Inflationary Twitter Audience Numbers Hurt Social Media Credibility

In yesterday’s New York Times, you may have read the article, Spinning the Web: P.R. in Silicon Valley, an interesting although not overly insightful piece.  From a social media measurement perspective, two items caught my eye.  The first, referring to Brian Solis, Principal of FutureWorks, about how he calculates social media audience figures:

“Instead of calculating the impressions an article gets by estimating a publication’s circulation and pass-along rate, Mr. Solis counts the number of people who tweeted about a company and their combined following, the number of retweets or clicks on links, as well as traffic from Facebook and other social networks.”

Toward the end of the article, we learn:

“By 6:30 p.m. on the day Wordnik went live, Brew’s staff calculated that 1.43 million people had seen tweets about it.”

Setting aside for a moment that the article and these sorts of audience metrics take a broadcast-oriented view of Twitter (Mr. Solis discusses the shortcomings of the NYT viewpoint here), the emerging view of audience measures for Twitter is to calculate the Followers of each person tweeting about the subject of interest, and then adding Follower numbers for each person retweeting the subject and so on.  The issue here, much as it is in traditional public relations, is that the audience figure that results from these sorts of calculations grossly overstates, by one or two orders of magnitude or more, the actual “audience” for these tweets.  It is a hypothetical number that assumes everyone that possibly could see a tweet has in fact seen it, and everyone who sees it is relevant to you/your brand.  This is fantasy of course.

On the issue of relevant audience, here is a quick example.  At the time I pulled these figures, the audited circulation of the New York Times was 4,974,000.  Most PR practitioners getting a ‘hit’ in the NYT would claim this as their audience.  However:

• If you were only trying to reach a C-Suite audience with your message, the actual audience reached would be 598,000 or 12% of the total circulation
• If you were trying to reach Women, your audience would be 1,937,000 or 39% of the total
• If you were trying to reach 25 – 54 year old Men, your potential audience would have been 2,930,000, or 59% of the total number.

There is a large difference between how many people theoretically can see a tweet, versus how many actually saw it/read it, versus how many of those seeing the tweet find it relevant to them, versus how many engaged with it by hitting a link or retweeting.  Part of my issue with this is the language we use to report the figures.  For the Brew staff to use these numbers to estimate 1.43 million people “had seen tweets about it” is wrong.  If they had said 1.43 million people had an opportunity to see the tweet, it would have been more realistic, although still greatly overstating actual relevant audience.

This problem of audience inflation has already been institutionalized in public relations.  The use of Impressions as an output metric does not mean a true impression in the branding sense, but rather an opportunity to see the content.  To make matters worse, many PR practitioners believe Impressions should be factored by either dubious pass-along readership figures and/or use of a multiplier to account for the mythical credibility advantage PR enjoys over impressions generated from advertising.  The simple fact is there is no research-supported, fact-based argument for using any adder or multiplier in public relations when calculating potential audience (here’s an IPR white paper on this subject I co-authored).

For Twitter and other social networks we lack demographics and data about tweet readership averages (i.e. what is the probability that any one tweet is actually read) that would allow for more precise audience estimates.  In the absence of data, believable assumptions should be used:

• Out of all the opportunities to see, how many actually read the tweet?  10%?
• Of those reading the tweet, how many find it relevant to them (or from the other perspective, how many of the readers are in your intended target audience)?  Maybe 10% again?

You can see how our audience estimate has already been reduced by a factor of 100.  This may well still be overstating the actual, relevant audience.  The issue here is that unrealistic and overstated audience figures have the potential to hurt credibility and call into question other data and metrics that may be more grounded in fact.  Actually the more meaningful metrics pertain to engagement or outcomes rather than exposure/outputs.  It is more meaningful that 40,000 visited the Wordnik website as a result of the campaign discussed in the NYT article than the overstated 1.43 million who were estimated to have seen the tweets.  40,000 is real.  1.43 million is fantasy.