Are those Superbowl ads any good? Twitter will tell us.
Even in an age of social media there’s still something about Superbowl advertising that appeals to us all. Maybe it’s because the game is the last big common media experience we share anymore; we know that everyone else will be hearing and watching the same jokes as we are. Perhaps it’s because we relish the fact that for one night every advertiser out there will actually attempt to entertain us, rather than insult us with the idiotic messages that interrupt most programming. Or it might just be all the pre-game hype that obligates us to pay attention. After all we’ll need an opinion come Monday morning.
Whatever the reason, we still look forward with anticipation to ads we hope will live up to all the anticipation.
But since this the age of social media – the ability to join in, share our thoughts, connect with friends and enjoy instant gratification – Mullen thought it would combine the two by creating Brandbowl2010. It’s a simple site that lets you tweet and follow posts from anyone using our hashtag #brandbowl. But thanks to our developers and UX team, and some help from our friends at Radian6, Brandbowl will also include close to real time reaction to the brands advertising on the game by analyzing conversation from everyone on Twitter, not just those using the hashtag.
If Twitter’s API cooperates, we should have a pretty robust experience, counting total number of tweets per brand (we can’t actually analyze each spot separately due to the challenges with isolating a comment, i.e. “loved that Bud spot,” would be hard to assign to a specific commercial), capturing sentiment, and calculating a ranking based on a combination of chatter volume and reaction. In addition, we’ll have word clouds to show you what people are saying. You won’t have to wait until until Monday morning and USA Today to know what brands did best; we’ll have it throughout the game.
It’s been a fun and challenging project, combining integrated teams from Mullen and Radian6, but not without its challenges. For example, how do we know we’re grabbing tweets that refer to the spots, not to a brand in another context? How do we resolve a brand named Dodge, when the word dodge would easily be used to describe a running back’s technique. In test runs we had to eliminate the word Charger, too, as there are lots of people on Twitter talking about San Diego’s football team daily.
So we’ve grabbed as many hashtags as we could find (#brandbowl as well as #adbowl, #superads, #SB44ads, et.al.) We’ve studied lists of spots and scoured the web for any information that would help – celebrities appearing in spots for example – and then created combinations of words to increase the likelihood we don’t grab anything that isn’t a comment about a commercial.
In addition we’ll monitor the game throughout, modifying keywords based on the storylines in the commercials. (If Honda has squirrels in its commercial, we need to add squirrels to a combination of words.)
Why are we doing this? Going to all this trouble ourselves, just so you and our friends and clients and employees can screw around on Twitter during the game? Because we can. Because it’s fun to make stuff, to experiment. It’s valuable to learn by doing. And finally because doing it is better than talking about it.
If it doesn’t work? Crashes? Falls victim to Twitter’s API? So what. Failure, too, can be a great lesson. Let’s hope that’s not the case, however. Wish us luck, and most importantly, join us at Brandbowl2010.com and #brandbowl.
This is fascinating - I think it'll be great, especially for those like me, who only watch the Superbowl for the commercials! I'm really looking forward to checking this out live.
Do you have any plans to analyze what makes a particular spot effective? Or to attempt to apply it to future works or clients? Or are the Superbowl ads too much entertainment to apply in the future?
.-= Erica Nardellou00c2u00b4s last blog ..ericanardello: Thanks for the #FF @SusieGeeee :) Right back at you! =-.
I'm not sure we have the analytics for that. We can measure conversation, sentiment and adjectives that describe sentiment. Perhaps we can speculate from that.
It's kind of sad how excited I am to see the online conversation around the SB ads - can't wait to see Brandbowl in action. @rmoede
Ah, such a fun experiment. And that's indeed what it is, right? An incredible experiment fueled by intrigue, the desire to create awesome interactive experiences, and a common love for all things advertising. :) I'm interested to see the lessons and findings that come from BrandBowl2010.
And to help out with David's question a bit, from a blog post our CTO wrote about our auto-sentiment feature:
"Radian6 automated sentiment reviews on-topic posts as they come in, determines the sentiment of the post at the sentence level, and aggregates a positive, negative, or neutral designation at the post level based on specified sentiment keywords and phrases. If a particular document or post touches multiple topics, sentiment can be determined for each separate topic."
Overall, I think these sorts of real-time experiences help fine tune the social media monitoring industry, provide some great insight into how people react to brands in the moment and how views of "interruptive" advertising have shifted, and continue paving the road for creative teams across the board. A+, Edward. I'm excited to see this thing in action. :)
Content Marketing Manager
.-= Teresa Basichu00c2u00b4s last blog ..Now The Ads Can Play, Too: Radian6 And Mullen Partner To Bring You BrandBowl2010 =-.
Thanks for the clarification. This is both an art and a science constantly improved from experience and previous learnings.
Great idea, Edward. And a nice vehicle for Mullen to engage CMOs of brands in the Big Game - especially those who don't perform so well. Very nice new business tool...
.-= Bob Rinderleu00c2u00b4s last blog ..Budget Balloon & The Coming Bust =-.
Edward - can you expand on measuring sentiment? I was trying to figure this out with the CNN Magic Wall during the State of Union. I assume you capture keywords like "love", "awesome", "amazing", "hilarious", but how do you ensure the context is positive? If a word such as "not" is used in the same tweet, does it re-categorize as negative?
Loving this - just trying to figure out the backend and accuracy.
.-= David Saxeu00c2u00b4s last blog ..Traps to Avoid in the Quest for Viral =-.
Basically you are right: awesome, love it, great, funny, liked it. If there is a not good that will be negative. There is obviously judgment and experience involved. Radian6 does this for a living and has all kinds of data and models to maximize the likelihood of getting it right. Then we modify their model with additional information and predictions based on words we think will get used. Obviously there will be a mistake here and there, but the overall read should be accurate.
I recall at SXSW last year that one of the panelists said Teleflora had the most tweets of any brand that had Super Bowl ads. It's not a perfect measure, but a good proxy as to whether the message is getting through.
The risk, of course, is there is a difference between what people talk about -- did you see the flying music dude in the soda commercial? -- and what people actually buy. The second risk is short-term chatter does not predict real trends in consumer behavior. But this is a great attempt.
.-= Ben Kunzu00c2u00b4s last blog ..The end of typing =-.
Always the scientist. Obviously no doubt nothing like this is perfect. And true brand sentiment needs to be aggregated over time. But we thought we'd have some fun trying the real time capture of comments, reactions, and model a ranking system. I'm lucky to have colleagues who love making stuff and taking on challenges. Hoping it's great and that we'll see you there, cleverly commenting on the best and worst the ad world has to offer.
As a quick aside - we will actually be ON Twitter this year responding (or I will be, anyway) ;). And Ben, you're exactly right about the difference between what they talk about and what the purchase.
The Super Bowl ad was great for us last year, but if everyone of the people commenting had purchased, it would have been even better, I think. :)
I look forward to seeing what happens this year and to watching BrandBowl2010 as I do.
Good play, Edward, by you and the Mullen and Radian6 team!
When you described to me this morning the Brandbowl function as that of sentiment engine, I found the idea of such large-scale, real-time brand-buzz gathering and analysis intriguing. The thought you put into the selection of which hashtags and word combinations to grab is impressive. I have no doubt that Brandbowl will be a great collective experience for us to be part of and learn from.
Let's hope that Twitter doesn't fumble during the game!
.-= Molly Blocku00c2u00b4s last blog ..Welcome to round 2 of @Pepsiu00e2u0080u0099s #PepsiRefresh grant program. Another round of ideasu00c2u00a0can be... =-.