How to (Qualitatively) Analyze an Instagram Post

Have you ever thought about using social media as data in assessment, evaluation and research? 

1. Turn on your phone, open up Instagram.  Scroll through the first 20 posts you see.
2. Now do it again, but look for patterns – differences – look with the eyes of a researcher.
3. Were there any common posts?  Were there any specific posts that really stood out?  Could you describe 3-4 major themes?

For those not familiar with my work, I am in the final semester of my doctoral program.  This April I will defend my dissertation, which took on a mixed methods approach to examining college student leaders’ use of and experience with digital communication tools.  Read more about the entire study (here).

With a mixed methods study, you have multiple phases in your research.   I began by exploring the topic qualitatively in focus groups.  Using 40 college student leaders at two institutions, I gathered a rich amount of data.  In the midst of the focus groups, I began gathering over 2,200 Facebook and Instagram posts that ranged in dates nearly an entire year.  I qualitatively coded these social media posts, as well as quantitatively analyzed them.  The final results can be found (here).

For this post, I would like to share how I went about analyzing, through traditional qualitative methods a social media post.  This is written through a formal research and assessment lens.  This unfortunately might take longer, but will result in a more fruitful result.

1.  Answer the Question: What are you Seeking to Research?

You need to have a focus.  In traditional research this is called a research question. Using a qualitative approach, you do not need to have a hypothesis for variables, but it does need to be measurable.

For example, it would be a hard study to approach answering the research question, “Do new students have a good time at orientation?” when analyzing their Instagram posts while at orientation.  A modified and more appropriate question would be, “What experiences were observed on Instagram at Orientation by new students, their families and the campus community?”  Still not a perfect question, but it continues to drill down into specifics.

2.  Get Access & Approval.

Continuing with our orientation example, if you are coordinating orientation you have an access point to this population, but that doesn’t mean you have the freedom to conduct research on them.  Here are elements to consider when getting access and approval:

• Technically, any formal research conducted on human subjects needs to be approved by your institution, through an Institutional Review Board (IRB).  Through this application process you will need to propose how you will gain approval from participants.  In this case, you would pull all social media posts that were shared during new student orientation.
• A few ways you could obtain approval from participants is making it part of the registration process, having all attendees agree to an Informed Consent.
• An Informed Consent  explains your research and the protections the participants have.  I could write a post entirely on informed consent, however I leave this section intentionally brief.

I hope to make the point that you can’t just start pulling data anywhere without asking permission first, even online.  There is much debate about this now, many times in the ‘big data’ conversations – like how Facebook captures data on us everyday and runs research.  In education, always lean to obtaining approvals up front.

3.  Connect Research to the Appropriate Platform.

In an example such as new student orientation, you could potentially have thousands of posts across many different platforms.  If this is your first attempt start slow.

• Only pick one platform to monitor and capture posts.  Based upon my research, I would recommend Instagram or Twitter.  Both have the ability to pull from a common hashtag without ‘friending’ or ‘following’ your participants.
• Using either of these platforms, establish a hashtag that you will encourage all participants to use.  This is another way you get ‘approval’ from participants and they can choose to participate knowing what it means when they use that tag.

One note about Instagram.  Unlike Twitter and Facebook, which have robust desktop websites, Instagram does not.  For my research I gathered participants posts from another site called iconosquare.  This is free and pulls directly from Instagram.

4.  Organizing your Data.

If you are lucky, members of the campus community actively used your orientation hashtag, so you’ll have a solid amount of Instagram posts.  However, you need to answer a few questions.

• First, how much data is enough?  For example, the weight of only analyzing 10 posts versus 100 is much different.
• Do you want every single post, or just a sample of each orientation session?
• Or do you choose just one orientation session and analyze every post there?  Ideally, this would be decided early on, but may become even more important once you see your data.
• You may also have a challenge of not having enough posts to feel you have substantial results.  Here you might want to expand also pulling posts from Twitter.

With decisions made on where you pull social media posts from and how many you hope to obtain, the time must be invested online to capture them.  Here I provide some tips for data collection through social media.

• For my research, I captured direct screenshots of the posts, including any comments, likes, etc.  In other words, on Instagram I was interested in looking at more than just the images, but also the interaction displayed such as comments.  Again, I did this for Instagram through www.iconsquare.com.
• Also consider where these images will be secure.  I saved my images on a hard drive that was password protected.
• These images were organized based upon institution, as well as individual pseudonym.  For orientation, you might organize your data in folders by orientation session.
• Unlike other qualitative research where you would transcribe interviews, this research is visual.  Formally it would be referred to as document analysis, which has also expanded to video and photos.
• Even while gathering posts you should be taking notes.  These will help you when moving on to a future step called coding and memos.

5.  Choose your Analysis Platform.

This could be an optional step, but one that would save you a lot of time.  In qualitative analysis, there are software programs that assist in organizing and analyzing data.  However, not all of these have the capability to add images.

• Because I work on a Mac I selected Nvivo.  I basically taught myself this program, with the helpful tutorials the program has, found (here).
• Other program options (based upon your operating system) include Dedoose, AtlasTi and Saturate.
• Take your time in learning these programs.  Most are very intuitive, but learning the ins and outs, especially in how to upload and code images will be worth it.

