Personalized experiences? Give me Consent, Control, Trade-off & Trust – A SXSW recap

March 17, 2012 Leave a comment

A vivid conversation

SXSW came and went – and so did the second installment of our discussion about personalization in “How to personalize without being creepy” on Sunday – a vivid conversation with journalists and marketers alike (search Twitter for #sxnotcreepy for the play-by-play):

The room was packed, to break the ice we kicked off with a few examples of personalized content and advertising “in the wild”, using green (not creepy) and red (creepy!) voting cards to gauge the mood in the room: What is acceptable, when does personalization start to feel creepy? We never made it through the Top10 that we had prepared, because after a few “warm-up” examples we put re-targeting and the recent Target story to the vote – and that’s when the real room discussion kicked off:

You can read the complete story by Charles Duhigg on the NY Times website, but the essence of the story is that a teenage girl suddenly received coupons for diapers and other pregnancy-related items in the mail, and her upset father accused Target of trying to encourage her to get pregnant. When Target called a week later to apologize once more, the father had to apologize instead – and declared that his daughter was indeed pregnant but hadn’t told him at the time. Target knew that his daughter was pregnant before he did – how did they?

Voting: Green = not creepy, red = creepy

Target habitually collects all customer data from credit cards, coupons, online orders, etc. It turned out that Target had hired statisticians to identify how shopping habits could be used to predict if a woman was pregnant and what their due date would be. The statisticians identified a number of products that were good indicators (think unscented soaps, lotions, etc.) and their algorithm had determined that said girl was most likely pregnant at the end of her first trimester.

Target learned their lesson from this PR disaster, however. Instead of sending more girls pregnancy-related coupons, they have now taken to obfuscate their knowledge by sprinkling in other, completely unrelated and largely useless, items into the same coupons – thus making the appearance of pregnancy related items “random”. But does that make it better or worse?

In the following discussion a few nuggets crystallized as important when discussing the topic of targeting, privacy, and personalization:

  • The “black box magic” breeds a lot of the concerns and rejection around personalization: If it isn’t obvious to a user how data was obtained, it starts getting creepy. If it turns out you’ve been collecting data secretly all along, even a very transparent opt-in to personalization has a foul taste (e.g. applying personalization for your news feed on all the stuff you’ve read in the past before opting in and didn’t know was tracked).
  • Context and the relationship with the user matter. Customers are much more at ease getting personalized experiences from a brand they trust and to which they have a relationship which warrants deeper knowledge (that’s why eHarmony’s “black box” matching algorithm appears much less creepy than it could).
  • There is distinct value in creating more relevance for the user. Cutting out stuff users don’t want to see or don’t care about is inherently a good thing, but getting this right is difficult: Intent is a fleeting state and once it has passed relevance changes dramatically (Amazon was cited as a great example which continues to suggest related items years after having bought something e.g. as a once-off present; recently they have added the option to “Fix this recommendation” to give users greater control).
  • Your data is the currency that buys a lot of the free experiences – applications, news, services. You can’t expect a discussion on internet privacy to remain inseparable from a discussion on monetization, advertising, and subscriptions. This isn’t a binary link, but there is a relationship that consumers need to understand.
  • We all like free stuff, especially on the internet. Of the entire audience (a group acutely aware of privacy and data usage for advertising) only 2 people paid for their email service – everyone else used a free service monetized with behavioral or contextual advertising. This isn’t inherently bad – but it’s important to remember “If you’re not paying for it, you’re not the customer – you are the product”.

