Unique Similarity Machine Learning Applications
simMachines revolutionizes what’s possible for today’s marketers. Enabled by “the Why” factors revealed behind every prediction, we provide marketers with rich machine driven insights at the speed and scale of today’s fast moving world.
[ chart ]
Dynamic Predictive Segmentation
Our similarity machine learning method enables us to generate dynamic predictive segments by grouping similar predictions together. These innately contain highly actionable insights at a segment level. Granularity can be defined by the user, and full transparency enables users to see the machine-driven factors behind each segment, compare segments, trend segments over time, and forecast the future behavior of a segment.
[ chart ]
simMachines uses its own similarity machine learning clustering engine to group predictions together. Clients can create clusters for different purposes as needed, query across clusters, and compare clusters and segments together over time.
Functionality Summary
- Cluster Creation: Create as many clusters as you like and define any number of classes for each cluster for analysis or campaign planning purposes
- Cluster Visualization: View the total number of segments tied to each class and see where each segment becomes distinctive from another at any level of granularity desired
- Segment Telescoping: Drill in on any segment to view its similar weighted factors and shift the order of weighted factors if needed
- Segment Labeling: Label segments to fit your objectives based on its defining factors
- Difference Comparison: Compare two segments together to see their differences
- Selection & Ouput: Highlight the selection factors you want to leverage in selecting an audience segment and output the objects with or without their associated data. A restful API can be used for automating selection output.
Customer Experience Optimization
In today’s highly demanding one-to-moment consumer landscape, the competitive battle field is based on how well you can manage each customer’s experience. Dynamic predictive segments can be leveraged to inform every customer interaction, with great precision and speed.
[ chart ]
Functionality Summary:
With Dynamic Predictive Segmentation for customer experience optimization you are able to:
- Create Multiple Clusters – Generate clusters that predict the likelihood a customer will buy a particular product or combination of specific products. Clusters can be generated in several hours and then updated and queried in real time.
- Query – Query a record in a cluster and immediately see which segment it falls into to tailor your marketing message and offer.
- Multiple Queries – Query a record across multiple clusters to determine what is the best recommended action to take during an interaction.
Explainable Pattern Detection & Forecasting
Explainable pattern detection and forecasting is essential to today’s businesses. Especially when using machine learning to consider all data sets that affect consumer behavior. simMachines enables businesses to detect emerging patterns, forecast future demand or consumption, and view historical trends in granular depth.
In this example, a large retailer wanted to trend users of VR equipment over time, for retention analysis and to forecast future device usage, as well as compare the user segments of one device to another.
[ chart ]
Functionality Summary
- Time Series Comparison: Compare two clusters, segments or classes over time
- Trending: Create time series analysis of the historical data by time stamping specific snapshots at any interval for reporting and analysis
- Pattern Detection: Reveal patterns in the data using “the Why” factors to explain the drivers of shifts and changes over time
- Anomalous Pattern Detection: Isolate non-similar objects as they occur based on distance parameters for analysis and reveal emerging patterns as these non-similar objects begin to appear more than once
- Forecasting: Build forecasts of predicted future events based on any time horizon or criteria and reveal the drivers of forecasts based on the weighted why factors
Customer Experience Optimization
In today’s highly demanding one-to-moment consumer landscape, the competitive battle field is based on how well you can manage each customer’s experience. Dynamic predictive segments can be leveraged to inform every customer interaction, with great precision and speed.
[ chart ]
Functionality Summary
- Justification: View the weighted factors behind each prediction to understand a prediction’s primary causes
- Hypothesis: Compare the gravity of a prediction’s unique factors to its opposite class to visually see its unique differentiators
- Continuous Learning: Continuous learning capability enables our algorithms to examine and adjust for local weighted factor differences between objects so that every prediction’s accuracy is further enhanced and automatically adjusts over time
- Customer Validation: Optimization of fraud detection and good customer detection are both important to ensure that as fraud detection rules are tightened, the degree to which a good customer is inadvertently identified as fraud is critical to avoid
- Audit Trail: Nearest neighbor object ID’s are captured and stored for every prediction so that they can be revealed to support each predictions identified weighted factors that were identified as the drivers
- Selection & Output: A restful API call is available for automating export of a prediction with or without its associated data to other systems.
