StakeSage-C
Last updated
Last updated
StakeSage-C is our liquid staking conversion rate estimation model designed for the Ethereum network. It accurately predicts the likelihood that an Ethereum wallet—without any previous liquid staking deposit (LSD) interaction—will perform its first liquid staking transaction. With over 95% accuracy on out-of-sample test data and availability for more than 130 million addresses, StakeSage-C empowers platforms to enhance customer targeting and optimize user conversion strategies.
Developed as a turnkey solution, StakeSage-C enables you to maximize return on ad spend (ROAS), dynamically adjust purchase funnels, and gain deep customer intelligence without the complexity and cost of developing an in-house solution.
StakeSage-C delivers actionable insights with high precision by identifying wallets that are most likely to convert to liquid staking. Its capabilities help improve marketing performance by focusing on users with a high probability of conversion, thereby ensuring advertising budgets are spent effectively. The model also supports dynamic purchase funnels by tailoring the user journey based on conversion readiness, guiding users directly to purchase or to further information as needed. Moreover, StakeSage-C provides enhanced customer intelligence through rich, machine learning–powered insights that enable better segmentation and personalized messaging.
StakeSage-C is powered exclusively by on-chain data from Ethereum. The model was trained on an extensive dataset that includes 483,050 wallets, 6,279,650 token features, and 16,374,701 transactions. This robust dataset has been applied to score over 130,960,187 wallets, ensuring that the model captures the full complexity of wallet behaviors. Such detailed and specific inputs enable StakeSage-C to provide precise predictions on conversion likelihood.
For the development of StakeSage-C, we rigorously evaluated several binary classification frameworks, including logistic regression, XGBoost, and artificial neural networks (ANNs). Logistic regression significantly underperformed compared to the other methods, while ANNs did not provide enough gain to justify their additional deployment and computational costs. Ultimately, XGBoost emerged as the optimal choice, balancing high accuracy with efficiency.
The model uses a multivariate prediction framework that analyzes complex wallet characteristics derived from on-chain data. This sophisticated yet efficient approach allows StakeSage-C to capture intricate user behaviors and accurately predict conversion likelihood, ensuring that its predictions are both robust and actionable.
StakeSage-C has demonstrated over 95% accuracy on out-of-sample test data, confirming its ability to generalize well to new, unseen wallet addresses. Out-of-sample testing means the model is evaluated on data it did not see during training, which is a strong indicator of reliability in real-world conditions.
Beyond these controlled tests, a two-month live validation study concluded in January 2024 found that StakeSage-C correctly identified nearly 70% of new stakers 30 days in advance—representing a 13% improvement over the baseline. These real-world results underscore the model’s practical effectiveness, highlighting its potential to significantly boost marketing and conversion outcomes in live environments.
StakeSage-C is designed to maximize marketing ROI by precisely targeting high-value users, ensuring that advertising efforts are focused on those most likely to convert. By accurately predicting conversion behavior, the model supports dynamic purchase funnel optimization—directing users to purchase when they are ready or guiding them to additional information when needed. In addition, the deep customer intelligence provided by StakeSage-C enables more effective segmentation and personalization strategies, which can significantly enhance overall user engagement and drive long-term business growth.