The themes that emerged from the discussion that would aid making an experience not creepy revolved around four main pillars:

  1. Consent – Let the user opt-in (or at least opt-out).
  2. Control – Allow the user to change their mind, control the depth and breadth of data being used; be transparent about what and how it’s being used. This is where re-targeting appears creepy – no opting-in, and few people know how to control it.
  3. Trade-off – People understand that even free stuff has a cost. Should advertising-financed offerings offer users a choice between increased costs and no tracking, or decreased costs and tracking? Let users participate in this trade off decision.
  4. Trust – Consensus appeared to be that for a company a user has a trusting relationship with or sees an authority in a field, personalized experiences are much more palpable.
There were a few big questions that remained at the end of the session that Mat and I will explore a bit deeper over the next few weeks on our blogs – there simply wasn’t enough time to get to the bottom of these:
  • Creepy or not – does it even matter? Do people use less Facebook because they see their friends pictures on ads? Will there be a dip in sales at Target because of above story?
  • How does all this personalization work? Where does the data come from? Do companies just identify my behavior, or do they actually know “me”? How does re-targeting work?
  • How can I navigate this topic as a company? What are the industry initiatives? Are there best practices and guidelines for using data? Most importantly – how is the legal framework changing in the US and in Europe?

Stay tuned on this blog for some posts delving deeper into these questions!

Filed under Ad Targeting, Personalization, Privacy, SXSW

The bad rep of personalization

March 10, 2012 1 Comment

Just ahead of our SXSW about “How to personalize without being creepy”  Pew Internet published the findings of a study on Search Engine Use in 2012 yesterday (here) with some interesting data on users’ preferences for personalized search results and targeted advertising.

Unsurprisingly, the key takeaway of the report is that: “Most search users disapprove of personal information being collected for search results or for targeted advertising“

More specifically, 73% of the polled say that they do NOT want a “search engine keeping track of your searches and using that information to personalize your future search results because you feel it is an invasion of privacy” and 68% stated that “I’m NOT OKAY with targeted advertising because I don’t like having my online behavior tracked and analyzed”.

It gets really interesting when putting these numbers into context with some of the other findings in the report – and I’d love to see the above variables cross-tabled with these other statements:

  • Most internet users don’t know how to control or limit the information that is being collected about them
  • “55% of search engine users say that, in their experience, the quality of search results is getting better over time, while just 4% say it has gotten worse”
  • 86% of search users “learned something new or important that really helped them or increased their knowledge”

It seems there is a theme here – people don’t like the “black box” of personalization and behavioral targeting and have often little idea of what is collected, how it is being used, and most importantly how to control what a website can or cannot do.

Looking at examples of personalization that theme seems to hold up: Personalization is welcomed and appreciated if it’s derived in obvious manners and can be easily controlled – but if it’s “magic” pieced together from unknown sources we get scared. A few examples:

  1. News aggregators or personalized newspapers (e.g. Zite, Washington Post’s Personal Post, etc.) – The user declares the sections he or she is interested in, and the consumption and ongoing rating of content continues to shape the news stream for the user. Hugely transparent, full control.
  2. Dating websites occupy both ends of the spectrum: While Match.com is rather straightforward with allowing users to set their “preferences” explicitly, other companies like eHarmony pride themselves for their proprietary algorithm to match people with results that are largely opaque to the users.
  3. Amazon’s recommendation lists and emails are an odd one – while they are very clearly based on stuff you’ve looked at or bought in the past (augmented with a bit of clearly flagged collaborative filtering) it seems to be impossible to effectively control by the user (even long after the original “intent” for the item has gone)!
  4. Music streaming services and video rentals – often flagged “because you viewed/listened to this, you might enjoy that). While the “magic” behind it (mostly collaborative filtering) is opaque to the user, the connection is easily made for the user. Not creepy at all.
  5. The infamous “following shoes” – re-targeting (while effective) is personalization at it’s worst. Looking at an item on a website without completing the purchase, and then finding that this very item shoes up featured in banner ads on completely different websites later. Black-box magic with no apparent way to control it.
In the future the possibilities for these “black box” magic tricks will be even more substantial. Just look at a few things that have recently been announced by some of the big data players out there:
  1. Facebook – for sponsored stories and social ads: Seeing your friend’s face used in an ad for a company he “liked” on Facebook was just the first step, together with sponsored stories this form of advertising is bound to step up.
  2. Targeted TV advertising through DVRs, based on behavior. Based on your TV viewing behavior the ads you see during the superbowl might not be the same ads your friends see – with little or no way to control it.
  3. Lastly Google+ is now a social layer across all of Google’s products with knowledge of a user’s real identity, demographics, likes, and behaviors – and judging by the recent Safari break-in with little regard for a user’s choice in opting out. This will be a huge playground for targeting and personalizing at scale (Susan Wojcicki seems to confirm this)

