Facebook retargeting, what Twitter should do next, and why this is just the beginning of a magical moment in mobile monetization.
I got retargeted on Facebook today – and it was magical.
Here’s what happened. I wanted to buy a Ted Baker gingham check sports shirt. So first I went to Google and searched for “Ted Baker gingham check sports shirt”. That’s a pretty explicit signal indicating that I’m looking to buy a pretty specific item. The top result was for a matching product page on Nordstrom.com, which I clicked. On that page, I then looked at different colors, scrolled down to read the reviews, but didn’t buy. Said another way, I signaled that I was engaging with the content but for some reason didn’t complete the purchase. From Google to Nordstrom to engaging with content… Close, but no sale. Yet.
Then I went to Facebook to see what my friends were doing.
Typically, the ads on Facebook are targeted based on my demographic info (age, gender, location, etc), and things I’ve liked (aka my “interest graph”).
However, this time, I got an ad on Facebook featuring the specific Ted Baker gingham checked shirt I’d searched for, that I could buy Nordstrom.com. In other words, I got retargeted. From Google to Nordstrom to engaging with content… to Facebook, to clicking on the Ted Baker ad, back to Nordstrom.com, and finally this time to purchase. This time they got me.
As a former ad-tech guy, I love re-targeting for a number of reasons:
- The revenue opportunity is huge. Millions of potential online sales get dropped just before the point of purchase. Before re-targeting, those potential customers simply vanished into the ether. Now they can be reeled back in and converted.
- The technology is cool. In the above example, cookies are dropped on my browser which indicate that I am in the market for specific shirts; the option to show an ad to me when I land on Facebook is made available on an Ad Exchange, along with the information that I’m market for shirts; Demand Side Platforms bid for those ad impressions on behalf of their advertising clients, pricing their bid on that shirt information; and the winning bidder dynamically creates an ad to show me specific products that I’ve previously looked at but haven’t yet bought. It’s the nerd-perfect blend of data flow, systems integration, targeting, and creative – and it all happens in real-time. Totally cool. And, according to super smart folks like Triggit’s Zach Coelius, it works.
- The user experience rocks. A lot of ads you see on the web are poorly targeted, meaning they are not relevant to the user and – surprise, surprise – are completely ignored. Worse, they can be annoying. A re-targeted ad shows me something I’m clearly interested in, and adds to the experience rather than detracts.
However, as a social technologist, this type of retargeting on Facebook is interesting for other reasons. Primarily because it indicates that Facebook can’t effectively monetize with “just” the mountain of the demographic and interest-based data it has about you.
The Interest Graph vs Intent Data
Demographic and interest-based targeting is good for top of funnel brand awareness. For example, an advertiser like Nordstrom could target males, 30-40, who have “liked” various fashion brands – and make them aware that Nordstrom.com offers a good selection of fashionable clothes. It’s like advertising in GQ Magazine – you know the audience is broadly in your target, and you hope that some of them end up buying something at your store. But it’s a bit hit-and-hope, which is why – as an advertiser – you wont pay top dollar for it. Or, from the other angle, Facebook can’t charge a lot for that type of advertising space.
The real money is made on bottom of funnel conversion. If Nordstrom know that I am this close to buying a Ted Baker shirt, and now have a chance to advertise that very same shirt to me, they will pay a lot of money to do so. i.e. Facebook can charge a lot of money for that type of advertising space.
But in order to offer that premium opportunity, Facebook needs to understand – and share with the advertiser – my recent intent to buy. And it’s never going to derive this from my Facebook activity or my interest graph. I go to Facebook to catch up with friends and family, not to signal that I’m market for a new shirt.
This is why Google, in comparison, is such a money making machine. Every time I search for a product on Google, I’m showing my intent to buy – and lots of advertisers will pay lots of money to show their ads intermingled with those search results. In fact, it’s this very need to understand intent – to command higher advertising dollars – that has so many commentators harping on about Facebook needing to build a search engine.
