People talk, they say good things about a brand, they say bad things about a brand or they just talk. Social media has become a place for people to make sure whatever they are saying is heard loud and clear by their friend and followers on Facebook, twitter, blogs etc.
Going through all this social media chatter to understand and figure out what is going on about your brand is a very labor and time intensive work. Even though many Social Media Monitoring tools have done a great job in aggregating the data from various sources, but to understand the context of those conversation and provide that to marketers in a meaningful way still needs a lot of work. Currently it really needs a human to understand the full context of the conversations and understand the sentiments of consumers.
However there are some metrics that can provide you a measure of the sentiments (positive or negative) of these conversations and indicate which direction they are going.
Social Media Sentiment Metrics
Almost all of the social media monitoring (listening) tools provide following metrics to help you understand the sentiment of the conversations about your brand in social media.
- Total number of conversation or mentions about a brand, product or topic
- Number of positive conversations or mentions about a brand, product or topic e.g. a tweet saying: I love brand ABC
- Number of negative conversations or mentions about a brand, product or topic e.g. a tweet saying I hate brand ABC
Many social media analysts use these raw number of positive and negative conversation to monitor the health of their brand, competitor or industry in social media (positive is good and negative is bad, unless that conversation is about your competitor).
However,these numbers are not KPIs. For example, if one day there are 100 negative conversation about your brand and the next day you have 50 negative conversations then what does it really mean? On the surface it looks like you have done something good to bring down the negative conversations by 50%. Is it really true though?
To fully understand those numbers we need a little more context. Looking at the total number of conversations about your brand might provide some context. Say, there were total 200 conversations about our brand on day1 and 100 on day two. If we take percentage or ratio of positive conversation to total conversations and ratio of negative conversation to total conversation we find that we did exactly the same on both the day.
On Day 1: Negative Conversations/Total Conversation Ratio = 100/200 = 50%
On Day 2: Negative Conversations/Total Conversation is 50/100 = 50%
As you can see these ratios or percentages provide much more information than the raw number did. Even though raw negative mentions were down, as a percentage your negative conversation were about the same on both the days. This is where many social media analysts and marketers stop and use the above 5 metrics as KPIs. Let’s relist the 5 KPIs discussed so far
- Total Conversation about a brand, product or topic
- Number of Positive Conversations about a brand, product or topic
- Number of Negative Conversations about a brand, product or topic
- Ratio or Percentage of Negative Conversations/Total Conversation
- Ratio or Percentage of Positive Conversations/Total Conversations
But something is still missing in these metrics. Those who have analyzed social media conversation know that majority of the conversations are classified as “Neutral”. Neutral means that there is no positive or negative sentiment in the sentence or the conversation in which that brand, product or topic is mentioned. In my experiences, over 90% of the conversations are neutral. So let’s take another example to show how that messes up the above KPIs.
- Day 1
Total Conversations: 1000
Negative Conversation: 5
Positive Conversations: 10
Negative/Total Conversation = 5/1000 = 0.5%
Positive/Total Conversation = 10/1000 = 1%
- Day 2
Total Conversations: 1500
Negative Conversation: 5
Positive Conversations: 10
Negative/Total Conversation = 5/1500 = 0.33%
Positive/Total Conversation = 10/1500 = 0.67%
In the example above, it looks like our positive and negative conversation both dropped on day 2. Though in reality, looking at the raw numbers there was no difference in the volume of positive or negative conversations. It just happened that “neutral” conversations went up on day 2 causing the percent of positive and negative conversations to go down.
So as you can see, in this case raw numbers are a better indicator than the percentages or ratios. So you can see how none of the above 5 KPIs provide an accurate view of sentiments of conversations in the social media. We need a better KPI.
I use another KPI, that I call “Sentiment Indicator” or “Sentiment Index”, which in my opinion, is a better indicator of sentiment then other metrics that we discussed. Here is how I calculate “Sentiment Indicator”:
Sentiment Indicator = (Positive Conversations – Negative Conversations)/(Positive Conversations + Negative Conversations)
(Note: Even though neutral comments are still good for analysis, I do not use them in my calculations of Sentiment Indicator.)
Using our example above Day 1, The Sentiment Index will be 10-5/(10+5) = 33%, which is exactly the same as that on the Day 2.
This metrics is more actionable than other metrics. If it goes in negative direction that means we are getting higher number of negatives as compared to positives and it is time to get into action. If it goes in the positive direction then we must be doing something good and time to find out what that is. Also, many times I will use the volume of conversation along with it to make sure that while we are maximizing the positive conversation we are also enabling the total conversation volume to go up.
Note: You should still dig deeper into those conversation, particularly the negative ones and see what is going on. Sometimes even one negative comment can quickly go viral and ruin your reputation.
Your turn now. How do you measure sentiment?
If you are not sure and need help, don’t hesitate to email me at batraonline (at) gmail (dot) com or leave a comment on this blog post.
Web Analytics Jobs:
- Senior database developer (Sr. SDE) Optizent Redmond, Washington
- Associate Games Analyst Arkadium New York, New York
- Director of Analytics Arkadium New York, New York
- Web Analytics Engineer – Contract (POP) POP Seattle, Washington
- Web Analytics Manager Gilt Groupe New York, New York
- Website Analyst Simplexity Reston, Virginia
- Principal Web Analyst AOL Virginia
- Senior Web Analyst VML New York, New York
- Director, Analytics VIMEO New York, New York
- Web Marketing Analyst (International) j2 Global Communications Los Angeles, California
- Senior Analyst, Web Reporting and Web Analytics TXU Energy Irving, Texas
- Director, Digital Analytics Deluxe Corporation El Segundo, California
- Digital Marketing Analyst Makeover Solutions Inc. New York, New York
- Senior Web Analyst VML Kansas City, Missouri