Every received FOIA request
Every request for information that’s received under FOIA (aka "FOI," aka "OPRA"). This should include the name of the requestor, the date of the request, and the information being requested. Tracking requests can be beneficial to agencies, so that they can look for patterns to determine what data to pubish preemptively, and it allows the public to see how many requests that the agency receives, something that is otherwise invisible to the public.
Note that some media outlets may object to this, because making their requests public could tip their hand to competitors about what stories that they’re working on.
Every FOIA response
Every response to every FOIA request. It should be possible to correlate this with the record of every received FOIA request (or both can be the same dataset). Should include the cost of fulfilling the request (if a cost was charged) and a copy of the material provided in response to the request.
Creating this dataset can reduce substantially the burden of FOIA requests, since there’s no need to request information that’s already available online.
For more about this, see Mark Headd’s “Making FOIA More Like Open Data.”
Staff roster and organizational chart
It is not uncommon for agency org charts to be published as a PDF. Render this as data, including all employees, to provide both open organizational data and a staff roster. This is perhaps more useful inside of government than outside of government, as people within agencies and in other agencies often need to know about personnel and the reporting structure.
The government should be providing the entire government’s budget as open data, but each agency should extract their own agency’s portion and republish it, in addition to their internal budget data. Many state governments established “open checkbook” sites in the mid- to late-2000s—when this data is still available, break out of the data by agency and provide the data on a per-agency basis.
If practiced by all agencies within a given government, this can become a standard data mechanism for the budgeting process.