Developing an analysis plan (ethnographic data)

The other day, I found myself overwhelmed by the amount of data I had collected over the past four or five years for my global surgery project. I’m trying to finish up the book proposal, and I realized that there were bits of data that I hadn’t taken into consideration: should that stuff be in the book, or should I write articles? Should I just assume that none of those materials would be used at all? After mulling it over, I decided I would create an analysis plan. Back in the consulting days, when I designed and coordinated survey-based data collection (and charged cash-strapped NGOs a daily rate for my services), I had very little time to collect and analyze data and write up my findings. To stay within budget, I would usually take a couple of planning days to develop an analysis plan — not only to get clients on board with the scaled back version of their overly ambitious terms of reference, but also to make sure I could write a useful report quickly and as thoroughly as these short timelines allowed.

I’ve since moved on to “academic research” using primarily ethnographic data (which, yes, can include surveys, but in my case, usually does not). But I think the analysis plan template remains the same. Here, I’ll use my Twitter data as an example, even though it is probably the “rawest” and under-theorized part of my data set. (In other words, I think there has to be more said about what Twitter as social data and as social practice can tell us about specific, evolving social phenomena. But that’s another story for another day.)

  1. Decide on an overall timeline for preliminary analysis. I would like to complete preliminary analysis in 8-12 weeks. This means that each of the column headings below will correspond to a week or set of weeks.
  2. List all sources/types of data (can be two columns, e.g. Twitter, tweets). I have “scraped” twitter on my topic for over four years, using a google script. The script has produced a fairly large spreadsheet full of tweets, with some basic information about each of those tweets (RTs, Faves, etc). I also have collected odds and ends on Evernote (screenshots, radio interviews, photos, maps/images),  compiled field notes during participant observation at meetings, written detailed notes on interviews, created a database of documents from various archives; and transcribed and translated interview transcripts. Each data type/source corresponds to a set of research questions that I came into the project with. As the project evolved, I added more (sub)questions to add texture to observations.
  3. Link data source/types with specific research questions. For my Twitter data, for example, I have the following questions: How, if at all, do my interlocutors on this topic (hashtag) use Twitter? What topics do people tweeting under this hashtag write about? What types of people are in the networks coalescing around this hashtag? How has this group changed over time? Who are the nodes in these networks? Has this changed?
  4. Add new sheets as needed. Other issues with analysis and collection may arise as you sort data piles and link them to their proper set of research questions. In the case of my Twitter data, I realized that I may need to reach out to more people for interviews so I added an “interviewees” sheet.
    • This question came up as I scanned the list of tweets and the most prominent Twitter handles: have I talked to these people, or researched the material they have made available about themselves?
    • The interviewee sheet includes columns like ID, Name, Title, website, twitter handle, link to interview notes or transcripts. (I also have an “interview” data source in my general list, so this is helpful for tracking those data).
  5. Finalize duration allotted for this piece of analysis. As far as timeline goes, I have decided that I can handle this set of Twitter questions in three weeks. (if you’re fancy, you might create a Gantt chart, or link this to an Asana project, so that you’ll have a nice visual representation of progress on the project). 
  6. Follow your plan and update as necessary. Flexibility is key!

So, this is a rough sketch of my plan. I believe it could have better/more detail that would allow for tracking themes, but I’m relying on my work in Dedoose for this. Do you have an analysis plan template? What does it look like?