Lityx was started about 7 years ago with a focus on analytics, especially marketing analytics. Lityx is focused particularly on more complex forms of analytics such as predictive modeling and optimization. Lityx works with companies from non-profit, media, gaming, financial services, healthcare and retail. Their initial focus has been on marketing analytics and involves optimizing customer marketing, optimizing operations and performance analysis. Having started off as a consulting firm they have increasingly focused on their product – LityxIQ.
LityxIQ is a cloud-based marketing analytics platform supporting data management, analytics, insights, modeling, and optimization. This offers speed to market, reduced costs (specifically no internal IT costs) and ease of use with built-in domain expertise and models. LityxIQ is focused on marketers and other business users, but also provides a means for expert analysts and modelers to be much more efficient while still building powerful models.
Within LityxIQ there are four layers:
- Data Management
Moving data to the cloud and managing it
- Analysis and Insights
Visualizations, reporting, and dashboarding
- Predictive Modeling – PredictIQ
Classic predictive analytic models including churn, value, risk, affinity etc. Automated model development, validation and management, version control etc. Models are implemented automatically and forecasting is also supported.
Marketing optimization, media optimization, and resource optimization using constraint-based optimization and linear programming
The product is all SaaS with the relevant security etc. and Lityx provides rapid results services as well as the option to manage analytics for a company longer term.
PredictIQ is the core of the tool. Within this a client has their own work space where they can manage multiple projects within which they have multiple datasets. Multiple models and model libraries can be defined and versioned with all models having development and production versions. Creating a new model starts with the business objective of a model (a response model for instance) and this constrains the variable selection, metrics, and algorithms used later in the process. For each model the user can build or edit the model, execute or schedule it to generate scores, analyze the performance of a model and manage approvals/versioning.
When building or editing a model, candidate predictive variables can be selected from the ones that make sense for the model (PredictIQ will filter this list as the modeling progresses). A new model has default sampling, validation (holdout) and algorithms selected, but a reasonably advanced user can change any model settings as well as attach any number of algorithms to be run and compared for the model. Settings for the selected algorithms can also be edited by a more advanced user. Output can be defined as the best model or all the models selected.
For any model a set of metrics can be displayed and analyzed, and PredictIQ automatically generates all the metrics that make sense for a given kind of model. Using the other solutions within LityxIQ (Data Manager, InsightIQ) you can also feed new data into models to track performance metrics and see when a model needs to be re-built. The environment also allows a model to be re-built automatically every time new data is pushed into the cloud.
Automated scoring is easy to set up and new datasets can be uploaded and scored on whatever schedule makes sense, or scored automatically when the data is refreshed. Scores are stored in a scoring catalog and are versioned along with the model so it is possible to see which scores were created from which model versions. Today results are exported as a file of scored records that can be downloaded and used in external systems, or exported automatically to an FTP server. Next year they plan to offer an API to get the results of a model in real-time from other applications.
More information is available at lityx.com Lityx is one of the vendors in our Decision Management Systems Platform Technology Report.