Ask the right question

Making sense of social media is not an easy task. There is a vast amount of data generated on a daily basis (Twitter, for example, has 500 million new messages are posted every day), and even if that data is harnessed and processed it’s easy to get lost in the variety of content.

Think about Twitter for example.

Say we’ve built a data set of tweets of our interest and we’ve captured who posted these tweets. We call this a data set and not a database because a database assumes that a bespoke software application is set up to store and manage data. In this example we don’t make assumptions how it is handled; it can be stored in text files or even in an Excel spreadsheet.

The features of this data set contain:

• the text from each tweet
• the time when each tweet was posted
• the handle of the Twitter user who posted it

It’s not an easy task to draw conclusions even from this simple data set. We can ask ourselves questions, like:

• how can a data set like this help my business to attract more people in my social media feed?
• are there any trends that I can identify to learn about topics Twitter users are interested in?
• what are the related features in this data, and what features are just noise that I can ignore?

Asking the right questions up front is probably the most important step in any analysis. To come to the actual answers that can drive business decisions, questions have to be broken down into smaller, more specific questions.

Identify the smaller steps that lead to conclusion

To illustrate the process, let’s take an example.

A business wants to generate more leads by increasing the number of social media users following their account so whenever they post new content all the followers will be updated.

1. The high-level business goal here is to increase the number of followers by a certain number (say by 25% over the next 6 months).
2. The strategy has to be defined to attract more visitors to the business' social media profile to increase the followers number even by a small degree.
3. The strategy must be backed by data to make sure it is as realistic as possible.
4. Using the data set described earlier, the key metrics need to be identified. It is generally good practice to let the data drive the analysis. The analyst must also look for signals that can guide her instead of being driven by prior assumptions.
5. By applying some data exploration techniques (such as calculating basic statistics like mean, simple number of occurrences of certain hashtags etc.) some conclusions can be drawn. This exercise does not need to be technically complex, although using advanced techniques such as machine learning can certainly help to understand the data better.

After initial findings the complexity of the exercise can be taken further.

Such an analysis, even by using basic statistical techniques, can reveal invaluable information about social media interactions.

The business might find that a hashtag they anticipated using might have a use they do not wish to be associated with. An example of this case is that the Twitter hashtag #UCL could be mistakenly thought to mean University College London while in fact it’s mostly used to refer to UEFA Champions League.

The business might also discover new hashtags or frequently occurring terms in those tweets that are of interest.

By avoiding certain keywords and using others the business can find itself reaching out to a broader audience which then increases the chances to attract more users to their social media profile.