The perils of polling Twitter bots

Australia’s prime-minister-to-be and self-proclaimed non ‘tech-head’ Tony Abbott now has another title to add to his list; he’s now supposedly the Australian politician with the most fake followers on the Twittersphere.

But before we use this information to leap to any conclusions about Abbott’s real popularity with voters, it might be worthwhile taking a closer look at the validity of the report that pinned the title on him in the first place.

Over the weekend, Fairfax Media published an in-house ‘study’ using two social media monitoring tools, StatusPeople and SocialBakers, to determine which Australian politicians had the most fake followers on Twitter.

The article found around 41 per cent of Abbott’s most recent 50,000 Twitter followers were fake, un-manned Twitter accounts. Julia Gillard came in a close second with around 40 per cent of her recent followers turning out to be bots.

But what it didn’t reveal, however, were the disclosures attached to these services, which potentially render them useless in dissecting accounts with large follower numbers.

As written on StatusPeople:

“For those of you with 50,000 followers or less we believe our tool will provide a very accurate insight into how many inactive and fake followers you have.”

If you're very, very 'popular' the tool will still provide good insight but may better reflect your current follower activity rather than your whole follower base.”

Gillard and Abbott both have over 50,000 followers, which renders StatusPeople somewhat ineffective in accurately measuring their accounts.

SocialBaker’s fine print goes on to further dampen the report’s claims, saying that it’s “expected” that  most Twitter accounts with a high number of followers will have between 20 to 50 per cent of their following as fake accounts.

To give you some perspective, the more followers you have the more bots you accumulate. Even Fairfax is not immune; SocialBaker calculates that around 38 per cent of The Age’s 110,415 followers are fake. Having said that, fake followers are relatively harmless unless you follow them back.

To add further doubt to the report, there’s also the fact that these figures can shift rather drastically over time and each service can provide a different results.

To give you an idea of just how variable these numbers are, here’s a chart of Abbott and Gillard’s fake followers count as monitored by StatusPeople and SocialBakers at 7am this morning. They don't match the figures from the Fairfax report. 

It’s a well-known fact that the media is obsessed with polls – particularly during an election year. As a result, this kind of social media reporting is only going to become more prevalent as we get closer to September. The media – as well as other outlets – are going to use social media analytics to measure and assess the performance of our politicians. It happened during the US election and it’s bound to happen here as well.

The problem, for the most part, is that social media analytics are not yet as credible or empirical as the proper Nielsen or Galaxy polls. Yet, as the Fairfax report highlights social media metrics seemingly hold a similar weight when it comes to measuring consensus.

There’s no doubt that social media analytics will eventually get to a point where they will be more empirical and accurate then our existing polls.

But for the moment, there’s nothing here for politicians or their acolytes (the real ones, not the bots) to fret about.