However, as clearly demonstrated in the Nordstrom example above, Facebook doesn’t necessarily need to do that. Thanks to the underlying ad-tech that traces a user from site from site, Facebook can instead exploit the fact that I’ve searched on Google, then landed on a product page on Nordstrom.com before going to the social network. From those clicks, before I got to Facebook, it can understand my recent intent and sell high priced advertising against that information.
So is the interest graph – from a monetization perspective – useless? Is making serious money all about understanding intent?
I’d like to think that re-targeting can be refined with a blend of interest graph data and intent data, and a big dollop of data science. Which is where the technology could get really cool…
Let’s say Nordstrom retarget as above, showing ads on Facebook to users who’ve previously been to their product pages but didn’t make a purchase. And let’s say Nordstrom could also access the interest graph of the users that subsequently converted to purchase as well as those who didn’t. A data scientist would have a field day trying to decipher a pattern between intent and interest data that resulted in the highest number of sales. Who knows, maybe 35 year olds in New York who like Joy Division and Manchester United are more likely to buy Ted Baker shirts than 40 year olds in San Francisco who like REM and like eating at good restaurants. Those two characters might show the same intent (e.g. both searched for the same thing on Google, and both checked out the same product page), but the data might show that one has a higher propensity to ultimately buy. And if you were Nordstrom, armed with that information… you might now spend more money to retarget to the very specific niche – based on intent and interest – that you knew converted at the highest rate.
Twitter, Retargeting, and… mobile magic
I’ve talked mostly about Facebook here, but what about Twitter? Well, of course, they can do exactly the same. At OneRiot, we discovered that Twitter has an implicit interest graph every bit as strong as Facebook’s. But, as Facebook are finding out, to make real money, they need to sell against retargeted intent. Likewise, Twitter needs to enable retargeting.
Now, the ad in a retargeting campaign can take the form of any creative – be that a sponsored post on Facebook, a traditional banner ad, or a Promoted Tweet. So, let’s say in the above example… I went from Google, to Nordstrom and then not to Facebook but to Twitter. Just as on Facebook I saw an ad that had been dynamically created to show me the exact product I had just been looking at, I could now see a Promoted Tweet talking about the same Ted Baker shirts at Nordstrom.com. It could even be hyper-personalized and structured as an @reply. That would be awesome.
Where this could get super exciting for Facebook and Twitter is if they supported retargeting cross-platform (web and mobile), and combined that with geo-location… to enable localized, multichannel, bottom of the funnel advertising. Lots of buzz words there – let’s break it down into plain English:
Let’s say I’ve searched for Ted Baker shirts on Google again, landed on the product page on Nordstrom.com, not bought… and then stepped away from my computer to head down town. While I’m walking the streets, coincidently in the near vicinity to a Nordstrom, I open my Twitter mobile app… and at the top of my stream is that retargeted, personalized, Promoted Tweet from the retailer. But now that Tweet also includes a link to Google maps giving me directions to the store. Using intent data gathered from my online activity, Twitter can deliver a bottom-of-the-funnel Promoted Tweet to my phone that’s informed by my current physical geo-location.
If you really push the boat out and get aggressive, Twitter could even send me a push notification, based on my current geo location, saying “Nordstrom just tweeted you about Ted Baker shirts”. Opening the app, I’d then find a Tweet that linked to not only to store directions but also included a 10% discount voucher redeemable for the next 30 minutes. That would definitely reel me in!
However, for the above examples to work – for that type of cross-platform, geo-located, personalized ad targeting to work – my intent (the information that kicks this whole cycle off, the information that says “this guys wants a shirt!”) needs to be explicitly tied to my personal profile. Today it’s buried in some cookie linked to an abstract understanding of an anonymous user tied to a desktop web browser. That’s just not useable in a cross-platform, multichannel world. Said another way, “@tobiaspeggs” on the desktop needs to be identified as the same “@tobiaspeggs” that opens his mobile device downtown 1 hour later.
