<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=1615255955451271&amp;ev=PageView&amp;noscript=1">

The Social Media Intelligence Blog

Insights on social media intelligence, marketing, and consumer insight

How to Conduct Social Media Research

By Rion Martin  •  March 14, 2016

How to Conduct Social Media Research

Posted by: Rion Martin on March 14, 2016
Rion Martin
Find me on:

As Donald Trump would say, "The amount of data out there? It's YUGE! I mean, it really is."

In fact, according to DOMO, every minute: Facebook users 'like' 4.1m posts, Twitter users send 350k tweets, Instagrammers 'like' 1.7m photos, and Snapchatters share 284k snaps.

Within this data is a treasure trove of potentially valuable information. However, the hard part isn't getting your hands on data, it's actually getting any meaning from it once you have it.

And this is particularly true of social media data.

Through years of trial and error, failure and success, we've come up with a six-step process we follow for conducting social media research in a way that achieves the best results.

Here's our step-by-step recipe.


1 : Get to know the category, the industry, or the subject (if you haven't already)

So, you've been given access to millions and millions of social media posts and your boss or your client wants you to analyze that data and bring them brillant insights.

But before you dive into what consumers are saying about your subject, you should take some time to get deeply familiar with the subject in general. 

What issues affect the subject you're performing research on? At an industry level? Category level? Who are the primary competitors? 

 

jetblue-competitors.png

 

Search Google News and Google Finance, scan company websites, read annual reports and industry publications, read Wikipedia entries and the articles referenced, and use UberSuggest to generate keywords from what people search for.

Make a list of the key things that stand out that consumers may be talking about.

 

2 : Formulate a list of questions and/or hypotheses 

Now that you know the ins, the outs, and the what-have-yous about your subject, it's time to take what you found out (or already know) and formulate business questions you can use social data to find answers to.  

Say we're working on project for Wingstop, we might want to know things like:

  1. How are people responding to the new Smoke9 flavor that was just introduced?
  2. What sauces are most popular? Least popular? Why? (grab a list from the menu)
  3. Does Wingstop's spokesman, Troy Aikman, drive positive attention for the brand?
  4. How does Troy Aikman compare to rapper and Wingstop franchise owner, Rick Ross in terms of driving attention for the brand?
  5. Outside of the common-sense competitors, what other restaurants are mentioned in conversations about Wingstop?

For the average report we write on an industry, category, or head-to-head brand comparison, we usually start out with 20-30 questions and end up focusing on just 8-10. 

Why? When conducting social media research we have to go off what people are actually talking about online, and often what they choose to talk about is different than what we might want them to be talking about. 

 

3 : Construct a list of relevant and irrelevant words and phrases

With your questions outlined, now you need to select the most relevant data to analyze.

How do you do this?

The most common way is to write Boolean queries that focus in on just the things you're looking for. 

We won't go into building queries in this post, you can read about that here, but in the simplest terms what you're going to do is to create comprehensive list of words and phrases that are relevant to your questions.

wingstop-mispelled.png

As people often use slang, mispell things, and use uncommon language online, your list needs to anticipate all the ways in which people might be talking about your subject.

For example, if we were to build a list for question #1, we would want to include the Smoke9 flavor by name, the hashtag #Smoke9, a more general phrase such as "smoke flavored wings", and possibly a vague phrase like, "new Wingstop wing flavor". 

 

4 : Double check your data sample! Refine. Filter. Include/Exclude.

Your query has hopefully just returned a bunch of results.

Do a quick scan of the topics and individual posts. Do they match what you're looking for? Are there any additional terms, phrases or hashtags that are showing up that you should add to your query to get more results?

Are you picking up additional data that you didn't anticipate? Perhaps an e-cigarette supplier is also using the #Smoke9 hashtag?

You'll want to exclude any mention of their name along with that #Smoke9 hashtag. 

*This is a very crucial step. Not confirming that your data sample conforms exactly to what you're looking for will yield an inaccurate picture of consumer perceptions of your subject.

 

5 : Begin answering your questions and establishing benchmarks

Go through your list of questions.

Did the queries you created to answer each question return enough results? 

Record the basic information to be used for benchmarking. What does the overall sentiment look like for the Smoke9 flavor?

wingstop-smoke9-positive-words.png

Purchase intent? Passion? Emotions? Themes like taste? Cost? Most popular topics?

Are more women talking about it than men? What distinct audiences are being activiated? How does each group feel about each attribute when compared to one another and against the baseline? 

Now dig deeper.

 

6 : Transition from broad to narrow for deeper, less obvious insights

With your benchmarks defined and your initial answers framed up, it's time to take things a step further. 

The idea at this point is to move beyond the surface level information and explore the sub-context within the data for each question. This is usually where the most interesting information is uncovered.

For example, maybe spicy came up in the topic cloud for the Smoke9 flavor.

Drill into this. Is the flavor too spicy or not spicy enough? How many people have responded this way? What groups have responded this way?

What about other flavor characteristics like sweetness, sourness, amount of vinegar? 

Be creative. Stem your initial questions into more specific questions and hypotheses. 

Final thoughts

"This step-by-step method is fantastic. It truly is. You know, I know alotta folks who have been very, very successful with this. It's just really great." -Donald Trump

Social media research can be an incredibly effective way to better understand consumers, but if you don't start with a plan, you're going to have a rough time. 

And nobody wants that.

Just remember to:

  1. Do your initial research, if you haven't already
  2. Create a list of business questions and hypotheses you'd like to answer
  3. Build a list of the words and phrases related to each question
  4. Once your queries have returned results, double and triple check the results
  5. Thoroughly answer your questions and establish benchmarks
  6. Transition from a broad view to much more targeted research

Best of luck!

-Rion


Related Posts: 

Setting Client Expectations With Social Media Research

How Social Media Data is Enriching Market Research

60+ Free Resources for Mastering Social Media Monitoring

5 Templates for Making Social Insights Actionable

 

Check out an example social media research report

New Call-to-action


 

Comments

Request a Personalized Demo Today!

Recent Posts

Popular Posts

Get Our Monthly Social Insights Newsletter

New Call-to-action

Follow Infegy

Receive Updates From Infegy