Marketing Optimization by Applying Machine Learning for Re-marketing of non-converted users


Monthly Subscription Service

- USD $1.450 per month - *

* opt out/one-click cancellation any time by client



FXclub, is interested in applying insights from data analytics to their Marketing needs.

The expectation is that it might be possible to apply machine learning modeling to produce the insights required, so that marketing results can be attributed as a function of propensity to convert to existing registered non-converted users.

To that end, FXclub submitted an initial dataset for exploration and feature engineering, so that data scientists on Naiss side can evaluate if this hypothesis is valid, and such a model can be defined.

More info about re-marketing is available in the technical approach section below.


Definition of Need

The objective is to improve current conversion ratios, and in particular find the propensity to convert of current registered users.

Based on defined propensity thresholds marketing actions can be tailored and optimized to increase and improve the number of converted users while reducing costs and/or increase the current conversion to cost ratio.


Customer process & execution

There are no a priori constraints or specific requirements. Project plan and task definition as well as exploration and approach are entirely open for Naiss team to decide. Naiss has architect ed and define the initiative as well as established any needs such as features, variables or other requirements (ETL tasks and work for instance) to eventually design and run the best possible machine learning models based on the current briefing from FXclub.

There is a baseline budget (USD $1450 per month) available to carry on the SoW agreed tasks against pre-approved work and deliverables.


Functional Use Case description - Re-marketing propensity ML modeling

Build a Machine Learning model to rank (scoring) current registered non-converted users by propensity probability to convert based on historical data.

The task involves deciding which type of model would be best for the problem at hand, as well as testing several tentative models based on previous feature engineering analysis as required for accuracy or other purposes.