As the UK’s first Dr of Social Media one of the questions I get asked the most is ‘how do you analyse social data?’It is a simple enough question.

It is a simple enough question.

It’s also massively relevant to my work [and I’m guessing, yours too].  You see, I analyse social data to find out how people make decisions and use this insight to help my clients position their marketing spend.

So, being able to answer this question is critical…

But it’s not easy to answer as I’ve found the specifics of the analysis change on the question being answered, the tool you are using, the business and how people talk online.

There are quite a few moving parts that need to be considered, and it makes it difficult to come back with an answer that is suitable for everyone.

You’ve probably been reading that and thinking ‘same old answer, different day… Why does no one give me something tangible that I can work with?

It’s OK.  That washed out ‘it depends’ answer doesn’t sit well with me either.

So, lucky for you that I’ve manned up to provide details in the process I use to analyse social data.


Or if you’re not ready for that level of trust let’s talk a little more.

We’re starting with the elephant in the room – social tools.

The inevitable question about the social tool you are using because the tool gives you the insight.  Right?!

Is it not about the social tool?

I think by now we all know that buying the latest social listening or social intelligence tool is only the first step in being able to generate actionable insights from social data.  If you’re new to social data analysis, you will learn this soon enough.

We’ve all seen the demos and how they manage to get insight easily from them but when it comes time for us to do the same its considerably more difficult and you want to pull your hair out because you’re struggling to get something that is good.

You’re not alone, it has happened to me too.

The tool you are using is the first step in a long process to generate actionable insight from social data.  Let me say that again… The tool you are using is the first step in a long process to generate actionable insight from social data.

If anyone else tells you otherwise run for the hills.  Leave.  And, don’t go back.  Ever.

This is the type of thing I see a lot….

Hi @DrJillianNey , I think what you are doing is quite interesting. I am curious to find out the source of the data you analyze?

— Anisha Surana (@AnishaSurana) September 4, 2017

Everyone starts by asking about the tool you are using.  But the most important thing to actionable insight is the steps you take to analyse the data.

The tool you are using and the functionality of this tool ultimately dictates the approach you use to analyse the data.  To put it simply, the tool you are using dictates the functional process of data analysis – how you segment the data for analysis.  The tools are all a little different and you need to get used to their processes and their Boolean.

I think the differences between the tools may be one of the reasons that people find it difficult to answer questions about how to analyse social data.

But what I’ve found through the last ten years of analysing social data is that there is a much larger process you go through before you even touch the data.  It’s this process that gets you prepared to analyse the data and specifically answer the questions or hypotheses that you set at the start – i.e. generate actionable insight.

I’m going to show you the process that I use in my own consultancy with my private clients and the process that I use to train my team.  This is how we generate actionable insights from social data when other people seem to get ‘nice to know’ insights.

What My Process Does and Doesn’t Do…

It’s important to note that this process is not for analysing interaction on brand owned content.  The work that I do goes far beyond that.  So, if you’re looking for analytics you’re are in the wrong place.  I’d recommend reading the book Ask, Measure, Learn.

If you want to be able to understand customer behaviour or have deeper questions about your audience then this is the place for you.  Keep on reading.

There are nine steps in my process.  With steps seven being when you finally analyse the data.  You may laugh at the fact that you have to go through seven steps before you analyse the data.

That’s cool.

This is new and you’ve probably have not heard someone say this before.  But if you trust the process, I guarantee that you will be better prepared for the analysis and you’ll ultimately generate better insights.

I also have to admit that I do have other ‘sub sections’ that I use depending upon the type of question we are trying to answer.  I could not put all of these in this download because there are just too many options – I’ve made over 50 social metrics that get to the heart of customer behaviour.  There are too many to mention all at once.

The process I’m about to show you will get you started and thinking more about the analysis you are about to undertake.

Nine steps may seem extreme but for me, the important thing is not to rush right into the analysis.  If you rush in you’re likely to analyse everything and anything, it’s a scatter gun approach.

