Managing brand safety online often means navigating murky waters. At any moment, a relationship with an influencer or a network of bots, may suddenly turn into the iceberg that brings the whole ship down. Wouldn’t it be great if brands had a radar system? With Scraawl technology, social bot detection and social listening, brands can monitor online conversations and keep their brand safe.
Casting a wide net
Monitoring online conversations for brand safety entails not only looking out for interaction with the brand’s social profiles but all mentions of a brand and its products. In order to conduct social listening effectively, a brand needs to follow a couple different search streams, essentially casting a wide net, to cover all mentions of their brand and all mentions of their product/services.
For example, we ran a streaming search on the brand Coca-Cola. The keywords we included were “Coke OR Coca Cola OR Cola OR #
#ShareaCoke“. As new data rolls in, analytics can be rerun and updated so we can view changes in influencers, communities, and even bot networks.
In addition to the streaming report, which we will come back to later, we also ran a Brand Monitoring report on the Coca-Cola Twitter profile. The brand report collected posts from the user and mentioning the user. We found that out of the 10,000 posts collected, 561 are from users identified search as potential bots.
The latest upgrades to Bot Detection now show user stats about the potential bot, even generating a basic statistics dashboard with the top words, urls, or hashtags shared by the user. It’s a useful place to go when looking for a common hashtag used in a coordinated bot attack.
When we completed the brand report with 10k posts, we were interested in understanding how the conversation was affected by the profiles identified as bots. Here is what the analytic Topic Modeling found:
Before bot filtering
Topic Modeling looks at the statistical pattern of word usage across and within social media posts to pull abstract topics. It is therefore more than a world cloud which relies on frequency counts. Topic Modeling is a more useful tool in identifying themes and sub-conversations.
After bot filtering
After the identified bots have been filtered out, we can see how the number of topics have changed. The percentage of the remaining conversations were not affected, however, indicating that the topics driven by the bots were pretty disparate from the rest of the digital conversation.
You’ll notice that topics two to four from above have now been filtered out. We now have proof that the bots found within the data set affected the conversation.
Going back to the social Bot Detection analytic, there is the option to export the list of bots users from the table which we can then plug into another brand monitoring report. For brand safety, it is one thing to identify bots in the conversation and another to actually detect a negative or positive shift in the bot conversations.
Most online conversations are going to have noise and spam, this is simply a given in today’s messy digital experience. However, it is crucial for brands to detect when the shifts in the digital conversation are taking place in real time.
Detecting tide changes
By running a brand report on the bots we’ve collected previously, we can look for behavior change at the source. We could simply look at the updating Basic Stats dashboard from the bot accounts and check for unlikely hashtags or words that may offer further insight.
Another, perhaps more efficient approach, would be to employ content and text features. We looked at Topic modeling previously, which would be useful analytic to employ here. There is also another content base analytic within Scraawl titled Sentiment Analysis. Sentiment is another useful tool for detecting audience behavior change.
You’ll notice in the graph below that there was an increase in positive sentiment around the 6pm mark on September 16th.
We can use Raw Data to filer the information here further to look more specifically at this time period using operators. Conversely, we can look at when sentiment dipped to higher negative ratios. By doing so, we can isolate bot activity based on the positive or negative sentiment found in the texts of their posts.
Perhaps when Coca-Cola monitors their online presence, they will look out for those dips in sentiment in the larger conversation and then run sentiment analysis on just the bots to see if the bots were the cause for the negativity.
We can also take a look at the Coca-Cola keywords streaming report from earlier, as new data comes in we can see how the topics change over time as well.
Essentially, with these reports and analytics we have the radars set up and we have established baseline. So what does a brand do next?
Responding to the radar
If a brand sees an iceberg a comin’, they may have enough time to change course. They may cut ties with the influencer before the scandal becomes too big or that brand may quickly change social media marketing tactics like changing their hashtags if bot activity hijacks their campaign.
Once a brand has identified changes in the digital conversation and have deemed the threat serious enough to change their social media marketing strategy, the best thing for a company to do is to get ahead of the storm. Brands should have social media teams directly reach out to any unhappy campers. Brands should be generating content to head off bad press. If the brand is inundated with bots, then the brand should report the accounts immediately.
All this goes to say managing brand safety online requires work, but the benefits of a thriving social media presence can do so much for a brand. In one example, a simple tweet from a high schooler and the resulting tweet-storm turned into a win for the company in, Nuggs for Carter.
And if your brand continues to put in the work, and maybe works with some good analytic tools, then it should be smooth sailing from there. Bon voyage.