Traditional marketing doesn’t rely on anecdotal evidence. I don’t think there’s a profession this side of the Census Bureau that loves data about people more.
If there’s an exception to that rule, however, it lies in social media. Print, broadcast and e-mail marketers have had decades to hone their craft and perfect the way they process data. Social-marketing marketers are still the new kids in the class, and there are still a few kinks in their system.
Many people using social media for business purposes (yours truly included) aren’t professional marketers or publicists. There’s a temptation to rely on soft impressions — personal experiences and secondhand stories that create the rough impression that you know what’s going on. When you finally decide you need some data on how your social content is performing, there are a host of tools, both free and paid, for measuring basic social-media metrics such as click-throughs, re-tweets and the like. Those are fine for measuring your own performance — but how much do they actually tell you about how people see your brand?
The next step, for many people, is to use some kind of professional, automated sentiment-analysis program. These programs will scan social networks for mentions of your brand and create a more robust picture of how people are talking about you. They rely on keywords, syntax and other factors to figure out what each tweet means for your brand. Unfortunately, computers still struggle with the complex, idiomatic way humans express ideas.
In the lead item from today’s SmartBrief on Social Media, Tom Webster asks some very pointed questions about the limits of what machines can tell us about human sentiment. These questions aren’t purely academic, either. If you’re relying on brand-sentiment analysis to make marketing decision and your data are bad, it’s only a matter of time until you reach a disastrous conclusion.
But what is the alternative? Ignoring the problem isn’t an option. For large, well-known companies, measuring social sentiment by hand would be prohibitively expensive. Social data need to be automated to some degree to be scalable. Perhaps all social marketers can do for now is to take in their machine-harvested data — and then consume with a healthy dose of skepticism.
How are you measuring brand sentiment? How do you cope with flaws in automated sentiment-tracking programs?
Image credit, Korovin Vitaly Anatolevich, via iStock
Twitter
Facebook
Linkedin
RSS
There's no doubt that automated apps on social networks and email have a part to play for any entrepreneur or business but that personal touch can't be ignored. I have had surprising results through social networks but if you really want those sales then you have to connect with your audience.
My recent post Ultimate Downloadable Toolbar from Conduit
Hi Jesse, If sentiment analysis is a part of monitoring your brand's reputation, I would argue that human analysis is not just an alternative, it is the only option. As it stands today, there is very little faith in the accuracy of automation and if you cannot trust the data, I fail to see the point. For larger, well-known companies, while they tend to have a larger volume of content, they also tend to have greater resources so it is proportional to an extent. Also, you do not need to analyze every brand mention in order to gauge a brand's perception in social (or traditional) media. Similar to traditional market research studies, proper sampling of your target audience doesn't require surveying all target audience members, yet provides marketers a dataset that has low error rates (+/-5%) and a high level of confidence (95%). Additionally, it is equally important to ask, what is the cost associated with making a "disastrous conclusion" when relying on flawed data?
[...] Your turn: Do you trust automated social data? | SmartBlog On … [...]
Automated sentiment analysis is never 100% correct due to the complexity of the human language. Sarcasm, acronyms, slangs, localised catch phrases and descriptions hinder automated analysis. Thats why at Brandtology, we have social media analysts to go through every single brand mention to ensure high relevancy and accuracy of the data presented to our global clients. For well established brands, guesstimation with a high margin of error is just not good enough. Furthermore, we provide sentiment analysis not just in English, but 9 different global languages.
Ashley Lim
Social Media Consultant
Brandtology
[...] Your turn: Can you trust automated social data? (smartblogs.com) VN:F [1.8.8_1072]please wait…Rating: 0.0/10 (0 votes cast)VN:F [1.8.8_1072]Rating: 0 (from 0 votes)SHARETHIS.addEntry({ title: "Sentiment Analysis: When Humans Are Not More Reliable", url: "http://blog.vmrcommunications.com/2010/04/28/sentiment-analysis-when-humans-are-not-more-reliable/" }); [...]
[...] Your turn: Can you trust automated social data? (smartblogs.com) Possibly related posts: (automatically generated)Trial and errorProblems and Solutions for Complex AnalysisDifferentiable Manifolds 51.542209 -0.317528 [...]
[...] Your turn: Can you trust automated social data? (smartblogs.com) Possibly related posts: (automatically generated)Trial and errorProblems and Solutions for Complex AnalysisDifferentiable Manifolds 51.542209 -0.317528 [...]