Data sources can be used and shared for both direct connections and extracted connections. In the event that your live connection is not fast, use extracts whenever possible. (But do make sure to hide unused fields and filter out extra data as mentioned above.)
Many of our data sources have more than 10 million records. By running these data sources off extracts, we can deliver data that’s both user-friendly and lightning-fast.
It’s also important to be agile. Iterate quickly, then roll out. View your data as a starting point, not the end. It can be tempting to try to create a data set that answers every question. Instead, focus on delivering a data set that can answer the most important questions. This will get the analytic journey started. You can always dig into more data later on.
And start collaborating in the early days. When you are building data sources, seek the perspective of a power user or an expert on the data’s subject matter.
Once you’ve got a data source, run a short pilot period (~1 week with ~20 users). Ask them to try it out and give feedback. You can make major changes during this period, even structural, without worrying about workbooks breaking.
Using this process, we’ve been able to roll out enterprise-ready data sources in a matter of weeks.
Once you’ve refined the data source, share it more widely. Write a note on your shared wiki, and send out an email to anyone who might want to use the data source. It’s better to overshare here. People want this data and will be excited to have a tool that makes their lives easier.