I would agree that data is directly observed, either by measurement or experience. We need to be careful to say that data need not be quantitative, but could also be qualitative. Quantitative data are readily used by computers, but whilst a computer can cope with categories (fields in databases), their ability to deeply understand qualitative differences is still a hurdle in the development of artificial intelligence. Computers' pattern recognition in high-dimensional data is more advanced, but the leap of the imagination that the human brain (and other natural brains in animals) can be observed to make is not something that is programmable in a standard way. Random treatment of information - having imaginative and possibly 'wild' ideas - plays a part.
But whilst data is directly observed, information is there whether we have observed it or not. 'Data' is a plural word, as we perceive that we can only observe things in bits (a 'datum' is a bit of 'data'), and that is how we programme computers. But information does not need to be such that it can be separated into bits. It can be a whole entity that has no obvious way to get in and look at the facets of it and analyse it. When we try to logically 'understand' something this is what we try to do to information, but it may be that our brains are capable of understanding whole information without breaking it down. But this means it is difficult to communicate to a classic computer the essence of that information and teach the computer how to analyse information of the same nature over and over again.
As a final point, I would say that data - the things we observe and combine together to make a body of information - are not necessarily useful. We collect the data in bits and then try to reconstruct the information, where the information is the entity we are in fact most interested in. After analysing the data, with discussion or with computers, we may decide that it isn't actually as informative as we had hoped it would be. Information is necessarily informative, whereas data, though we can talk about it, write it down, express it in computer code and so forth, is not necessarily informative. Collecting data costs time and money, so being sure beforehand that it will be useful for our improved understanding and wellbeing and progress is key.