Social data is powerful.
We should know, our performance-driven campaigns are based on extracting the right demographic, geographic and psychographic data from unstructured social conversations. It delivers results. Knowing your social data works.
It works because we subscribe to the 10:90 rule by Avinash Kaushik who says that “For every $100 you have available to invest in making smart decisions, invest $10 in tools and $90 in big brains”
Big brains also mean smart brains, and when it comes to social data, you need to know your stuff. You can’t guess with data. You must interrogate hard before you interpret. If you don’t, you will fall down holes that can impact your business quite quickly.
So how can you avoid falling into the trap of wrong social data?
- Watch out for dirty data spikes
Analytics is as much detective work as it is maths. With the right tools (we use third gen monitoring tools) you often see spikes of brand mentions. But if you don’t investigate you might be drawing a conclusion on increased mentions that is just spam. It is often the social ‘bots’ that catch you out. Little algorithms that retweet a piece of news endlessly till they get caught by Twitter. So go look at those peaks and make sure it is not a spurt of spammy content.
- Refine, carve and hone the search
Dirty data in and you end up with garbled results. Setting out your search terms for social monitoring tools is as much an art as it is a science. It really does take time. You start with a well thought through Boolean search string. Then you refine. You review the results. Drill the details, look at the spikes. Then do it again. Constantly filtering until the data you have is clean. No mean feat when you are looking at natural language and unstructured data. So don’t rush it.
- Don’t assume a relationship is a cause
This is an old one. Don’t assume correlation is causation. Certainly a mantra for statisticians, it should be in ingrained into every social analyst too. Want to know more, there is a brilliantly straightforward explanation on the Guardian.
- Be significant with the numbers
It’s really easy, especially when using visualisation tools with social data, to not notice the size of the sample data. But if you are going to base recommendations on your insights then you must have statistically evidential sample sizes. Without it, you cannot trust your data. For instance, suppose we were doing medical research for a new drug. And if we were able to show a 50% improvement after a dose of our new medicine that would be good, yes? Not so good, or evidentiary, if we only tested two people. You need a bigger sample size to ensure robust extrapolation of insight. So watch your numbers. This is a useful sample size calculator if you want to check.
- Avoid the hidden data
Tools for uncovering social data abound. Not just paid ones, but plenty of free ones too. Do watch out though. Some of the free tools that purport to make life easier, don’t. For instance, Klout scores or the LinkedIn’s new social selling index. They hide how scores are created and what values they assign to all the data variables. They make it all look simple. But they are usually not a good measure of influence or selling ability. They can be gamed. And either way you don’t have the details to give you the rigour required to check the analysis – take a look at the superb discussion on @wadds blog.
- Once is never enough
Social data is constantly changing. It never stays still. Yet, so often deep analysis of a brand’s social mentions is carried out only once a year. Smart analytics is ongoing. It notes the changes to inform activities in real-time to meet the changing behaviours, demographics and attitudes. Don’t just create an audit and never review – your audience will move on and evolve and you might be left behind.
As you can tell I love data, but I do have one further piece of advice for anyone analysing social conversations: look up and take in the big picture too. Especially when you want to innovate or capture a new opportunity. Look at other data sources, listen to employees, and understand the wider issues. It will mean you don’t just deliver great data analysis, but invaluable insight, recommendations and results.