This publication addresses the adaptive control of manufacturing deviations in micro gear hobbing. We aim to establish Zero Defect Manufacturing in a series production using manipulated parts with a small sample size and machine learning. Therefore, optical focus variation metrology is used to measure gears inline. Afterward, the evaluation of measurement results based on trained models and the transfer of correction parameters back to the machine tool through control algorithms are established. Critical parameters of the manufacturing process are identified through preliminary tests, which are varied using Latin Hypercube Sampling. The resulting experimental plan defines manipulated deviations for manufacturing 200 sample gears representing production variations. The evaluation according to parameter-based gear deviations enables the modeling of influencing quantities using machine learning. This information provides a control algorithm for the feedback of correction values to the machine tool based on data analyses. After validation, it is shown that the current state of measurement technology enables the inline quality control of micro-components. The final control loop achieved accuracies in the micrometer range at detection levels of over 90%. Consequently, these results form a basis for implementing future adaptive quality control loops within data-driven production.