The Future of Twitter Ads

Since Jan 1, 2011, this number has already increased 82% Our active users are now growing faster in 2011 than they did in 2010 US growth is neck and neck with overall global growth And, we are on pace to add as many active users in the next 4 months as we added in all of 2006, 2007, 2008 and 2009 combined (26M) … Every team in the NFL is on Twitter and more than 50% of NFL players 75% of NBA players 82% of the US Congress and 85% of US Senators 87% of the Billboard Top 100 Musicians of 2010 93% of Food Network chefs 100% of the top 50 Nielsen-rated TV shows


(image) As I posted earlier, last week I had a chance to sit down with Twitter CEO Dick Costolo. We had a pretty focused chat on Twitter’s news of the week, but I also got a number of questions in about Twitter’s next generation of ad products.

As usual, Dick was frank where he could be, and demurred when I pushed too hard. (I’ll be talking to him at length at Web 2 Summit next month.) However, a clear-enough picture emerged such that I might do some “thinking out loud” about where Twitter’s ad platform is going. That, combined with some very well-placed sources who are in a position to know about Twitter’s ad plans, gives me a chance to outline what, to the best of my knowledge, will be the next generation of Twitter’s ad offerings.

I have to say, if the company pulls it off, the company is sitting on a Very Big Play. But if you read my post Twitter and the Ultimate Algorithm, you already knew that.

In that post, I laid out what I thought to be Twitter’s biggest problem/opportunity: surfacing the right content, in the right context, to the right person at the right time. It’s one of the largest computer science and social engineering problems on the web today, a fascinating opportunity to leverage what is becoming a real time database of folks’ implicit and explicitly declared interests.

I also noted that should Twitter crack this code, its ad products would follow. As I wrote: “If Twitter can assign a rank, a bit of context, a “place in the world” for every Tweet as it relates to every other Tweet and to every account on Twitter, well, it can do the same job for every possible advertiser on the planet, as they relate to those Tweets, those accounts, and whatever messaging the advertiser might have to offer. In short, if Twitter can solve its signal to noise problem, it will also solve its revenue scale problem.”

Well, I’ve got some insights on how Twitter plans to make its first moves toward these ends.

First, Dick made it clear last week that Twitter will be widening the rollout of its “Promoted Tweets” product, which pushes Tweets from advertisers up to the top of a logged-in user’s timeline (coverage). Previously, brands could promote tweets only to people who followed those brands. (This of course drove advertisers to use Twitter’s “Promoted Accounts” product, which encouraged users to follow a brand’s Twitter handle. After all, if Promoted Tweets are only seen by your followers, you better have a lot of them).

Just recently, Twitter began to allow brands to push their Promoted Tweets to non-followers. This adds a ton of scale to a product that previously had limited reach. Remember, Twitter announced some pretty big numbers last week: more than 100 million “logged in” users, and nearly 400 million users a month on its website alone. Not to mention around 230 million tweets generated a day. All of these metrics are growing at a very strong clip, Twitter tells me.

All this begs we step back and ask an important question. Now that advertisers can push their Tweets to non-followers, how might they be able to target these ads?

Twitter’s answer, in short, is this: We’ll handle that, at least for now. The first iteration of the product does not allow the advertiser to determine who sees the promoted tweet. Instead, Twitter will find “lookalikes” – people who are similar in interests to folks who follow the brand. Characteristically, Twitter is going slow with this launch – as I understand it, initially just ten percent of its users will see this product.

(The implication of Twitter finding “lookalikes” should not be ignored – it means Twitter is confident in its ability to relate the interest graphs of its users one to another, at scale. This is part of the issue I wrote about in the “Ultimate Algorithm” post, a major and important development that is worth noting).

Now, I’ve spent many years working with marketers, and even if Twitter’s lookalike approach has scale, I know brands won’t be satisfied with a pure “black box” answer from the service. They’ll want some control over how they target, who they target to, and when their ads show up, among other things. Google, for example, gives advertisers an almost overwhelming number of data points as input to their AdWords and AdSense products. Facebook, of course, has extremely rich demographic and interest based targeting.

So how will Twitter execute targeting? Here are my thoughts:

Interest targeting. Twitter will expose a dashboard that allows advertisers to target users based on a set of interests. I’d expect, for example, that a movie studio launching a summer action film might want to target Twitter users have shown interest in celebrities, Hollywood, and, of course, action movies.

