Modeling of risk for device launch using device lab test data and friendly use test data. Model has high correlation to number of device issues and device returns post launch.
How to predict the post-launch performance of devices based on test results from the Lab?
How can we prioritize tests which are most relevant to device performance?
How can we make go/ no go launch decisions with quantitative data?
Extensive analysis of handset test data using inferential statistics.
Anomaly detection to flag outliers.
Identification of key variables from test data impacting smartphone performance.
Deriving rules and thresholds using Ensemble Modelling techniques.
Variable reduction carried out from 1000s of initial test results to pinpoint key variables.
Model for predicting post launch test results with an accuracy of ~ 80%.
Model for Predicting customer’s propensity to buy a specific smartphone based on variety of Customer and Handset metrics.
Which of our users are most likely to switch to smartphone?
What is the most likely handset they will switch to?
Creation of customer and device usage features by mining CRM data, CDR data and network data.
Analysis of historical subscriber switches and the overall device switches patterns to create custom metrics.
Modelling using advance boosting algorithms and hyper-parameter optimization.
Model created for predicting users with upgrade propensity to smartphone with an accuracy of ~70%.
Model deployed nation wide with 15% improvement in positive targets.
A comprehensive device performance benchmarking and analytics solutions using network, probe and crowd source data.
Can we create a unified metric to rate and rank smartphones being uses in the network?
Can we identify consistently poor performing devices in the network?
Device metrics formulation using call trace and customer call data.
Threshold analysis and device performance score formulation using multidimensional statistical analysis and unsupervised algorithms.
Geographical and Temporal consistency analysis.
Identified devices with consistent performance issues in the network highlighted to OEMs.
High degree of correlation observed between device performance scores and metrics derived from Business / Marketing Data.