The 4 YouTube Analytics Tools You Need
By Joshua BoydJan 24
Published September 6th 2016
Social media research can surface consumer insights that can be difficult and expensive to find in any other way. The volume of conversation on the web gives this research method a unique ability – uncovering qualitative insights on a quantitative scale.
For many people, however, it can be daunting. There is so much data: millions upon millions of conversations happen online every day. If you don’t know how to refine what you are listening to, you’ll drown in the data.
It’s also easy to rely on the metrics that are constantly produced on social: likes, followers, fans and so on. Knowing how to cut through the noise to find the information that can drive business decisions is a real skill. I recently spoke to Bex Carson, Head of Brandwatch’s Research Services team, to get an insight into best practice when it comes to social media research.
Uncovering insights requires a certain approach. You need to move beyond the simple social metrics to uncover robust insights that can recommend a real business action. The problem can be knowing who to listen to, how to spot a trend, knowing what’s significant and putting your findings into context.
Many people will start with the metrics they know about. They then think about the questions it is possible to answer with those metrics, and develop a research plan based on that. A good social intelligence tool has the flexibility that renders this way of thinking very restrictive – it starts with what was already known and therefore limits the possibilities of what can be discovered.
The foundation for good social media research is asking the right questions. If you start with a bad question, you’ll receive a bad answer.
Start by forgetting any concerns about how you are going to conduct the research. Focus on the problems and questions you need to answer, without thinking about the methodology. Once you have the question, you can work on the methodology.
With Brandwatch there’s a way to answer pretty much any question. The joy of social media is its flexibility and Brandwatch Analytics has been built to be flexible to match that.
The questions need to be specific. They need to be able to deliver an answer that can be acted upon. Examples of this sort of question might be “What do women in their 50s want from a fashion brand?” or “Is our content resonating with our target customers?”
Spend time developing questions, and brainstorm with others where possible. Consider the following:
There are two main approaches you can take to answering your questions. The first involves uncovering metrics that answer a defined question. The second is more of an exploratory approach, discovering insights as you work through the data. Where time and budget allows, a combination of the two approaches will generally reveal the most interesting and robust answers.
The best social media research that produces the best results will combine some elements of both. That way you know you are investing time to get definite answers to specific questions, but you’re leaving yourself open to discover something new about your audience.
For this type of research, you are looking for an answer to a specific question. You need to identify specific metrics that you can bring together to answer that question. By using the power of segmentation – Brandwatch’s Rules, Categories, and Tags – you can arrive at specific answers.
This method is great for answering defined, objective questions. For example, if you have run a campaign and want to know if it drove spontaneous awareness, you can look at the volume of discussion for both the campaign assets and the brand baseline over time. These relatively simple queries will result in metrics that uncover the success of your campaign.
This approach is exploratory, listening to the data to discover the story within it. Think of it as social media ethnography. You can arrive at a fuller understanding of your consumers, uncovering different audience groups as you move through the data and spot trends.
This methodology can be a challenge. You need to enter the project without preconceptions and ensure you don’t draw your own conclusions. It’s a harder way to conduct social media research – using open listening to find underlying themes.
The benefits are plentiful, though. First, you are closer to the voice of the customer. There is no substitute for listening to your customers verbatim, understanding the nuance and sentiment of their conversations.
As a human analyst, you can understand and categorize things that a machine simply can’t. Someone might be talking about the same topic even if they don’t use the same words. Reading through the conversations, you can find people might talk in a way that you hadn’t thought of, and use that to enrich your research.
Broad topics and themes often drive more interesting results than brand queries. You want to analyze of a type of conversation rather than a particular brand. Within the conversation theme, you still need to define the problem clearly so you can create a targeted query and reduce noise in the data.
As an example, let’s say you wanted to know what people who have seen the movie thought about the new Ghostbusters film. If you just wrote a query on Ghostbusters, there would be a lot of conversation around the all-female cast, or the abuse that Leslie Jones received on Twitter.
By writing a query that includes personal pronouns, mentions of cinema brands, people saying “just seen/watched” and so on, you will have a much cleaner data set without manipulating it too much.
This can be a really important way of revealing the interesting, more concealed consumer insights. Many topics can be dominated by a few obvious and popular themes, so that the insights get lost in comparison.
You want to surface underlying themes with a unique value, rather things you can see easily on a topic cloud/trend line. To do this, you can look at you initial query results to identify the popular themes, and then exclude those mentions by using tags.
The data that remains lets you see what else people are saying beyond what you already know. It can also be useful to remove retweets, leaving you with only the original conversations.
While automation can be a useful tool, the type of analysis we are talking about here is human led. This means you need a manageable dataset. It needs to be large enough to be representative of the mentions, but small enough so it can be read completely, and in detail. Mark off the mentions as you work through them to ensure you cover your whole sample.
This allows you to undertake the next step…
There is no right way to do this step, and it will largely depend on your dataset. This is where you go on a journey of discovery, hunting down the insights in the data. You want to start grouping themes and topics.
You can start by creating categories that you think you are going to find. To go back to the Ghostbusters example, you would expect some people to say they “loved it”, others that “hated it”, and everything in between.
It’s important to remember that this is just a starting point. You need to be completely data-led. New themes will emerge as you work through the data, so you need to add these.
Human analysis allows you to have nuanced emotional categories, such as anger, frustration, humor, joy and so on. A human analyst is able to pick these up where automated processes would not be able to.
You can also set up categories for author type, to understand the different conversations among different actors.
Brandwatch allows you to group categories together, so each emotional or author response can form a sub-group within a parent group. This can prove useful during your analysis.
Crucially, you want to liberally add categories as you read through the data. At the end, you might find there were only a couple of mentions of one category and decide to delete it. It’s better to do this than be conservative and realize you missed an important topic of discussion only once you reach the end of the dataset.
Now you need to dive into the charts and start analyzing.
The best way to start is to simply create a standard chart component and apply filters to the data. You can start looking at your different categories, moving the data around to spot interesting patterns.
Some of the most useful insights come from crossing your parent categories. So if you have author type on one axis, and emotion on the other, you can start to see different emotional responses from different author types.
The fundamental tenet of social media research is noticing a difference between things. You don’t know what good looks like without seeing the bad, but you are looking for a significant difference. You don’t want to recommend sweeping changes because of a tiny difference in the data. Be curious and keep exploring.
If you want to be interesting, be interested.
If you notice a difference, start digging into it. Then dig some more. Keep digging until you have read something that helps explain why that difference exists.
You can’t create a suggested action based on a piece of data if you don’t understand why that piece of data is the way that it is.
First, remember that you are telling a story. With any story, you need to consider your audience.
There may be several people or departments that are going to read your report. They won’t all have the same amount of time available to read what you have written, and they might not all be interested in the same things.
While it’s a good idea to start with a methodology to give the report transparency, hit them with the key findings and consumer insights straight away.
Then, as you continue through the story of the data, you can explain each insight more thoroughly with charts, graphs, and analysis to back it up.
Having insights buried in a 27 page PDF is frustrating and risks your hard work being overlooked.
Finally, remember that you’re not really telling your story; you’re telling the story of the people you’re listening to, so make sure to include the voice of the customer. Real examples of social posts, with real customer profile pictures, will bring your social media research to life.