I call this a ‘Kitchen Sink Analysis’.

The Kitchen Sink Analysis

I call it the Kitchen Sink Analysis because you are effectively trying to analyse everything and you are throwing everything at it to try and make sense of the data – you throw the kitchen sink at it.

The Kitchen Sink Analysis is typically where most people start out.  I know I did.  But I quickly realised that it wasn’t giving me the best output.

The good news is that you don’t need to do that.

By structuring the analysis and knowing what you want to get out of it and what you want to be able to do with the insight once you have it, you’ll start to get better results.

Here goes…

Don’t forget to download my free interactive worksheet that takes you through each step so you have the answers to get your analysis started.


Step 1: Define the Purpose

To not throw the kitchen sink at your analysis you need to know the question you are trying to answer and what you want to be able to do with the insight once you have it.

By answering these two purpose questions you lay the foundations of how to approach the analysis and what the insight needs to do.  You prepare yourself to focus on the data that matters.

If you’ve been analysing social data for a while then you’ll know that there is too much ‘white noise’.

This noise is a total distraction and pushes you into a kitchen sink analysis.  To overcome this problem, define the question.  And, know what you need to do with the insight once you have it.

The common questions I get asked to answer are:

  • What do people say about my brand?
  • How do people feel about my brand?
  • How can I sell more?
  • Who is my customer and how do they behave?
  • Can you help me win a pitch?

The common uses for the insight include:

  • Create a marketing campaign.
  • Create a content marketing ideas strategy.
  • Develop a new product or app.
  • Brand positioning.
  • Conversion rate optimisation.

Step 2: Deconstructing Your Question

Now you know what you want to answer and how you want to be able to use the insight, you’re now ready to deconstruct your question.

This step is almost always forgotten.  It might be that people don’t realise that you need to go this deep or that they feel like the data should guide their path.

Don’t fall into the trap.

What you want to do is to take your question and break it down into smaller segments.  This is extremely important because it helps you to break down the data into smaller relevant parts that you can analyse properly.

To get actionable insights from social data you cannot rely on the ‘topics’, ‘trends’ or whatever the tool calls their automated wordclouds.  This automation is generated through volume and you might miss the topic that’s important to you.

Just look at what a topic wordcloud says about ‘Green Bubbles’… Don’t use this automated ‘insight generator’, it really doesn’t give you insight.

Another example of this would be when you analyse festival data, you tend to find that the artists make up most of the chatter between 60% and 90% of the chatter.  It’s just common sense that the artists will make up a lot of the talk but it skews your analysis.

Say that we want to understand the festival experience we want to know about weather, camping, travel, food, drinks, toilets, other attendees, atmosphere, the at home experience and the at festival experience…

There’s a lot of things that make up the festival experience that gets drowned out by the talk about the artists.  By deconstructing your question, you spend time thinking about the different elements that make up your larger question.

Step 3: Keywords and Phrases

Now that you have the question and you’ve deconstructed it properly you are now ready to start getting the keywords, phrases and language.

We all know that social listening is dependent upon you creating a ‘search query’.  This is the stage where we get all the information we need to make up our search and segmentation queries.

This includes all the hashtags, keywords, nicknames etc etc.

One of the things I do when I start working with a new client is to review their current tools and set up along with their dashboards and other outputs.  What I tend to find is that not enough time has been given to getting the keywords right.

You need to know how people [your audience] talk online.  If you do not get the right words and phrases you will not have all the data that you need.

To do this stage properly go back and look at how you have deconstructed your question, now spend time looking at the language around each of your areas and get all the keywords, phrases and everything else you need to gather the right data.

Step 4: Create Your Query

After you have all the phrases and keywords you can now start to construct your search query.

The biggest pitfall I see at this stage is that the search query is too loose.  Go back to your original question and think about the best way you can gather the data to answer that question.

Too much data is not good.  Too little data won’t help.

For example, if you only have a brand name you will get everything about that brand.  That’s fine if you are looking at brand equity but say you only want to understand purchase intent.

