That Randy Brown

Bootstrapping new Features with Data

When adding a new piece of data or metadata, such as tagging, to an application, the feature must come with at least one use case beyond the ability to create, update, delete, and add metadata to existing data structures. This incentivizes the usage of the new data for users, and trains them to the usecase at the same time.

This can be something like 'now you can search, sort, and filter by this new data attribute' or more rich data visualization / drill-downs / intersections of data. In either case, it is important that the user be able to use the data for something immediately. Sometimes, this can be short-circuited and a feature can be introduced without additional data, and then data can add enrichment.

To simplify the leveraging and adoption, the closer to only one unique use case that a dataset gives, the better. Search, sort, and filter, may already exist on data in the app, in which case, adding a metadata attribute to these capabilities might be sufficient to facilitate adoption, but it also might mean that something else is already facilitating these activities, as this data -- especially tagging, is much more user driven segmentation.

This complicates the addition of new data, and there's a natural performance curve where it no longer makes sense to add new data elements because there's either no new feature to encourage its use. Because of this, user defined data attributes are probably the most powerful data model that can be added, and after they are added, there are diminishing returns on any structured data. Similarly, I conjecture that the natural performance curve plateau is probably when the data model is finally complete, and further features don't add data, but recontextualize it in new views and combinations. Similar to how user defined data attributions are the ultimate end of this journey, user defined data contextualization and visualization is probably the final feature: the ability of a user to put whatever data in a view that they want.