In order to harness machine learning effectively, the data used for training and evaluating the models needs to be of high quality and relevant to the problem at hand. Some of the key characteristics of the data that is needed include:

1. Quantity: Machine learning algorithms typically require large amounts of data in order to learn and make accurate predictions. The more data available, the better the model can learn from it and generalize to new data.
2. Quality: The data should be accurate, consistent and free of errors, outliers or missing values, as this can impact the performance of the model.
3. Relevance: The data should be relevant to the problem at hand. It should contain information that is useful for making predictions or decisions and should be representative of the real-world problem.
4. Labeling: The data needs to be labeled, meaning that it needs to have the correct output or target values for each input example. This is important for supervised learning algorithms which learn from labeled examples.
5. Variety: The data should be diverse and come from different sources, to avoid bias and to generalize to new data. This can be achieved by having a diverse set of examples or by augmenting the dataset with different versions of the same example.
6. Format: The data should be in a format that is compatible with the machine learning algorithms and tools being used.
7. Privacy and security: The data should be kept private and secure, and comply with any relevant regulations, standards, and laws.