How might that interest be known? There are plenty of clear signals: What a user posts, of course. But also what he or she retweets, replies to, clicks on in someone else’s tweet, or who they follow (and who that followed person follows, and, and….).

Geotargeting. Say that movie is premiering in just ten cities across the country. Clearly, that movie studio will want to target its ads just in those regions. Nearly every major advertiser demands this capability – consumer packaged goods companies like P&G, for example, will want to compare their geotargeted ads to “shelf lift” in a particular region.

Twitter has told me it will have geotargeting capabilities shortly.

Audience targeting. I’d expect that at some point, Twitter will expose various audience “buckets” to the marketer for targeting based on unique signals that Twitter alone has views into. These might include “active retweeters,” “influencers,” or “tastemakers” – folks who tend to find things first.

Demographic targeting. This one I’m less certain of – Twitter doesn’t have a clear demographic dataset, the way Facebook does. However, neither does Google, and it figured out a way to include demos in its product line.

Device/location targeting. Do you want your Promoted Tweets only on the web, or only on Windows? Maybe just iPads, or iOS more broadly? Perhaps just mobile, or only Android? And would you like location with that? You get the picture….

Given all this targeting and scale, the next question is: How will advertisers actually buy from Twitter? I think it’s clear that Twitter will adopt a model based on two familiar features: a cost-per-engagement model (the company already uses engagement as a signal to rank an ads efficacy) and a real-time second-price bidded auction. The company already exposes dashboards to its marketing partners on no less than five metrics, allowing them to manage their marketing presence on Twitter in real time. And its recently announced analytics product only adds on to that suite. Twitter has also said a self-serve platform will be open for business shortly, one that will allow smaller businesses to play on the service.

Next up? APIs that allows third parties to run Promoted Tweets, as well as help marketers manage their Twitter presence. Just as with Facebook and Google, expect a robust “SEO/SEM” ecosystem to develop around these APIs.

The cost per engagement model is worth a few more lines. If an ad does not resonate – is not engaged with in some way by users – it will fall off the page, an approach that has clearly worked well for Google. The company is very pleased with its early tests on engagement, which one source tells me is one to two orders of magnitude above traditional banner ads.


Finally, recall that Twitter also announced, and couched as very good news, that a large percentage of its users are “not logged in,” but rather consume Twitter content just as you or I might read a blog post. Fred writes about this in his post The Logged Out User. In that post, he estimates that nearly three in four folks on are “logged out.” That’s a huge audience. Expect ad products for those folks shortly, including – yes – display ads driven by cookies and/or other modeling parameters.

In short, after staring at this beast for many years, I think Twitter is well on its way to cracking the code for revenue. But let’s not forget the key part of this equation: The product itself. Ad product development is nearly always in lockstep with user product development.

Twitter recently surfaced a new tab for some of its users called “Activity”, and I was lucky enough to get it in my stream. It makes my timeline far better than it was. The “Mentions” tab (which we see as our own handle) is also far richer, showing follows, retweets, and favorites as well as replies and mentions. But there’s much, much more to do. My sense of the company now, however, is that it’s going to deliver on the opportunity we’ve all known it has ahead. It’s mostly addressed its infrastructure issues, Costolo told me, and is now focused on delivering product improvements through rapid iteration, testing, and deployment. I look forward to seeing how it all plays out.

14 thoughts on “The Future of Twitter Ads”

  1. I think a much better solution would be for twitter to handle most of those signals internally and invisibly. They can then perfect their algorithm, to maximize clickthrough rates.

    At the very least they should offer an “automatic” option for any of the signals they allow you to target, where Twitter’s algorithm determines the optimum targeting.

    Thus an advertiser should be able to (for example) target a specific region, but let Twitter decide who within that region to show the ad to, based on demographics and interests.

    I have often wished Google offered such a feature for their search advertising. So I could submit an ad and have them determine what keywords to show it to.

  2. I suppose ads in my Twitter stream are inevitable, but I don’t like it. My experience with both Klout and Google’s targeted ads has been, to put it simply, creepy. As you say, “…surfacing the right content, in the right context, to the right person at the right time…” isn’t easy.