I can hear privacy wonks screaming already – and that’s another discussion altogether. But if we can get over that hurdle, cross-platform retargeting becomes a reality. i.e. Bottom of the funnel mobile ads can be targeted based on your earlier, desktop web browsing behavior. And the ad-supported properties with massive numbers of logged-in users who engage with them in pervasive, cross-platform ways (i.e. switching from desktop to mobile and back) become the big winners. Which is why I’m so hot about Twitter and Facebook. They are with us all the time, whether I’m tied to a desktop or walking the streets. And they can show me retargeted ads – focused on conversion, and driving me to purchase online or in a physical store if I’m near one – at any time.
Broadening the target on mobile
Of course, one of the paradoxes of targeted advertising is: the better you target, the smaller the audience you can hit. If you are Nordstrom, and you want to send a promoted Tweet with a discount code to Twitter users who open their app within half a mile of a store, who have recently displayed intent to buy a shirt, who like Joy Division and Manchester United because they tend to convert at a higher rate… then you are looking at a very small audience. Broadening the audience, without losing site of the target, is key.
This is where Facebook have got the technology lead right now – and I’m thinking here specifically about Facebook’s new mobile ad network. Launched in September, it means an advertiser can now run ads “Powered by Facebook” across thousands of other 3rd party mobile apps. Which surely means, one day soon, advertisers will have the ability to retarget, from web to mobile, not just to users who open the Facebook app (relatively small number)…. but potentially any app (relatively huge number!)
Think about this example. Let’s assume Rovio becomes a publisher in the Facebook mobile ad network. Now I could search on Google for a Ted Baker Shirt; poke around on Nordstrom.com but not purchase; go to Facebook and see a retargeted ad… but still not bite; head down town… stand in line at Starbucks, open my phone to play 2 minutes of Angry Birds… and see a Facebook-powered ad telling me to buy that Ted Baker shirt at the Nordstrom two blocks away. Now that’s powerful!
So where’s this all going?
Dave Morin has a great line about the disintegrating distinction between “online” and “offline”. He argues that there’s now just awake and asleep. When I’m awake, I’m connected – at my desktop or via my phone. This is really interesting from a retail perspective. It means it’s time to bring all the techniques we’ve developed to increase conversion and basket size online and bring them into the “awake state” – i.e. to bring them to the phone and to make them work seamlessly cross-platform.
Of course, we’re already seeing a lot of this beginning to happen. Mobile apps that show online reviews of a product I’m looking at in the store; QR codes that link to mobile web pages show more product information; realtime mobile chat that connects you to an expert in the very product you’re standing next to right now. Etc, etc, etc. But now, by tying intent and interest information to a user profile – a user profile that’s consistent cross-platforms – we could start to bring over all manner of ad-based conversion techniques to his “awake state” as well. Such as retargeting; Such as personalized ads (dynamic creative); Such as recommendations (“people who looked at your shirt also bought these jeans”); And even context-sensitive calls-to-action (e.g. directions to a store two blocks away that’s selling those jeans that would look great with your new shirt).
Nerds will rule the world
A lot of nerds love Minority Report for the cool UIs. I always loved Minority Report for the ads. In the movie, Tom Cruise walks past a billboard downtown which scans his retinas to trigger a personalized ad based on past purchase history and inferred intent. Something like: “You bought Kakhis last week, now you need a vest. The nearest GAP store is two blocks away”.
What I’ve outlined in this blog post is a small step away from what was envisioned in Minority Report… and you could pretty much piece it together today (replacing retina scanning and billboards for a Twitter ID and a mobile phone ;). Sure, there are plenty of holes and inaccuracies in what I’ve outlined – but that’s why it’s a blog post not a business plan. What i do know for certain is that there are smarter folks than me who are thinking more diligently about this space, and building out the required technology platforms that will turn into humungous business.
The first time I was retargeted on Facebook, I thought that was magical. But we ain’t seen nothing yet…