Why ROMImate?
The main goal is to improve quality of measurement in marketing through better use of latest technology. To achieve this goal, I strongly believe, marketing industry needs to attract more technologists to innovate and develop tools and techniques to improve processes of marketing. Therefore, the primary purpose of ROMImate is to facilitate learning of technologists and promote research and development in the field of agile marketing investment measurement.
How does it work?
This prototype of ROMImateis is designed to showcase a simple workflow of media mix modeling. User can play with data to see how any change in media or sales data affects models that is visualized through a pie chart. The dashboard is designed to detect any change in data and notify user to run regression analysis in the real time.
Other supported features at the moment are:
- CSV file upload: User can upload their own data in predefined format as suggested in the upload section of the dashboard to see result of modeling.
- Social Login: Social login with top 4 social media (Facebook,Google, Twitter and Linked-in) is supported to facilitate easy sign-up.
- Social Sharing: User can use social share options to share their experience on top social media.
What technologies are being used today?
The system architecture of the prototype is entirely based on serverless architecture with right blend of NoSQL database and micro services coding pattern, as shown in diagram below:
- Analysis Engine: A linear regression analysis micro service is developed in Python to perform basic media mix modeling on collected sales and media data. This service is an important piece of prototype. It can be used as reference to enhance regression analysis further. Another service for Sentiment analysis is also available as part of analysis engine though it is not used in prototype implementation.
- Report Interface: The ROMImate report interface is entirely developed in Node.js
with extensive use of Express.js web application framework and Pug.js template
engine. Some of the key Node.js or JavaScript packages used in the construct of
ROMImate are:
- Passport.js for Social login
- dc.js and d3.js for Charts
- crossfilter.js for in memory data handling
- More package details are available inside package.json of report interface.
- Conversion Hub: A reference micro service is developed in Python to pull conversion data from Google Analytics through REST API. Though this service is not used in the prototype implementation but it can be used as a reference to develop other service as per the need of developer.
- Media Hub: A placeholder reference micro service is developed in Python to pull data from a media data provider through REST API. Though this service is not used in the prototype implementation but it can be used as a reference to develop other service as per the need of developer.
- Database: The ROMImate framework database is named as “mediamixmodeling” that stores all the collections used by service hubs, analysis engine and report interface.
Where could you take ROMImate tomorrow?
That's depends on the developers community, but we've got some ideas that could help in giving it a direction.
Here are the few features that we would like developer's community to work on in near future:
- Interface for Sentiment analysis engine – data from sentiment analysis can be used as a media parameter to see how sentiment can affect sales through media mix modeling.
- Interface for media investment planning to allow marketers to analyze and optimize media investment.
- Support for dynamic dependent and feature variables – user should be able to upload data in CSV format with their own attribute definition for media and sales.
- System and application level exception handling to make framework robust and crash proof.
- More data visualization through graphs in dashboard.
- UX redesign.