Feature

4 social data metrics that might be lying to you

Karin Robinson, head of social insight at OgilvyOne, has kicked off Social Media Week London with her talk, Lies, Damned Lies and Social Statistics: why raw data can tell the wrong story, and why that matters.

Karin Robinson: discusses social data at Social Media Week London. Photograph: Laurie Close (‏@Laurieec1)
Karin Robinson: discusses social data at Social Media Week London. Photograph: Laurie Close (‏@Laurieec1)

Social data is often publicised as the window into the 21st century consumer, but Robinson (@karinjr) thinks that taking some of this data at face value could mean making the wrong assumptions which ultimately, may lead to wrong outcomes for a brand.

See all Social Media Week London coverage

Here are four social data metrics that bear closer scrutiny.

1. Demographic data

Robinson looked at her own followers to test the accuracy of the demographic data from Twitter’s analytics tool. The tool claimed Robinson had a 69% male following. "However, I tweet a lot about feminism and pop culture, so I’m curious as to how and why such a large demographic of my social following on Twitter is male," she said.

By downloading and recategorising the raw data, Robinson found that 33% of her followers were male and 39% were female, with 17% classed as neutral users – brands or companies – and 11% uncategorised.

"Consistently, the demographic of women on Twitter seem to be overwhelmingly underrepresented," she claimed.

2. Age

"In terms of traffic data, Facebook is one of the stronger social media platforms," Robinson said. But she is cynical about whether everyone abides by Facebook’s policy that states users must be at least 13 years old.

"There’s some potential for inaccuracy about the age of social audiences already on this platform," she said.

Robinson revealed only 0.45% of users on Twitter disclose their age, suggesting despite the social media platform’s insights into consumer age, the data may not be enough for brands and agencies to rely on alone.

It's one example where marketers should be sceptical and consider the context. Robinson suggests making sure these insights are validated against customer data elsewhere. If it doesn’t match up to website referral data, you need to think harder about it.

"Do not rely on demographics, look at behaviour," Robinson said. "You want to reach people that act like your brand."

3. Sentiment

Brands want to know if people like them and commonly look at sentiment analysis, but Robinson questioned the value of this perspective.

Robinson explained: "Is it accurate to class all posts as being negative or positive? There’s a much more complex emotional dynamic going on. And is it really about your client’s brand or someone else?

"We’re British, sarcasm is a thing here and this is hard for computers in terms of context, as well as slang. So how can you account for this?"

4. Visibility

It's easy to give prominence to public conversations but "not all conversations are taking place on Twitter," Robinson pointed out.

"We can’t see all the conversations on Facebook and LinkedIn, and there are password-protected blogs and forums that sometimes we can’t take into account."

Dark social, Robinson explained, "is dominant. 71% of all social media referrals come from unidentifiable sources. It is incredibly powerful as this is where consumers talk to people who they trust and are honest with."

Exclusive Vines

Brand Republic asked Robinson what data should you use to show the impact social media has had on a business?

And how should marketers incorporate social media into the marketing mix?

Read Karin Robinson's article about social data on OgilvyDo: Why Raw Social Data Can Tell The Wrong Story


Image courtesy of Laurie Close (‏@Laurieec1), innovation manager at Ogilvy & Mather