Skip to main content

Twitter Inc.'s SEC filing in advance of its IPO came in a little bit longer than 140 characters, but it offered investors simplicity as well as an easy-to-understand value proposition – not unlike the microblogging service itself. But despite rapidly growing revenue from digital advertising, some observers voiced concern over its $69.3-million (U.S.) loss in the first half of 2013. If Twitter's not making money now, skeptics ask, why should we believe that it will once it goes public?

Boosters fire back with ammunition from Twitter's S-1 application form. Twitter's overall revenue grew 198 per cent from 2011 to 2012, and in the second quarter, it generated $120.9-million in advertising revenue, accounting for almost 90 per cent of its revenue. The application form also notes greedily that by 2017, total online advertising (excluding mobile) is projected to increase to $124.7-billion, up from $91.1-billion in 2010, and that mobile ad revenue in 2017 could reach $52.2-billion.

However, Twitter's filing also included the remarkable disclosure that cost per ad engagement (a user clicking on an ad) has been decreasing steadily in the last five quarters. And for digital advertising, a wildly-expanding profit horizon may not be in the cards.

In a much-cited October 4 report, Pivotal Research Group analyst Brian Wieser looks at the automated real-time auctions that take place every time an ad is served online. Every time an ad appears on a website like this one, the website (or an advertising firm acting for it) conducts a split-second auction with other servers programmed to place their ads in the best possible locations, and who have specific amounts they're willing to spend per impression (the act of an ad being displayed on an individual computer or device). The Mad Men world of deal-making between advertisers and clients has been replaced by computers making millions of calculations among each other with little or no human interaction.

Mr. Wieser presciently compares the rise of so-called programmatic ad-buying to what happened when Wall Street began replacing humans making decisions with computers: high-frequency trading, which he cites as having driven commissions paid for equity trading down from $14-billion in March 2009 to $9-billion in March 2013. The high-frequency trading machines' hyper-charged trading efficiency took a huge amount of cash away from middlemen, and Mr. Wieser expects online ad exchanges to do the same for middlemen like Twitter. "With programmatic buying… the buyer knows with great speed what inventory is working for them, and can play off sellers who know less and less about the specific characteristics of demand that a given advertiser or trading desk has for a campaign it is executing against."

Small wonder, then, that firms like AOL are trying to get a piece of the action. In August, it bought Adap.TV, which makes software that serves ads for online videos, for $405-million. Twitter's acquitision of digital ad startup MoPub in September gives the social media giant its own top-shelf exchange, but Twitter still has to convince advertisers to place mobile ads on its platform rather than on, say, Facebook's, which is generating strong growth in mobile advertising (in the second quarter, it reported $657-million in mobile advertising, roughly 41 per cent of total ad revenue), and players like Google and Yahoo which are scrambling to catch up in mobile.

With the expected growth in the number of robot-ized ad exchanges, and the ensuing competition for ad dollars, Mr. Wieser sees the revenue per digital ad impression declining, not rising. That means unless players such as Twitter can increase the number of ads they serve, or somehow escape the tyranny of programmatic ad buying through products like their Promoted Accounts that advertisers buy from them directly, their profit outlook depends on growing the number of ad impressions they sell faster than the price of each impression declines. For investors, the question is stark: in a race against man versus robot, do you really want to bet on man?

Report an editorial error

Report a technical issue

Editorial code of conduct

Tickers mentioned in this story

Study and track financial data on any traded entity: click to open the full quote page. Data updated as of 15/11/24 2:07pm EST.

SymbolName% changeLast
GOOG-Q
Alphabet Cl C
-2.33%173.22

Interact with The Globe