So, influencers influence. Twitter states the obvious.

PR Measured’s five great tips to optimise measurement
12th May 2016
Influential influencers – hard to find…
19th May 2016

matchboxIn a previous life, I created a successful integrated marketing business, linking PR to above and below the line marketing. One of our offerings was the creation of data, researched specifically for PR purposes. This kind of media coverage-generating research has exploded in recent years… and it was therefore with some interest that I read a recent piece in AdWeek (and now in many other sources), covering a research project, sponsored by Twitter, into the “clout” of influencers. The research was actually undertaken by Annalect, who don’t present themselves as a market research firm, but as a supplier of data mining and analytics tools.

AdWeek’s headline reads “Twitter says users now trust influencers as much as friends” – quite attention grabbing. Seems that Twitter’s focus was to push the claim that “nearly 40% of Twitter users say they’ve made a purchase as a direct result of a Tweet from an influencer” (

The research that Twitter is publicising, claims, as AdWeek puts it, that “social media influencers might have nearly as much clout as a friend or neighbour”.

I’m a little puzzled about how this claim is arrived at – and perhaps you’ll forgive a little digression into the Barcelona Principles, # 7 of which states: Measurement and Evaluation Should be Transparent, Consistent and Valid. (You can skip the next couple of paragraphs if research type stuff isn’t your thing…)

This is how Twitter describes the methodology used for their study:

“In the first phase of analysis, more than 300 respondents were surveyed to understand how consumer receptivity toward brand influencers on Twitter compared with their perceptions when interacting with alternative ad formats and word-of-mouth marketing. The second phase of research explored our findings further through an online lab experiment in which over 500 users were exposed to traditional digital ad formats as well as promoted brand and influencer Tweets in-situ. The goal was to study the direct impact each ad format had on brand health metrics like awareness, favorability, and purchase intent.”

Call me old fashioned, but I’m really not clear how far (or indeed whether) this description accords with the transparency component of the BP #7.

Firstly, there’s a total lack of information about the sampling regime used, including whether/how the respondents were profiled and segmented. Knowing whether they are heavy or occasional social media users would be useful… We also know nothing about their purchasing behaviours and whether they are already predisposed to persuasion by 3rd parties.

But more importantly, there’s a major definitional issue at play here. The claim that influencers influence is supremely tautological.  As influencers, by definition, are those who influence, the finding that influencers are a significant factor in influencing purchases or behaviours is really without much weight or insight.

In fact, the real meat in the research lies in the finding that respondents were far more likely to indicate a purchase intent where a user tweet and a brand tweet were experienced together. The study claims that exposure to a brand tweet by itself increases “purchase intent by three times” but when reinforced by a user/influencer “tweet”, this increases to 5 tines (would be nice to know what the baseline measure was that these figures were compared to)..

It’s Marketing 101, though, that reinforcing advertising with “endorsements” creates greater brand acceptability and purchase intent. It would have been more surprising if no, or less effect had been noted. Quantifying the increase in impact is useful perhaps (if the research method supports it – which isn’t clear).

While we may not have learned anything terribly significant from Twitter’s research, beyond the meaningless finding that influencers influence, it does bring to mind a rather more important (and difficult) issue for social media analytics. How is it possible to identify genuine influencers (using the definition that an influencer – whether friend or not – is someone that influences the mental or behavioural state of another person)? As a marketer, I want to know that my efforts to create a halo effect around influencers are impacting just those influencers having either very targeted or mass influence. In the dynamic social media world, influencers shift rapidly over time, and tracking their impact by means of traditional engagement metrics tell us very little about their real impact over behaviour.

None of the current horde of so-called “influencer” tracking tools, as far as I am aware, capture the crucial “last mile” – i.e. whether a social media influencer has genuinely impacted on a person’s awareness, propensity to purchase, reputation perception or actual behaviour. To do this, of course, requires primary research, which costs, so is generally simply not done, even though to do so would genuinely close the loop, measurement-wise.

The Twitter research reminded me of an excellent, rigorous piece of work undertaken by HP a few years ago (you can read the paper here). There’s a little maths in it, but the abstract is eminently readable. The paper demonstrated that metrics like shares, likes, retweets, updates or “engagement” reflect merely passive button- or link-pushes, rather than expressions of any active intent to act. Most social media analytics are a good example of a more general problem in measurement – that metrics are more than often selected on the basis that they can be measured rather than being based on what should be measured…

Our mission at the Measurement Practice is to help clients create and use string, rigorous data to support their business goals. This research is a salutary reminder that it is all too easy to take headline research on trust – especially where it is designed to support a communications campaign. In this case, to support Twitter’s influencer marketing service, Niche.

So, what did the research show us? That influencers influence, and that reinforcing marketing efforts makes a difference. Didn’t we already know that?

Caveat emptor – not all “research” that glistens is gold!


  1. You’ve done a great job of calling out the flaws in the Twitter research; very interesting. One other flaw – apparently they used a variety of ad formats during the experiment. We know that people respond differently to various types of ad formats, so mixing in another variable may have further confused results. Influencers may influence, but even this is changing as paid influencers are now asking for too much money. Anyway, good thought piece.

  2. […] Twitter “influencers influence”.  But, the Measurement Practice dug a little deeper to find that was not what the data was telling us.  Meanwhile, overspending on influencers was uncovered in “confessions of a social media exec on […]

Leave a Reply

Your email address will not be published.