Data Product Integrity

[AUDIO LOGO] It's a fully curated and easily understood set of data that is made available for an express purpose. All right? So instead of giving people data that they could do something with, here is the data to look at your year-over-year performance in product lines X, Y, Z. So identifying the data products that rely on my data source. I don't know how you do this. You absolutely want to do impact analysis. And if there's data products downstream that are relying on data that you currently have in one system and are migrating to a different system, Yes, you have to make sure that you're made whole through that migration so that the nature of that data product doesn't materially change. Or if it does, A, there should be a good reason behind it, a business reason, not a technical reason. But if there is some sort of impact on it, that that material change needs to be well understood, documented so that that data product will reflect that moving forward. Tell me more about the impact on us. It's coming from the data model. Well, the data model is a great place to do and perform impact analysis. It depends on the level of impact analysis that you want. But yes, if you make a change to a column where what is the impact analysis of that. Now, impact analysis gets into a bigger world with more of a data intelligence, data catalog type approach, where you're starting to see a lot of the end-to-end life cycle of that data so that again, if we change something here, what are all the ETL jobs? What are all the testing jobs? What are all the profiling pieces that are going on, the observability? All of that will be impacted in some way, shape, or form by the migration. And the beauty of having a good, strong data intelligence backbone in your organization is to be able to identify potential impacts and plan for those impacts and then mitigate any negative impact that may come from that. And negative impacts generally come from not knowing and then realizing after the fact. So again, I'm a bit of a broken record where, what is it they say? Measure twice, cut once. I think you're a fan of data modeling. I could be.