I set out to write a quick correction on a bad article that was discussed on the NY-Tech mailing list earlier this week, but this ends up being half about why Technology journalists and bloggers should just stop – as they rarely know what they’re talking about.
The article “How Much Are Twitter’s Tweets Really Worth?” on BusinessWeek.com has been gaining a bit of buzz across the industry this week. It’s a pretty good summation about how advertising works on Twitter – not because it’s a concise overview, but because it’s about as mindless and poorly conceived an article as the concepts that it speaks about. The writer, Spencer E. Ante, is an associate editor for Business Week. He has an impressive resume and articles behind him, so perhaps this was a postmodern experiment, or maybe he was just hungover from New Years eve. Whatever the explanation is, I’d love to hear it – as its the worst written article I’ve read in ages. The article is no longer online, so I’ll have to use quotes from a cached version in my criticism below. Let’s all take a moment and thank the “Fair Use” clause of US Copyright Law.
# UPDATE
The article’s disappearance was not because of a paywall issue, but because it was – indeed – a steaming pile of shit. Businessweek now states:
> This story contained a factual error that rendered its premise incorrect. The story is no longer available. We regret the error.
I’m keeping this up, not to “rub it in”, but to note that the “factual errors” and “incorrect premise” are something that are pandemic to technology journalism. Writers at BusinessWeek, TechCrunch, Mashable, etc rarely know what they’re talking about – and giving them a podium to stand on is just… dangerous.
# Bad journalism is worthless , Twitter is worth a lot
The first half of Ante’s story is a schizophrenic overview of the recent search deals Twitter signed with Google and Microsoft. Ante starts:
> Google and Microsoft are paying Twitter $25 million to crawl the short posts, or tweets, that users send out on the micro-blogging service. It sounds like big money.
Sounds like big money? That **is** big money – Twitter is making $25 Million dollars to give two search engines a ToS license and access to index their data. In a world where Search Engine Optimization is a skillset or service, Twitter is getting paid by the major engines so they can optimize themselves. This is pretty much unheard of.
For whatever reason though, Ante then goes on to comment:
> But do the math and the payments look less impressive. Last year, Twitter’s 50 million users posted 8 billion tweets, according to research firm Synopsos, which means Google and Microsoft are paying roughly 3¢ for every 1,000 tweets. That’s a pittance in the world of online advertising.
This is where Ante shows that he must be drunk, hungover, or a complete idiot: This deal has absolutely nothing to do with online advertising. Google and Microsoft aren’t paying to advertise on Twitter, they’re paying to be able to show tweets in their own search engines. In fact, given how the integration of this deal works – where Tweets appear in the search engine results with a link back to Twitter – it should be Twitter who is paying the search engines. This is a syndication deal, not an advertising one. And this is to syndicate user-generated-content, not editorial! Twitter now has a giant ad, at the top of most search engine pages as syndicated content , and they got **paid** for it! Getting paid to advertise your brand, instead of paying for it, isn’t a pittance – it’s brilliant, revolutionary, and (dare I say) mavericky.
One of my companies is a media site. We’re not a “top media site” yet, but we’re hoping to grow there. Handling technology and operations, I deal with advertising networks from the publisher side a lot. Another one of my companies is advertising oriented, with a focused on optimizing online media buying and selling. Suffice to say, I know the industry well – which is why I find Ante’s next bit of information troubling:
> Top media sites often get $10 or $20 per thousand page views; even remnant inventory, leftover Web pages that get sold through ad networks, goes for 50¢ to $1 per thousand.
Here’s a quick primer. If you’re a media site with a decent enough brand or demographic, regardless of being at the “top” , you’re getting a fairly decent CPM. I don’t think Ante’s numbers are “right” for “top media sites” – in reality, top media destinations are a bit higher per inventory slot. Additionally, most web pages have multiple slots which together create a “Page CPM” that is the combination of the two. While each slot might get $10-20 , an average of 2 slots on a page would net $20-40. If you look at ad networks that publish their rates (like the premier blog network FederatedMedia.net) , or speak to a friend in the industry, you’ll get instant confirmation on this.
In terms of the remnant inventory, I think these numbers are even more off. Remnant inventory for random, run-of-the-mill websites and social networks will absolutely run in the 10¢ to $1 range. “Top” media sites are of a different caliber, and will monetize their remnant inventory at a higher range, usually in the $2-8 range, or utilize a behavioral tracking system that will net CPMs in that similar $2-8 range.
My main issue with this passage has nothing to do with numbers. What I find even more inappropriate, and wholly irresponsible, is that Twitter is not a “Top Media Site”. Twitter is undoubtedly a “Top Site”, however it is a social network or service. Twitter is not about providing media or content, it is about transactional activity and user-generated content. This is a big different in terms of online advertising. For a variety of reasons ( which mostly tie in to consumer attention span and use cases ) Social Networks have a significantly lower CPM – with most monetizing at a sub $2CPM rate, and a few occasionally breaking into a $2-8 range.
Ante’s comparisons just aren’t relevant in the slightest bit. Across the entirety of his article. But hey, there’s a quote to support this:
> The deals put “almost no value” on Twitter’s data, says Donnovan Andrews, vice-president of strategic development for the digital marketing agency Tribal Fusion.
