Whenever we search for our preferred jam at the supermarket, we are engaged in a complex visual search task. But how is this skill acquired? In this study, we analyzed data from an online educational game similar to the card game "Set". The game’s objective is to identify sets of three cards that either share features or have entirely distinct features across several dimensions. The objects on the cards vary in number, color, shape, and shading, and players must identify the correct set of targets among a range of distractors. We developed measures of perceptual salience for targets and distractors by counting shared attributes (e.g., all targets have two symbols) and distinct features (e.g., the symbols on all targets have different colors, such as purple, green, and yellow). Across 1,169 game items played by thousands of elementary school students, we found that the more features targets share, the easier it is to identify the set. Likewise, the more similar the targets are to distractors, the more challenging it becomes to find the set. However, the similarity among distractors did not influence item difficulty. Furthermore, we find that not all feature dimensions are treated the same by the children. Card sets that share filling and color were easier to identify for kids then card sets that share shape or number. In addition, the difficulty decreases with each additional shared feature. Overall, the study provides a new approach to investigate more complex search tasks.