6.  Coding & Memos.

After selecting and uploading your images into the software program, you are ready to start coding!   To provide background on what is a code – I’ll provide some definitions straight from my dissertation.

As Merriam (2009) describes, the researcher assigns codes to data to develop themes or categories.  Further, Creswell (2012) defined coding as the, “process of segmenting and labeling text to form descriptions and broad themes in the data” (p. 243).  This process will be continuous.  After developing sufficient codes, themes will begin to emerge, which are “similar codes aggregated together to form an idea in the data-base” (Creswell, 2012, p. 248).

Most research would approach the first round of qualitative coding through what is called ‘open coding.’  This activity is open to any and all patterns that emerge when analyzing your data.   Ideally all these codes would still fall within your research question(s), but in opening coding you are open to all possibilities that could be found in your data.  This could also mean you may end up with results that won’t answer your research question(s), but can lead to other publication or presentation opportunity, research studies or programmatic initiative.

To give you an example of how I would open code a post, here is an example of an Instagram photo I took last year at a Super Bowl gathering: 

Possible open codes: *Selfie * group * event * friends * football * YouTube * couples * evening * reunited * smiling * engagement *

If you think about it, it is kind of like ‘tagging’ descriptions like you would include on a blog post or hashtags on Twitter or Instagram.  This is a very simplistic view and is without a research question, but also hopefully gives you confidence that coding posts does not need to be complicated!

It is also important in this stage of your research that you are taking notes.  Formally, this is called a memo, which is a formalized means of journaling.  Memos can also be included in your data analysis.

Ideally with open coding, one would code every post, then you would go back for a second round of coding that was more focused based upon preliminary results.

For example, looking at the post above at my Super Bowl gathering.  Another code that could have emerged would be location, so I would add *Indoor.

7.  Advanced Options.

If you’d like to further establish how you analyze your social media data, after your initial open coding, look to ground your research in theory and practice.  This could be established through a rubric.

• For example, what were the learning outcomes for the program, mission of the department, student developmental theory, etc?  For my study I built the Social Change Model into my social media rubric, looking for posts where students were presenting one or more of the seven dimensions.
• You can also use ‘quantitative’ measures, as simple as counting how many times a behavior occurs.  For example, you could include how many likes, comments or even the Instagram filter.  For example, I could code how many ‘likes’ each post received.  These could be codes such as 0, 1-10, 10-25, 25-50, 50-75 and so on.  Let’s do this using a new example.

Below is a photo at an event at Disney World with my husband.?

• This photo received 100+ and even had 10 comments.  Using Iconoscare, you also see the Instagram filter (Ludwig).
• If I applied a quantitative method to all my Instagram posts, I could look for an average ‘comments’ and even ‘likes’
• I could also look for posts that included humor, romance, my dog, inspiration, school work, etc.  I would assign 0 for no and 1 for yes.  A simple excel doc can provide frequencies and descriptive statistics.  (Hint I just basically gave you want I post about on Instagram, ha!)
• What is important at this advanced point, you only code what content will help answer your research question(s).

8. Find your Themes.

Establishing your themes is like sorting jellybeans.  They may be clustered by color, size or a grouping of each color.

Look big picture and drill down to the details and back out again, so obvious results from your coding may actually reveal deeper meanings.  If you have used a qualitative software, the fastest way to see your initial results is to find how many times you coded certain categories.

Going back to our orientation example, let’s say 80% of posts included group photos.  This is something to pay attention to.

Why?

• This fuels to answer your research question, of what are the experiences at orientation, as presented on Instagram.
• By having 80% of posts being group photos, this means that participants using the Instagram hashtag at orientation are presenting images with others.  Note: one can’t necessarily argue that this leads to lasting friendships or the quality of the orientation program.
• It does inform you that at orientation, the campus community is posting a lot of group photos.  This could be a positive sign.  For example, rather than a higher percentage posting selfies, group photos displays participant interaction.  Tie this back to your learning outcomes and you have something!

9.  Major Findings.

Your major findings may or may not be the same  as your themes.  Your findings should directly answer your research question(s) and aid in shaping future programmatic and policy direction.  Findings are your recommendations based upon the established results.   Using the orientation example, if group photos were the highest behavior presented on Instagram during orientation – this may fuel an argument to keep as many group interaction opportunities in the orientation program.  They are ‘sharable’ content to participants.

Ready to really dig in deeper?  Form a focus group of those community members that posted group photos, to find out qualitatively their experiences at orientation and their use of social media.  Multiple research methods is a form of triangulation, which will result in more salient findings.

10. Share Widely!

Please share…share…share.  This is an emerging method in research, both in qualitative and quantitative means.  Look to conferences, journals or even blogs that you can share results, as well as lessons learned.  In my final chapter of my dissertation I will clearly communicate the limitations of my research gathering and analyzing social media data.

I’d love to hear from you if you are thinking about this type of research/assessment.  Don’t hesitate to reach out, using the contact page (here) or tweet at me @josieahlquist.