Really? Really? A $25 Million Dollar deal to syndicate user-generated-content, puts “almost no value” on that data ? Either this quote must have been taken out-of-context, Donnovan Andrews has no idea what he’s talking about, or I just haven’t been given keys to the kool-aid fountain yet. Since Donnovan and I have a lot of friends in common (we’ve never met), and journalists tend to do this sort of thing… I’m going to guess that the quote is out of context.
# Twitter advertising is not (worth a lot)
The second half of Ante’s article is a bit more interesting, and shows the idiocy of Twitter advertisers:
> A few entrepreneurs are showing ways to advertise via Twitter. Sean Rad, chief executive of Beverly Hills-based ad network Ad.ly, has signed up 20,000 Twitter users who get paid for placing ads in their tweets. To determine the size of the payments, the startup has developed algorithms that measure a person’s influence. Reality TV star Kim Kardashian, with almost 3 million followers, gets $10,000 per tweet, while business blogger Guy Kawasaki fetches $900 per tweet to his 200,000 fans.
Using Twitter for influence marketing like “Paid Tweets” is a great idea – however these current incarnations are heavily favoring the advertising network, not the advertiser.
There is absolutely no way, whatsoever, to measure “reach” on Twitter – the technology, the service, and the usage patterns render this completely impossible. The number of Followers/Fans is a figure that merely represents “potential reach”; trying to discern the effective reach of each tweet is just a crapshoot.
When an advertiser purchases a CPM for an ad, they purchase 1000 impressions of the ad in a user’s browser. Software calculates the delivery of each ad to a browser, and those programs are routinely audited by respected accounting firms to ensure stability. Most advertisers, and all premium rate (as above) advertisers have strict requirements as to how many ads can be on a page (standard: max 2-3) and the position (require ads to be “above the fold”). 1000 deliveries roughly equates to 1000 impressions.
When an advertiser purchases a CPM on an email, they purchase 1000 deliveries of the email, featuring their ad, to users’ inboxes. When emails bounce or are undeliverable, they don’t count against this number – only valid addresses do. The 1000 deliveries are , usually, successful email handoffs. A term called the “Open Rate” refers to the percentage of those 1000 emails that are actually opened by the user, and load the pixel tracking software (this method usually works, it is not absolute but good enough). Typical Open Rates vary by industry, but tend to hover around a global 25%; with content-based emails around 35% , and marketing messages at 15%. With these figures in mind, 1000 email deliveries roughly equates to 250 impressions.
When an advertiser purchases a CPM on a Twitter, they merely purchase a branded endorsement (which is very valuable in its own right) that has a potential reach of X-Followers. This number of followers does not equate to the number of people who will see the tweet “above the fold”, nor does it equate to the number of people who will see the tweet on their page at all. Twitter has absolutely no offerings ( at the current time ) to count the number of people exposed to a tweet on their website – either at all, or in accordance with an optimal advertising situation. Twitter has itself stated that 80% of their traffic comes from their API – which makes those capabilities technically impossible for that traffic.
Gauging the number of Tweets sent out over the API won’t work either — Twitter applications built on the API tend to have “filtering” capabilities, designed to help users make sense of potentially hundreds of Tweets that come in every hour. When these client-side lists or filters are used, sponsored tweets may be delivered to the application- but they are never rendered on screen
Looking at common use-patterns of Twitter users, if someone is following a handful of active users, all Tweets that are at least an hour old will fall below the fold… and tweets that are older than two hours will fall onto additional pages. This means that twitter users would effectively need to be “constantly plugged in” to ensure a decent percentage of impressions on the sponsored tweets.
A lot of research has gone into understanding usage patterns in Twitter, as people try to derive what “real” users are: a significant number of Twitter accounts are believed to be “inactive” or “trials” – users who are following or followed-by less than 5-10 users; the projected numbers for “spam” accounts fluctuates daily. Even in the most conservative figures, these numbers are well into the double digits.
Social Marketing company Hubspot did a “State of the Twittersphere, June 2009” report. Some of their key findings make these “pay per tweet” concepts based on the number of followers even more questionable. Most notably, Hubspot determined that a “real” Twitter user tweets about once per day (the actual number is .97). Several different Twitter audits have pegged the average number of accounts followed by ‘seemingly real’ accounts ( based on number of followers, followings, and engagements with the platform, etc ) to be around 50 – so an average user should expect about 50 subscribed Tweets daily as well. The Twitter.com site shows 5 tweets “above the fold” ( which represents 20% of their traffic, and a quick poll twitter clients shows an average of 7 ). Assuming Tweets are spread out evenly during the day, an average user would need to visit Twitter about 9 times a day in order to ensure seeing sponsored Tweets. In the online publishing and social media world, expecting 9 visits per day, every day, by users is… ridiculously optimistic. Realistically, users likely experience a backlog of older, unseen, tweets on login – and sponsored tweets get lost in the mix.
As I stated before, the “celebrity advocacy” concept of a sponsored Tweet is very desirable concept for advertisers — and one that would decidedly command a higher rate than other forms of advertising. However, the concept of “Actual Reach” on Twitter is nebulous at best. A better pricing metric for Twitter-based advertising would be CPC (cost per click ) or CPA ( cost per action ) , where tweeters would be paid based on how many end-users clicked a link or fully completed a conversion process.