is a big gap in the market for this solution. During the interview with restaurant owners owning one to five restaurant chains were the one who are actually in need of our business solution. We believe that due to high number of restaurant chains and profits the franchise owners use analytical solution to grow their business. We see a possible clients for our business model and we can see our business to convert from B2B2C to B2C model, which will be our ultimate aim.