Now I am not at all opposed to personalization or targeting at all – I believe it is a useful application of technology to eliminate waste advertising (I really don’t need to watch TV ads for female products, I don’t want to buy golf clubs, and there is no point pitching a vacation in Mallorca to me) and provide me with relevant offerings (Do tell me about the outdoor gear sale, do pitch that sailing trip to me, and yes it’s noon and I’m downtown – so it’s ok to talk to me about restaurant lunch offers).

What is paramount, however, is that (a) you have my consent – so you better be a company I trust and that I’m ok with holding some of my personal information, (b) you’re transparent about the data use – so don’t go off selling it to third parties or do stuff behind my back that I didn’t consent to, and (c) you let me control this relationship – I’m happy to share a lot if I think my data is safe and I trust the company, I might share more if there is something in it for me in return, and I want the ability to turn it off completely at any time.

No need to be creepy!

Filed under Ad tech, Personalization, Privacy, SXSW Tagged with , , , ,

Ad tech: Someone fix this mess, please?

January 13, 2012 Leave a comment

Just to get this out of the way first – I love advertising technology; I’ve worked in marketing and ad tech for a number of years and I love what technology can do today. Ad tech is an industry evolving at blistering pace that has taken us from the blinking banners of geocities through the lows of pop-up banners and the current heights of deeply interactive, highly personalized brand experiences that no longer feel like advertisements but rather content in their own right.

That said – as a whole, advertising technology is a nightmare to deal with at the moment.

Over the last years billions of venture capital money have been poured into this sector creating an industry landscape with countless players, each of them adding some value or preventing some costs along the value chain of the advertising and money flowing from advertisers to publishers to their audiences. How fragmented and cluttered this industry has gotten is probably best demonstrated in the LUMA charts (created by the Luma Partners) which have been wildly circulated in articles and presentations alike. These charts show (conveniently for each major subsector of the advertising ecosystem) the main players and their types of play.

(For a publisher the Search LUMAscape isn’t as relevant).

For anyone trying to find their path from the left side through to the right – it’s a huge, cluttered mess screaming for consolidation.

Having previously worked on the agency and the technology sides of the industry, I have come to find that working on the publisher side might be the most challenging of all for any number of reasons:

(1) A maze with a million valid options

Fundamentally, choices are good. Nobody wants to go back to world in which a single player dominates an entire market. That said – in today’s world publishers have to make a decision what service or product they want to buy or offer and how it will add value to their value chain or product lineup, and then usually run a lengthy investigation to ensure that it (a) actually improves performance and/or bottom line, (b) works with the incumbent advertising solutions, and (c) doesn’t conflict with your product roadmap.

Quite often there is more than one vendor offering a largely similar service, and often enough what’s offered by one is mutually exclusive with something else in your existing lineup (or doesn’t “talk” to one of your existing systems). It’s poorly standardized (i.e. how data is exchanged or described) and requires a lot of creative engineering to set up properly. 

(2) Everything nibbles on *your* bottom line

The revenue waterfall for a publisher’s CPM is quite staggering – between the ad serving, the audience segmentation, the third party data, the DMP, the SSP, and a whole slew of other acronyms a CPM can erode quite quickly. Once you start using these services and technologies not only on the “good” premium inventory but also on the “bulk” inventory, an already slim CPM might result in zero or negative yield once you really factor in all costs. 

I get it – each of the tools adds value or makes audiences or buyers available that otherwise might not have been. But for each of these tools there is the question: Is the marginal revenue or cost prevention of employing this tool for this piece of inventory greater than the cost. Which brings me to #3:

(3) You never *really* know how well you’re doing

Most publishers have a rough idea when they are doing well on raw CPM, and what their floor is where revenue turns into bad revenue (with close to zero or negative yield). Sure, it’s easy to see that at $25 CPM you’re doing well for that inventory, and selling remnant with $0.15 is still more revenue than $0.

