Facebook Sentiment? 100% Positive


Facebook is the single most used app out there, the social platform of choice for majority of internet users. It is also great source of data, yet very few know how to extract value from it. In this post we try to highlight different approaches to analytics and we will demonstrate them with an example of Facebook’s sentiment.

On one side I’d like you to picture the broadly accepted approach which uses several commercially available black box solutions, from specialised vendors; products that usually serve only one purpose to thousands – in some cases, millions – of customers. On the other side, imagine custom-built solutions that can completely reflect your individual needs and views of the world. I believe it would look something like this:



It all comes down to mindset

My take on this is that the out-of-the-box solution would probably work well in the beginning of your transformation journey, when your questions are not yet too advanced or complex. But as you progress, as it often happens, your black box toolkit would run out of breath and particular limitations would have negative effects on your delivery and the success of your business. In this sense, every organisation is a unicorn: It has its own and unique data set as well as sets of business problems and objectives. By this definition, your business can’t be satisfied with a generic solution.


There goes your social listening

Social Listening as the successor of traditional news clipping practice and media monitoring has been around for years and it’s widely used. It is, however, somewhat ironic that the by far most important platform in terms of usage – Facebook – doesn’t really allow much room for these tools to do what they should do, which is to scrape conversations of the users in the public places and categorize them into topics or sentiment, so that you as a marketer can figure how your brand is faring against your competitors, what the current sentiment in the market is and which products are the most hated and loved. This was an important change introduced by Facebook about a year ago.


Playing catch-up & why agility is key

Facebook has a monstrous development team and one simply must marvel at the pace of innovation coming from them on both lines: towards its customers (read: the advertisers) and its users. This single fact makes the ‘catch-up game’ for the vendors who build their solution leveraging Facebook’s platform nearly impossible and they often get outpaced by the innovation of Facebook itself. Ask anyone who tried to build a better product than Facebook native ads editor.

One example, look up how many of the available community management tools support Instagram and how long after this type of services were enabled in the API by Instagram did they roll out the functionality. The short answer to spare you the search: Some still don’t, hello future.


Agile approach wins

I believe that building your own solutions, while not trying to reinvent the wheel, is the winning strategy. To demonstrate how, I will show you what we built in less than 2 days of work (15 hours, to be specific).

We don’t focus on building social media analytics tools and we sadly don’t have 100 developers at our disposal like your Meltwater, Simply Measured or Hootsuite might. In fact, what you are going to see was built by one person, who isn’t even a developer. Take a minute to digest this if you need to.

The key reason why he was able to make this happen so quickly was the choice of the underlying technology. For this we used our own Keboola Connection, a smart data integration and preparation platform which provides fully scalable solution to our problem. If you are not a very technical person, think of Keboola as a Lego brick for data solutions and products.

Easy building blocks are configured quickly to consolidate available data from Facebook, Instagram, Twitter and other platforms to then deliver the insight you want wherever you want. For the front-end visualisation we set ourselves up with a dashboard housed on our cloud-based server using javascript libraries to visualise the data in the way we want.


Competitive overview and basic vanity metrics

Why do we call these vanity metrics? Because they have no direct relationship with your business’s bottom line. Disagree? Tell us in comments, please.

Simple report capturing changes in audience, the speed of growth or decline and total numbers. You can see the full dashboard here, but when you do, bear in mind that this is still a working prototype, not yet a polished marketing product.

facebook social community overview

Custom definition of metrics

The whole idea of custom solutions is that you don’t need to subscribe to the definitions you are told to live by, imposed on you by the black box tools; in a custom environment you can define these yourself. The common problem may be the definition of engagement, i.e. Are all Facebook interactions equal? Is a 10x share more valuable to you than 1x like? Easy, set these metrics up yourself just like we did here:
Screen Shot 2016-05-25 at 11.50.39

Screen Shot 2016-05-25 at 11.50.17

Facebook sentiment

Going back to the innovations, Facebook recently introduced ‘reactions’. When ‘Like’ isn’t enough, you can now use other emotions – both positive or negative – to express yourself.

reactions 1

This obviously is just the beginning for Facebook and things like video comments are coming soon, making your analytics efforts harder than ever. Now these different reactions represent very different emotions that may, in some use cases, be useful to your business (think product launches, PR crises). Since we are working with the raw data and we are collecting these reactions for each post for our control group of pages we defined, it is very easy to build the following visualisations.


Overview of sentiment in the last 30 days on selected pages.

Communication Strategy

In the example below you can see perfectly clearly how some brands are simply utilising negative (PETA) or positive content strategy (See Australia).

sentiment spider chart

Methodology – How did we get here?

  • • We collected all interactions on all posts. For non-techies, this is what happens in Keboola Connection: We seamlessly connect to Facebook API and collect all the data points we are interested in but not more (data volume makes a difference).
  • • These are still early days of the reactions feature for Facebook, so in the control group about 92% of all interactions were ‘Likes’. This will change quickly, no doubt.
  • Then we categorised these interactions into positive or negative and assigned them values; extra points for love and marking angry as more strongly negative than sad.
  • The data, once organised and structured in Keboola Connection and all metrics calculated as we like them, is then loaded to our visualisation layer (in this case to the custom dashboard, but we could easily use any BI tool available on the market since Keboola is built to integrate with those). If you are using Yellowfin, Qlik, Tableau or Power BI, we got you covered.


Business implications

 Social media provides incredible wealth of data beyond the vanity metrics and if your organisation isn’t leveraging this golden mine, you’re missing out!

If you’re still reading this and you’re like, ‘OK, so what now?’ rest assured that this is not the end of the journey, but rather the first outline of where the journey might lead. We showed you an example of smart data consolidation and custom interpretation and we think it is a very promising start, especially once you start applying advanced analytics to this data.


We are all about data integration and we believe that data has much higher value for business when connected. The question is, how far do you dare to take it?


Click here if you would like to see the full dashboard.. I am pretty certain this topic may cause some controversy and I would love to hear what you think. Feel free to send feedback to pavel.bulowski@keboola.asia

Did you like it?

Let us know so we can improve

0 Vote DownVote Up