- Company name: XICA CO.,LTD.
- Interviewees: Product Development Department Manager Mr.Soyama, Mr.Shimojo
Visualize Promotions' Statistical Analysis
XICA CO., LTD. provides the marketing analysis tool "XICA magellan", which visualizes offline advertisement measures such as TV commercials and magazine ads. By analyzing the input data of advertisement placement and achievements calculated daily, Magellan visualizes the customer's advertisement measures' effects and the relation with subsequent web measures and allows them to optimize their advertising expense allocation. Magellan statistically analyzes how much sales will increase when investing in a particular advertisement measure. Since Magellan handles many statistical analyses, we mainly use Python for development.
Amount of Calculations and Parallel Processing
XICA magellan runs many regression analyses. Because it analyzes the relations between various advertisement measures, the number of regression analyses can reach up to a few hundred, which would normally take several hours to compute. However, since the calculations for each regression analysis is independent, we reduce the calculation time by parallelizing each of them. Currently, we're using the
multiprocessing package of Python and parallel process multiple processes on a single high spec machine. Added to this, we are planning to use virtual container systems such as Docker to parallelize multiple machines so that Magellan can quickly handle large amounts of statical calculations.
A Renewal to Make Introduction and Configuration easier
In having many companies introduce XICA magellan and observing actual use cases, we found several possible improvements. One of them is to make configurations when newly introducing or starting new measures quicker so customers can undergo the PDCA cycle more speedily. In the current stage, we have users input values in Microsoft Excel format which is converted to be input to the system. We're planning to make improvements so that users can to do this from XICA magellan's GUI, and make configuration quicker. Together with that change, we decided to renew our system so that further improvements could be made smoother.
Along with this renewal, we changed our development environment, changing source code management from BitBucket to Github, making code reviews more active. Together, we introduced SideCI as a static tool to continuously secure quality. Many Python developers put emphasis on coding standards, so the introduction of SideCI went very smoothly. We're being able to keep the renewed system at a state where all the codes follow the coding standards. With this, we will increase the speed of development of future maintenance, improvements, and add functions.
Making A Simple Tool You Can Know About Inside Out
As a concept in developing XICA magellan, we want it to be a tool that customers can master using. Rather than adopting new technologies and doing advanced and complicated things, we want Magellan to be a tool which can be reassuring to customers, can be used with ease, and improve one's creativity. Analysis utilizing statistical analysis and machine learning tend to become 'black-boxes', but by maintaining simplicity, we aim to support our customers with a simple foundation they inside and out.
Signup and try Sider for free in just 30 seconds!
You can try 14 days for free.