HeaderSIS.jpg

IS480 Team wiki: 2018T1 analyteaka projectscope

From IS480
Revision as of 00:11, 15 August 2018 by Solomon.teo.2016 (talk | contribs)
Jump to navigation Jump to search
Analyteaka Header.png


HOME

ABOUT US

PROJECT OVERVIEW

PROJECT MANAGEMENT

DOCUMENTATION



Analyteaka Scope.png
Analyteaka customer profiling.png
Customer categorization Based on the historical sales data provided by Scanteak, we are going to work out the customers’ race (from name), gender (from name), age (from NRIC), income level (based on their housing district), and if they are return customers (based on the past transaction records).
Customer profiling Moving forward from customer categorization, which isolates various identifiable traits (age, race, gender etc.), we are going to generate several profiles/personas based on a combination of identifiable traits.

Examples of descriptive analytics would include:

  • Age/race/gender/housing district composition of customers
  • Customer return rate
  • Sales and quantity sold for each identifiable trait or profile
  • Main payment method

Examples of business questions that will be answered:

  • What is the average amount spent by customers between the age 40-45?
  • Are they mostly new customers or return customers?
  • What is the main target profile (40-year-old Chinese male)?
Analyteaka store profiling.png
Part 1 This module is responsible for providing descriptive analytics for different products and their respective categories. It will provide the foundation for predictive analytics (e.g. recommended product and quantity allocation for each store).

Examples of descriptive analytics would include:

  • Sales figure and quantity sold for each product and category
  • Sales composition of each product and category
  • Grouping of products which are commonly purchased together

Examples of business questions that will be answered:

  • What is the best performing product/category for this outlet?
  • How many products do customers usually buy in a single purchase?
  • What other products do customers usually buy when they purchase a coffee table?
  • What is the best performing store/location?
  • What is the best performing month/day/period?
  • Which is the race/age/gender demographic for each store?
  • What kind of item sells better as combination?
Part 2 This module is responsible for providing predictive analytics for different stores and their respective locations based on part 1.

Examples of descriptive analytics would include:

  • Number of customers for each store
  • Number of sales and sales figure for different months/days/periods
  • Customer (age, race, gender, etc.) composition of each store

Examples of business questions that will be answered:

  • What is the best performing item category next quarter?
  • Which store sells the most item next quarter?
  • What kind of product can we push as an add-on/bundle per season?
Analyteaka Bootstrap.png

A new sales system, meant to replace Scanteak’s legacy system, is currently being developed by Scanteak’s in-house developers. As the new sales system is still in the midst of completion, the bootstrap module will allow the user to upload the customer data exported from the new sales system. Once the new sales system has been completed, the manual bootstrapping of CSV files will be phased out and the bootstrap module will be modified to interact with the new sales system directly through API calls.

Furthermore, the new system will allow for higher level of data quality, providing better predictive analytics result.

Steps for bootstrapping

1. Data cleaning by removing the duplicate row (double entry, invalid rows)
2. Infer columns required

  • Gender and ethnicity based on first name and last name
  • Residential district/housing type/ house value based on postal code
  • Age and citizenship based on NRIC.

3. Convert into datastore request objects
4. Uploading data to datastore

Analyteaka data visualisation.png

This module will make use of Bootstrap, flask and Dash by Plothy to generate charts based on data output generated by customer and store profiling module.

Analyteaka Marketing planning.png

This module aids the management in


Analyteaka machine learning.png

This module will contain the machine learning system. As we are using Python as our main programming language, we will be utilizing libraries such as – SciPy, NumPy, matplotlib, pandas, Scikit-learn to help us complete this module. Using the training dataset (6 months’ worth of offline retail data) we have prepared, we will train the system to provide predictive analytics for both customers and stores.

The entire process can be automated, whereby the system can retrieve raw sales data from the in-house sales system, upload it to the server, process the raw sales and payment data in the machine learning module before returning it to the database. Examples of predictive analytics:

  • Recommended products for different customer profiles
  • Recommended price range for different customer profiles
  • Recommended products to be displayed for different stores
  • Best selling type of item and item category for different stores

Examples of business questions that will be answered:

  • What kind of furniture (based on price range) should you recommend to a certain customer profile (e.g. 30-year-old Chinese male at the Suntec Branch)
  • What kind of customers (e.g. Chinese) are you expecting at a certain branch (e.g. Suntec) and what kind of furniture (e.g. Oriental-style) should you display?
Analyteaka Footer Transparent.png