You’ll have too much data by looking at a brand name alone.

If you have too much data you can be easily led back down into a Kitchen Sink Analysis.  Keep focused on the purpose of your analysis.

Query writing at a high level is complicated but it is a skill that can be learned.

Ask your social listening tool for a bit of help if you get stuck.  This is also a topic I’ll be covering in more detail.  You can sign up for updates HERE or search through the blog.

Step 5: Develop Segmentation Criteria

Now that you have the search query written and the data pulling through your tool you now need to start segmenting the data.

These segments are related to how you deconstructed your question in step 2 and the phrases and keywords you collected at step 3.  Go back and start to create your segmentation queries.

The social tool you are using will have different terminology for what you are doing.  For example, Brandwatch uses rules and categories for this step.  Whereas Crimson Hexagon uses quick filters.

If you don’t do this step properly you become reliant on the tools ‘volume automation’ and you’ll never get the insight you are looking for because it’s hidden.

Step 6: Unknown Segments

Because social intelligence research is driven by naturally occurring social media conversations, you cannot account for every eventuality.

The process I’ve outlined will help massively to prepare you but you’ll always find a chunk of data that you’ve collected that doesn’t fit into any of your segmentation criteria.  These are our unknown segments.

The unknown segments are where I’ve found that the magic happens.  They are the segments that you didn’t know about.  They are probably where the insight that you hadn’t considered is hiding.

Like I said, I find that this is where the magic is.

To do this you need to find the data that has not been segmented.  The tool you are using will dictate how you go about this.  Ask them for help.

What you are looking for in this data is to find a pattern and create a new segment or change up the segmentation criteria in the segments you had already created.

Like I said, this is all down to language and if you’ve not thought of all the words and phrases properly you may miss some data.  This is the stage where you refine what you have done before.

Step 7: Analysis

We’re now finally ready for analysis.

Because you have segmented the data you’ll find the data easier to work with and analyse.

The question that you are trying to answer will dictate the analysis.  I advise looking for the context of the communication – discourse analysis on all the comments.  You need to find out what the comments mean, the context around the keyword, and what it tells you about that segment of your ‘deconstructed’ question.

I talk more about this step in my interactive PDF and get you to map out your thinking…


Step 8: Interpretation

Your work doesn’t stop at analysis, you need to interpret what this all this means.

I often see too many people putting easy to measure metrics in reports and trying to pass that off as insight.  It is not insight.

You’ll have an easier job at this step if you have worked through the process that I have set out here.  To explain…

  1. You know what you want to answer and what you need to do with the insight – this helps you to identify what is important in the final interpretation and report. You don’t measure the easy to get things, you prepare to answer the question properly.
  2. You’ve deconstructed the question and segmented the data properly – this lets you know the volume of conversation in each segment.
  3. You’ve analysed the data to find out the context driving the conversation – this lets you know the specifics.

This step is about putting all of this together and answering the original question.  If we go back to the festival we may be able to say how people prepare for rain and what they need from the festival site.  We can say what type of food is desired and why, and so on.

There is an art to interpretation but again it is a skill that can be learned.

Step 9: Additional Analysis

There are areas that have not been considered during the other eight steps.  For example, who is talking or where they are.

You might want to understand festival experiences from people in the UK compared to people in the USA or Australia.

Follow the same system but add in the extra segmentation.  In this case demographics, psychographics or location.

What’s Next?

This is the overarching process that I use when analysing social data.  As I said before there are other things I do at each stage depending upon the question we are answering and the metrics we use – this is done on a case by case basis.

If you want to learn more about these sign-up to my People Science updates HERE or if you want to go deeper you can request an invite to my membership community HERE.  If you have a specific question then you can email me and my team on

Your immediate next step should be to review the searches and dashboards that you already have running, use this process to find out if you could be doing more and amend what you have accordingly.

Then the next time you go to do any analysis follow the process…  Don’t forget to download the PDF!


 And, do let me know how you get on!