Social media analysts must constantly sift through enormous amounts of information, and often times, it’s frustrating to deal with irrelevant content that falls outside of the intended research parameters.
I know, I run into these issues on a daily basis. The secret to successful analysis all starts with the search query.
Without an appropriate query, the analyzed data might be irrelevant and the end results inaccurate. Fortunately, through my experiences I’ve come up with a straightforward approach that repeatedly gets me great results. Let’s take a look!
1. Start with a question
What are you trying to solve for? What specifically do you want to know more about? Great research starts with a fundamental business question you’re looking to answer.
When Tylenol’s digital agency set out to improve the brand’s digital strategy they started with one fundamental question: “Who is talking about headaches and migraines online?”
2. Think broadly
There are many variables affecting a product, brand, or subject such as competitors, macro industry trends, unintended uses, and complimentary products. Rather than starting your research with the brand itself, focus first on the industry or category.
For example. In preparation for an upcoming product launch, a major consumer electronics manufacturer was seeking to create messaging and imagery that caught the attention of consumers.
They performed a category analysis using social media monitoring to determine the features that were discussed most often and with the highest levels of passion.
3. Get comfortable with boolean operators
Boolean language can be intimidating at first for users who are not familiar with this kind of advanced search method. However, with an understanding of the fundamentals and some practice, using boolean operators is really pretty easy.
Most social media monitoring / listening platforms utilize the same basic boolean operators, and a few, including Infegy Atlas, can leverage more advanced operators for greater depth. To get you started, we’ll cover the universally used functions.
Include a word OR another word. The following example searches for posts containing the word iphone or the word ipod, but it does not have to have both.
Include a word AND another word. The example that follows searches for posts containing both the words iphone and ipod at least once and in any order.
Do not include a word. The example that follows searches for posts containing the word “iphone” which do not include the word “ipod”. This can be combined with an OR operator to exclude many words, such as: iPhone NOT (iPod OR Mac OR Macbook)
Phrase search operator: “ ” (quotes)
Include posts which contain the exact phrase as specified. The following example searches for posts containing both the word apple and iphone in the exact order specified. This means “apple releases iphone” will not match.
Searches for a word, phrase or complex block within so many word spaces of another. The below example will match documents which have any of the words car, cars or automobile no more than 3 words from the exact phrase detroit auto show.
With an understanding of basic operators you’re ready to start writing great queries!
4. Build your query one piece at a time
Whether you’re focusing on a specific product or an entire industry, the best possible way to begin is by starting small and building up your query. After adding each new term you should also test or preview the results to ensure you’re on the right track.
In the early stages of creating a query, consider the following:
1. Common language
What are the common ways in which your subject is referred to by people online? Coca-Cola, for example, is commonly referred to as simply, “Coke,” but this also raises some challenges, particularly because of drug references.
By not including the word, “coke,” you would severely limit your results. However, by including it you’re also introducing irrelevant content. To get around this you will need to incorporate exclusions into your query (more on that below).
The majority of content collected from short-form online channels has a low writing level, and as a result is rampant with misspellings. Your approach should take this into account in order to obtain the largest relevant sample size.
3. Hashtags and campaigns
What hashtags are being used for the subject you’re researching? To determine this you can run a preliminary search and look at a topic cloud or other data visualization to see what is being used with the greatest frequency.
Because space is limited on some social networks, authors often use hashtags without also directly mentioning the subject or brand. As a result, just adding hashtags can significantly increases sample sizes.
4. Brand / subject bias
If you’re not careful you might pull in online posts that were created by the very brand/subject you’re trying to research. In some instances this might be okay, but if your research is focused on what people are saying about that brand/subject, you’ll want to exclude their own comments from the scope of your query.
With a solid base established, tighten, focus, broaden, or exclude terms
Let’s say your subject is Walmart, but you mainly want to focus on their grocery offerings. If you search broadly, you’ll more than likely pick up things that have nothing to do with what you’re looking for. So, what are words that would help you get you to the content you’re looking for?
Now go further. Put yourself in consumers’ shoes. Ask yourself, how would someone talk about Walmart’s groceries? Sayings like, “just went” OR “bought” OR “paid for” OR “need to buy” are some phrases that are used when people talk about going to the store.
Incorporating terms and phrases that consumers typically use can help cut out noise and narrow into the specific crowd you want to listen to.
The idea behind broadening your query is to find similar terms and phrases used online that mean the same thing as your subject. We’ve already touched on this a bit in the previous section, but there is a lot that you can do in this area.
For example, brands will often times have different online handles for different markets. In fact, Pepsi India has its own Twitter handle, @PepsiIndia, with more than 130k followers. Not including this would color the results in an international brand analysis.
As mentioned before, the internet is filled with misspellings. Adding on to the JetBlue example above, broadening the query to look for mentions of Jet Blue pulls in significantly more data about the exact same brand.
If you’re dealing with a subject that is universally used as another term, or is typically associated with another subject, you’ll want to exclude terms that will pull in irrelevant content.
The brand COACH is a prime example, as the brand name is a common term used to discuss very broad subject matter. Below is a basic example of how to exclude some terms that color the results for COACH, the complete query is more complex.
The NEAR operator is particularly useful for tightening up your research where multiple brands or subjects might be discussed in the same post. Say you want to know what people think how waterproof iPhones are, for reference, they aren’t.
If you do a broad search for iPhone AND waterproof you will likely pull in content that references other phones that are waterproof within a post that also mentions the iPhone, which isn’t what you’re looking for. To get around this, you would utilize the NEAR operator to look at the term waterproof within a specified distance of iPhone, say 15 words.
Finally, apply filters to tighten results even further
Filters provide an additional step after building a solid query to refine your results even further. From geographic location, social media channels (Twitter, blogs, forums, etc.), language, time of day, writing level, and audience segments, such filters are just one more way to ensure you return exactly the data you’re looking for.
No matter how many features or widgets a social media monitoring tool comes equipped with, everything starts with the query.The approach outlined above has proven time after time to achieve great results and should serve as a good starting point.
Just remember to:
- Start with a question
- Think broadly in the beginning
- Get comfortable with boolean operators
- Build your query one piece at a time, and test at each stage
- Identify common language, misspellings, and hashtags
- Remove the brand or subject itself from adding to the conversation
- Tighten, focus, broaden, and exclude terms
- Apply filters to the finished query to tighten even further