This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Find out more here

FEATURE17 February 2017

A hypothesis-driven approach to social media insight

Data analytics Features Media UK

Speaking at yesterday’s Social Media Research Summit, organised by the Market Research Society (MRS), Pulsar’s co-founder and vice president of product and research, Francesco D'Orazio, demonstrated how to use social media to validate a specific research hypothesis. 

Social media data has transformed the scope of research. The availability of 11 years of Twitter data at the touch of a button is just the tip of the iceberg, said D'Orazio, who went on to extol the benefits of both the granularity of publicly available data (Twitter and Instagram) and the aggregated nature of Facebook and LinkedIn data. 

But the sheer quantity of interactions can make analysis challenging, which is why D'Orazio believes in the value of an emerging research approach: using social data as a tool to validate specific research hypotheses, rather than as an exploratory tool. 

The traditional ‘emergence’ approach, said D'Orazio, relies on "seeing what crops up" and is based on keywords and stories. The ‘hypothesis’ approach involves framing data: looking at it through the lens of a specific question. In short, the hypothesis approach shifts much of the analysis to before data collection, rather than afterwards. 

The key advantages of this approach are as follows: 

  • Keeps the data collection focused and reduces ‘noise'
  • Makes analysis faster, more structured and standardised
  • Makes it easier to replicate results across teams
  • Makes it easier to integrate social data with third party sources such as surveys

D'Orazio took the audience through the process of this method: from client brief, to hypothesis, to data query, to insights. 

Translate

In order to move from the client brief to a research hypothesis, the researcher must break the brief down into three elements: What is the business objective? What is the target audience? What are we trying to understand? 

This is then further broken down into two elements: Who is the audience you're trying to reach? And what type of behaviour and moments should be investigated? 

The researcher can then create a hypothesis for each of these elements that can then be investigated in the data. The more focused this is, the better, said D'Orazio. It should be considered as a frame for looking at the data, rather than simply a theory to be validated. 

Transform

The next step is to transform the hypothesis into a study definition. A hypothesis will contain – and be related to – a number of elements: language; behaviours; attitudes; moments and occasions.

Transforming the hypothesis means defining the ‘signals’ to look out for, such as audience demographics and a list of terms (and sub-terms) to look out for.

For example, if you're investigating fast food consumption among UK/US millennials, a hypothesis could be that the UK/US millennial audience buy into authenticity and not the fast casual proposition. The terms to look out for could then be: ingredients, pairings, sustainability, price, health, occasions, behaviours and quality. Within quality, for example, there are terms to look out for such as: premium, chef, better, best, amazing, etc.

Test

The last stage is to test the study outputs, including comparing the results across demographics for context, looking at the language used and how it compares to the hypothesis. 

This can offer insight into how consumers talk about a category, including how terms are conflated or distinguished, what behaviours and attitudes relate to the category (for different demographics), and what types of conversations people have. 

This can either validate the original hypothesis, disprove it or drive completely new insight, D'Orazio explained. 

1 Comment

3 years ago

Wow! What an excellent, insightful article. With social media monitoring softwares changing all the time, we are able to do more and more with social data and weed out the noise. Location based monitoring and digital image searching are both available and are a great additional resources for researchers and marketers.

Like Report