But what does your average eCPM look like? And how do you know it’s spot on where it could be? What if you had bought that new data sharing tool? Or decided to kick the creative optimization out the door again? Even with the best modeling the choices in designing and running your ad tech ecosystem will always remain only part science – and part art.

If nothing else, the ongoing investments in this space have ensured that an entirely new consulting branch has sprung to life – monetization audits and ad tech best practices.

Luckily consolidation of the market has started and is ongoing (such as Google buying Teracent, AdMeld, etc. or InMobi buying Sprout), and increasingly we will be able to buy more comprehensive solutions instead of individual puzzle pieces – but new ventures are springing up every months and consolidation is likely to take time.

Until then: Please dear venture capitalists – when you kick off your next technology funds and investment rounds, how about funding some meta-solution providers that help publishers identify the best possible design of an ad tech ecosystem in order to manage your inventory most profitably!

.. or at the very least an audit tool that tells you how well you are doing compared to what’s possible!

Filed under Ad tech, Monetization, Personalization, Technology Tagged with , , , , , ,

SXSW: How To Personalize Without Being Creepy 2

December 18, 2011 Leave a comment

It’s official – the advanced second stage of our quite successful “How to personalize without being creepy” panel at the 2011 South by Southwest Interactive festival (linked here) has been accepted as a core conversation for SXSW 2012.

Core conversations are relatively new formats – introduced in 2008 they seek to provide some form and guidance around the “hallway conversations” that occur between sessions: Discussions around particular concepts, sharing of viewpoints across industries, and arguing the pros and cons of a particular solution.

We’re thrilled to take the discussion about personalization in this new year of Facebook Open Graph to a new level: Beyond the core demographics and simple personal trivia that were the foundation of our SXSW2011 panel, the level of insight all of us are offering other people – and companies – has surged dramatically to heights few of us would have imagined not long ago.

No longer is it just “who I am” and “where I’ve been”, but companies like Spotify let me share in real time every track I’m listening to; every news article I’m reading online is instantly visible to all my friends and open for commenting, and video streams I’m watching quickly make their way onto the suggestion lists of my friends as well.

In just a short year our understanding of what constitutes an in-depth, comprehensive view of “our data” has been dwarfed by a new level of capability and detail that monitors us and our behavior in real time, makes it accessible to friends, and provides companies with insights that allow them to react in real-time and present us with a tailored message of offering.

Now this can be really, really cool and powerful and offer value and service to us customers in never-before-seen ways – but it also has the potential for malice and being really, really creepy (!).

There is a lot to talk about! Over the next months Mat (@matharris) and  I will be preparing the ground for our core conversation on our blogs – so stay tuned here or follow us on twitter (@muuque).

In the meantime I’ll see how on earth I’ll find a hotel room in Austin for March!

Filed under Personalization, SXSW, Technology Tagged with , , ,

Broadcast Is Dead. Long Live The Narrowcast.

July 24, 2010 Leave a comment

(This entry is cross-posted with the blog from my previous video personalization startup Streamvine here)

Making TV commercials is a feat of terrible compromise. You have a great product with some great benefits and features – but airtime is too costly to talk about all of them, let alone that nobody wants to hear the entire laundry list anyways. So you gotta water it down: Who are the target customers for the advertisement, what message will – on average – resonate with them best, and how can you fit it into 30 seconds? Broad message + broad audience = broadcast.

Designing a message for the masses and their averages is bound to shut a lot of people out – those who don’t care about the product. Those who don’t care about the features you picked. But that’s part of advertising. Or was part of it.

Alright, so broadcast isn’t quite dead yet. There is plenty of TV to go around for a long time to come, but things are shifting noticeably. Viewership on Hulu is through the roof. Google’s Web TV is imminent. YouTube is transmitting live sports events, and even Microsoft Silverlight is sponsoring live streaming of Nascar races and some Olympic content.