    Klout says I’m an influencer of “jeans” since on Twitter, I’m known as @jenajean. Google serves me ads on “cats with bad breath” after I send a Gmail to a friend about seeing a stray cat eat a squirrel outside my back door. Yes, funny now in a sick way, but certainly offensive at the time. And definitely not targeted at all. How does anyone write an algorithm that incorporates common sense?

    Frankly, if ads are inevitable, let me pick the 10 or so categories I prefer. Perhaps change it up and surprise me once in a while, but please – Twitter, Facebook, Google, Klout and all the rest – don’t serve me ads based on what you think I’m interested in. You’ll almost never get it right.

    Anyway, John, great article as usual.

    Jenifer Olson @jenajean

  3. John,

    If you’re reading this, I’d love to hear your thoughts on Jenifer Olson’s remarks, especially the category concept and how it might help Twitter here.


  4. It’s great that Twitter is finally figuring out how to be the same advertising goldmine that Facebook and Google are. The question is how it will affect Twitter users. And I’m not talking about business accounts, but average users who don’t have business interests. Ads do clutter social media accounts.

  5. Hi Lozj, and Jenn. I am not sure the category thing will work yet, because while I see where you are going, not that many folks are engaged with ads in that way – yet. I do see it happening over time…

  6. No matter what products Twitter comes up with, they will never have logo as daring and compelling as the ExecTweet bird. That guy had panache.

  7. It’s not just that Twitter is chasing advertising that is alarming. It’s that they really don’t have the digital real estate to support it. Who uses Twitter on How will they distribute ads? With the Twitter stream being so real-time how can advertisers find value in it?

    This said, all bets are off if they discover an innovative way to deliver advertising value in an “out of the box” manner. Maybe TweetDeck will get redesigned to accommodate advertising beyond “promoted Tweets.” I never understood the value in Promoted Tweets anyway.

    Good topic of conversation Mitch….

  8. I find interesting social media gives control to people while socialmedia advertising follows old paradigm of taking control from people. So I think – in reverse – somewhat similar to your ‘interest targeting’ – but allow people to ‘opt in’ – Facebook has their Like button – Twitter can have another ‘opt in’ device. The more you ‘opt’ the more you get – follows a somewhat Groupon strategic mindset. Allows tracking of opts. Stop the intrusion. Must think SuperBowl of ads — you know people want to watch the SuperBowl ads. Same here. Make them want it. That’s what Twitter is all about.
    In addition – reverse thinking – re local – simply stated – if could envision Twitter as a map of ZipCodes/Census Blocks (that’s U.S) – and profile allowed either ZipCode shopping preferences (from geo perspective) – that might help consolidate / target data – same with interests – if on profile as ‘friendly data’ – that would be # interest and ability to change overtime. Giving the consumer power to find/seek brands/services.
    My gut feeling. Twitter like the Library of Congress – stacks – daunting. Need a Dewey decimal system to navigate – both from brand perspective and consumer — a win-win proposition. This is not all about brand control — social media is also about consumer control — never forget this. Consumer is King (I believe) – think Twitter believes. If Twitter breaks this trust – believe this is huge strategic issue.

    The Question: 1. How important is end-user – the customer – the you and me – re Twitter’s plan to monetize their company? 2. How will Twitter balance the strategic issues of monetizing (doing what brands want) at expense of maintaining brand current brand love?

  9. I have always felt that the hit-and-miss nature of Twitter, where you see lot of not interesting or relevant content before you come across something really good in the timeline, makes twitter very compelling and addictive. this, or human nature where it needs something compare to realize the true value may be the reason for this phenomenan.

    Question for Dick Costolo is:

    Do they see any value in such phenomenan and when they start building algorithms of surfacing the right content, do they see the risk of disturbing such rather chaotic phenomenan which may have adverse effects on the user experience?

  10. My three questions for Dick Costolo:

    1. As Twitter monetizes tweets (firehouse access, etc), will there be a revenue share option for the producers?

    2. How will Twitter work around the challenge the logged-out user presents, in terms of attaining highly prized user data, in order to capitalize on that audience. Additionally, does Twitter have plans to migrate these users from observers to engagers, or will they develop separate services for these two different user groups?

    3. What are Twitter’s plans to maintain the “serendipity experience” while increasing relevance?

Leave a Reply

Your email address will not be published. Required fields are marked *