Data isn’t either open or closed. It’s not a switch that gets flipped, but instead a sliding scale, in which closed data is made increasingly open.
One popular assessment method is 5 Star Open Data. You probably should not use this to assess your data. That’s because it’s a standard that is promoted by people who are excited about something called “linked data.” The fact that you are reading this guide is solid evidence that you are not going to care about publishing “linked data,” a thing that is ultimately really about data, rather than open data, anyway.
So let’s invent our own three-star standard here (which is really just the first three levels of the five-star standard).
Take some data (say, an Excel file), put it on a website for people to download. Yay, you opened data! Give yourself a pat on the back.
Now put an open license on that Excel file. That is, explicitly state that it’s in the public domain (as it probably is, since you work for a government). If it’s not, release it under an open license, like Creative Commons Zero. Now people have permission to redistribute your data and incorporate it into other works, which is great news.
Now publish your openly-licensed Excel file in an open format (e.g., CSV). Congratulations—your data is now very open data!
A Little Open is OK
If you have important data that is only currently available in non-open formats (e.g., Excel) or even non-structured data (e.g., PDFs, Word), then go ahead and publish it in that format! It is far better to publish not-very-open data than to not publish data at all.