„If you can’t make me love you, at least make me laugh”. Creating valuable KPI from Facebook reactions
Around three years ago, since publishing of this post, Geoff Teehan, who was a Product Design Director at Facebook, published an article about new reactions. Team’s implementation sticked and now we enjoy emojis in our daily social media scrolling. They impact the way we interact with the platform, but for business owners, they may serve as valuable customer insights source.
- How to map Facebook reactions with actual well-researched emotions
- What are the basic principles of sentiment analysis and how to run it?
- How to implement the sentiment scoring as an easy-to-use KPI
in your day-to-day work?
I don’t want to go into details how the reactions work, because you already know that from your daily experience.
The only thing it’s important to mention is that currently we have six different types of Facebook reactions:
That’s it. You can also add share and comment, but these are more of the engagement, rather than emotional metrics - we’ll go in to that later.
You may ask: „Are six reactions enough to cover people emotions?”** **
The answer is: yes, majority of relevant ones. Facebook team did some research and they were actually pretty good in mapping them. They have analyzed popular comments and realized these are the ones that are the most common in Facebook world and it worked. You can read more about it in Teehan’s Medium article.
However, we can also map the reactions using more scientifically-proven methods. Paul Ekman, American psychologist from University of California San Francisco, world-class expert and pioneer in emotions research proved in his paper from 1972 that there are 6 basic emotions (link here):
We may try to map some of them together:
Legendary „Like” can’t really be mapped. In theory, it can mean: „I like it, I’m happy”, but also: „thanks for raising this issue”, or „I’m notified”. The reason is quite simple.
You have probably seen people liking the information about somebody’s death. It provides limited emotional value and should be categorized as an engagement, rather than emotion.
Going further, we can come up with two buckets of Facebook reactions, emotional and engaging . We will focus on latter here.
Initially, when I was planning this article, I wanted to include „Like” as a potential positive reaction, but after reading the article of Chris Pool and Malvina Nissim from University of Groningen, Netherlands, I changed my mind (link available here). These two researchers trained a machine-learning, pattern recognition model based on comments from media broadcasters (e.g. Fox News, CNN) and figured out that “like” is not emotional enough
Similar view was also proved by Florin Badita. It’s one of the newest articles in the area, published in 2019 (link here). Author proposed to subtract “Like” from comparison to get the essence of emotions. Similar thing happened to Wow, as it is not emotional enough to drive significant sentiment power.
To conclude, research papers and articles proves that Facebook reactions are good indicator of actual emotions. The only problematic ones are “Like” and “Wow”.
In general, it’s an application of quantitative scoring to text data. For example, if I say:
„I was satisfied with overall customer support, but the product lacked its features”,
then it’s really hard to say what I really feel. I’m positive about support (one category) and negative about product features (second category). I can also add a quantitative metric of +1 for customer support and -1 for features. In this case the sentiment is 0.
One sentence may have many sentiment expressions, e.g. Amazon reviews. In such situations, customer feedback will be shared by many discussion categories. I strongly recommend running such exercise once, either remotely or using natural language processing algorithms. There are many tools that allow you to do so as well. The links I’ve provided before also share methodological input if you’re interested.
However, in our example, we want to zero in on much simpler thing. We want to use Facebook reactions as an applicable KPI without further categorization (even though it’s possible to do later on).
It’s super simple. I came up on this idea, while I was scrolling through feed and realized no one has done it before.
In general, you need to download the list of your posts and their reactions. After that one just needs to apply this simple measurement:
Positive: (+1) for Love and Haha
Neutral: (0,0) for Like and Wow
Negative: (-1) for Sad and Angry
Once you have it, you deduct % of positive reactions from % of negative reactions. Net sentiment is achieved, which is our main KPI.
It’s best to show the concept of the analysis on the graphical presentations.
I have downloaded dummy data from here.
In the following section there will be two types of analysis presented: 1st one focusing on posts effectiveness and 2nd one presenting more high level view of overall sentiment trends.
In this analysis, you need to aggregate the reactions into one simple dataset. Here is how it should look like:
Calculate shares of your positive and negative comments, subtract them from each other and here you go. Now you better understand effectiveness of your posts as you have access to net sentiment scoring.
From the substantial volume of reactions, it’s much easier to understand if our post was liked or hated. At the same time, it’s now possible to double-click into the ones that resulted in poor customer feedback.
You can use this analysis to:
- Better track your campaigns and communication results, one by one.
- Identify the trajectory of your discussions. It’s easier to see if you’re moving into right direction or not.
- Deep-dive into posts that drive the best sentiment or find the ones that are not liked at all. It’s easier to scale up good content or remove the wrong one.
- Run qualitative analysis on actual comments and categorize them properly, so you generate richer customer insights.
Facebook reactions data have many other use cases, but there is additional one that I personally like. You can easily use this to create a very simple, yet powerful managerial dashboard for this social media channel.
By plotting volume of positive, neutral and negative comments, marketers and community managers may quickly understand whether their posts generate higher discussion buzz and what is their sentiment. It’s also valuable to understand the engagement vs. emotional affinity.
As you can see on the chart above, audience reacts on posts, yet most of their reactions are neutral (most probably “likes”). In this case we can think of boosting the emotions. I think I don’t need to remind you that emotional reactions also increases the reach and visibility of your fan page.
Such analysis is valuable from various standpoints:
- You can see the trends in a longer time frame
- It can be used as a part of social media team dashboard
- It presents the discussion spikes next to sentiment, so you can see the audience engagement
- It’s easier to understand if our audience is emotionally attached.
- Potential for deep-dives to identify most engaging and emotional posts
To conclude, Facebook reactions can be used to drive valuable customer insights. Despite their high business insights utility, I haven’t found out too many uses cases online, therefore I decided to publish my own.
We can drive customer experience knowledge from pretty much everything. It’s just good to have your eyes opened and be cautious as free knowledge is available for you everywhere.
By connecting Facebook reactions with actual emotions and quantifying them, you can quickly create easy-to-use metric, which may help you with improving your communication with potential audience. I suggest you to try it.
Let me know if you have any questions or shoot me an e-mail at: email@example.com if you want to discuss the details.
Good luck and enjoy your new KPI!