So, you’ve decided which monitoring tool you’re going to use to listen to online conversation (Brandwatch, of course). Now what?
You’ll already know that you can get loads of data from your monitoring tool, so here we’re going to discuss how to analyse that data. Y’know, so you can present all that insightful data in a meaningful report.
This is just a snippet of the advice we have on offer – for a more in-depth look at report writing, have a look at this.
So, it may be you who is writing the report, or someone else in your team, or you may outsource it to be written externally. The reasons behind who produces your reports could be down to your budget and resources, your goals and so on. Bear in mind that analysis from an outside source could be more objective, but that an insider may have more of an understanding of your industry.
There are different types of reports, but they generally fall into one of the following categories:
- A brand audit
- A competitor audit
- A research report
1. Initial steps
You need to consider the following before you start:
- Who: What your report shows and explores will differ according to who is going to be reading it. For example, is the report for board members, department heads, other employees, third parties or other teams?
- Why: In order to narrow down exactly what you are going to report on, you need to decide why you are producing the report in the first place. What are you hoping to find out or get from the report?
- What: If the scope of your report is too broad, it will be very difficult to get interesting insights from the data. A generic goal will make it very difficult to produce an interesting report. For example, defining that you want to find out “is there demand for bacon flavoured ice cream?” is clearer than “what do people think about bacon ice cream?” Monitoring tools can provide you with lots and lots of data – without any direction as to what you want that data to answer, it can be overwhelming and difficult to segment and understand.
- Where: Do you want mentions from across the world? From a specific market? From a specific website or group of sites?
- When: What date range do you want to look at? Make sure it’s a long enough period to make the results representative. We find that a date range of 3-6 months is usually sufficient, though sometimes a year if we are analysing long-term trends. However, if you are tracking a specific, short-term campaign, then you might prefer a shorter time date range.
Once you’ve considered all of the above, it is helpful to write a brief before starting on your report. This is obviously particularly important if you are outsourcing your report writing, as they will need a clear idea of what you expect the end result to be.
2. Getting started; finding your data
Once you’ve written a brief, you’ll need to then go about collecting the data you want. In Brandwatch, you find the data you need by writing Queries (a form of complex search string).
Your query is important; it is what everything else in your report is based on. Bad query = bad, irrelevant, useless data.
That’s why we give you a wide range of Boolean operators to use within your queries, to make them as accurate as possible.
Some queries are very simple: if you are searching for a brand or product with a very distinctive name (Nescafe, say, or Logitech), then clearly this is going to be much simpler than our classic example, the phone network Orange or the pharmacy Boots.
You may take the approach of including relevant terms in the query, or excluding irrelevant ones, or – most likely – a combination of the two. The process should be one of iteration; use the test search to see what your query is bringing back and continue to exclude and include terms to refine your data.
Bear in mind that there is no such thing as a perfect query, though; aim for roughly 90% or above accuracy.
3. Lots of lovely data. Now what?
Now you’ll be wanting to segment your data to get some juicy insights. Firstly, it’s a good idea to familiarise yourself with the data a little, so have a browse through the mentions to get a feel for the conversation.
It’s normal to have preconceived ideas about what the data is going to show, but try to keep an open mind.
You should find from reading mentions that certain topics or trends emerge regularly. Use this insight, along with your brief, to decide how to segment your data. There are many different possible ways, for example:
- Topic: Use categorisation to group conversation into common topics. For example, if you were analysing cars, you might want to categorise by common topics of discussion such as Price, Design, Comfort, Engine and so on.
- Brand: If you are researching a type of product (e.g. ‘tea’) you might want to segment by the brands mentioned (e.g. Tetley, PG Tips etc).
- Author type: You might want to take a look at the kind of people talking about your brand/topic – are they advocates, prospects, competitors etc.
- Mention types: You could consider what type of mentions you have – complaints, compliments, questions, leads, reviews etc
And so on.. we could go on, but take a look at your data and you’ll start to find different ways of segmenting your data.
You can manually categorise your mentions by looking at a sample (400 mentions is the magic number for accuracy), or use rules to categorise them if the topic allows it. More about using categories and rules here.
4. Analyse, analyse, analyse
Now that you have a clean data set you can begin to answer the initial brief. Some common ways to report on the data are:
- Volume over time
- Volume by site type
- Top sites
- Volume by category
- Sentiment by category
When you are reporting this analysis, always keep in mind the ‘so what?’ question. Think about what these findings mean for your brand or business.
Make sure you are answering the questions posed in your initial brief, but don’t be afraid to work outside of the brief if you discover something interesting.
Want more help?
You can find a more in-depth guide about writing social media reports here.
If you’d like to take a look at some reports that our insight team have put together following the methods outlined above, take a look here.