A central aspect of online decision tree solutions is evaluating the incoming\ndata and enabling model growth. For such, trees much deal with different kinds\nof input features and partition them to learn from the data. Numerical features\nare no exception, and they pose additional challenges compared to other kinds\nof features, as there is no trivial strategy to choose the best point to make a\nsplit decision. The problem is even more challenging in regression tasks\nbecause both the features and the target are continuous. Typical online\nsolutions evaluate and store all the points monitored between split attempts,\nwhich goes against the constraints posed in real-time applications. In this\npaper, we introduce the Quantization Observer (QO), a simple yet effective\nhashing-based algorithm to monitor and evaluate split point candidates in\nnumerical features for online tree regressors. QO can be easily integrated into\nincremental decision trees, such as Hoeffding Trees, and it has a monitoring\ncost of $O(1)$ per instance and sub-linear cost to evaluate split candidates.\nPrevious solutions had a $O(\\log n)$ cost per insertion (in the best case) and\na linear cost to evaluate split points. Our extensive experimental setup\nhighlights QO's effectiveness in providing accurate split point suggestions\nwhile spending much less memory and processing time than its competitors.\n