Deep learning for marketing & sales

The challenge

Timeneye is a time tracking and time management web-app (with apps for smartphones and tablets) that allows medium-large companies to track their teams’ time across several projects. Timeneye offers a detailed reporting system typical of project management systems. The service offers a free 15-day trial, during which hundreds of activity logs are generated for each account, representing the actions performed by users while using Timeneye. The challenge is to transform this data into quantitative information and to use it to predict the conversion probability of each account into a paying customer.

Axyon's approach

A wrapper was built to gather log data from Timeneye’s databases, anonimize it, transform it into quantitative information regarding users activity during the free trial. Using Axyon's technology a statistical analysis was performed and data was normalized; then, a Deep Learning model was developed and tuned to predict the conversion probability of each account from its users’ activity during the trial period. A trained agent was developed and installed directly on Timeneye systems to update conversion probabilities in real time and display them on Timeneye Administrator Panel (used by marketing specialist to select specific marketing strategies for each segment). A weekly process automatically re-trains the agent using new data.

Results

The system returns the conversion probability for each trial account. The prediction is used by the marketing department to deliver tailored campaigns to different users segments.

The client’s system (benchmark) used a linear model implemented in Excel, and reached an average AR of 65%. The Axyon model reaches an average AR of 81%.

The performance enhancement led, for client, to a 60% improvement in conversion rate (thanks to better targeted marketing campaigns).

Sales