Increasingly watching TV doesn’t involve the TV at all anymore.

Yet TV advertisements have– apart from gaining some color in 1954 – seen little change. Sure, advertisers have become smarter about placing them, identifying value propositions, increasing reach. But the basic premise is unchanged: one broad message, distributed shotgun to all viewers.

Does anyone really think it’s a great idea to show me an F150 Super cab commercial, when my IP address should make it obvious that I’m watching from downtown San Francisco? Is it really a great idea to have a show sponsored by a well-known feminine product, when I’ve clearly stated that my gender is “male”, when I registered on the site?

Sure, there is a slight chance that my wife is watching from my account. Or that I’m an avid pickup fan who just “happens to be” in S.F. at this moment. In all reality though, odds aren’t on the advertisers side here.

Even if you hit me up with a product I could care about – with all the CRM data companies have collected over the years, shouldn’t they know by know WHICH features I’d ACTUALLY care about?

Enter the possibilities of narrowcasting

Narrowcasting gives us the unique option to specifically cater to countless of atomic known audiences – one at a time. No longer do advertisers have to craft a broad message that appeals to the largest common denominator across what they expect the viewership of their message to be, but instead they can apply everything they know about each viewer, and put it to work to speak to each viewer with a unique voice!

It gets even better if you stop thinking about delivering the right speech for a product, and consider that many companies have more than one brand in the race. Watch some TV in the evening and take note how many ads are really from the same company, just a different division.

With narrowcasting there is a huge opportunity for marketers to think about budgeting across products and divisions in a whole new light. Just take a look how many different commercials a huge enterprise like Johnson&Johnson runs over the course of an hour: Acuvue. Listerine. Neutrogena. Carefree. Rogain.

And how much overlap is in the target audiences for these brands and products? What if you could automatically and instantly decide what is the “best” product to advertise to each individual viewer?

What products across your entire brand and product portfolio do you advertise to 45-year-old Females from New York? Out of those, what has the highest margin for you? What has the highest likelihood of being bought?

How would your product portfolio look, if spread out on a multi-dimensional graph. Do you really want to show the feminine products to all men? Even to all women? Would you tell every viewer about the same features? Or is there things you would tell blondes, brunettes, or red-heads differently if you’d know who’s watching?

Filed under Ad Targeting, Ad tech, Personalization, Rich Media, Technology Tagged with , , ,

Streamvine personalization with Facebook data!

May 21, 2010 Leave a comment

(This entry is cross-posted with the blog from my previous video personalization startup Streamvine here)

With the launch of BizGreet 2.0 we recently enabled out-of-the-box personalization of video channels using viewer data supplied through the Facebook Graph API. This is a great way to personalize anywhere on the web, but with it come all the concerns about the increased availability of private data anywhere.

Facebook and its approach to users’ privacy have been constantly in the news over the last few weeks, so given a few questions we have received lately from customers and users alike, we want to clarify how we use (and don’t use) Facebook data.

First and foremost there is always your consent. Be it before you access our Facebook app or when a website would like to personalize our video streams with your Facebook data – we don’t get to see or use any of your data without that. When you do agree to let BizGreet access your personal information on Facebook, this permission is for BizGreet only and can be revoked by you at any time. It will not enable other web sites or services to use your information (any access is tied to our own domain name, e.g. BizGreet.com), and we will never share this information with others. And really – we have no interest in even keeping your data around for ourselves!

We are passionate about making video content on the web more relevant to viewers (there is so much noise out there, we like to make some content stand out and a little more interesting), and to do that we usually take a quick peek at your demographic data and then decide how to tailor a video stream especially for you (this could mean looking at your first name to make sure that the video addresses you the right way, or looking at your language settings to play the video in your native tongue, or even seeing if you are married or what your hometown is).

However, in the vast majority of BizGreet uses, any information that we are granted access to for a viewer (doesn’t matter if we use it or not) is instantly destroyed after the personalized video has been assembled. Even though we sometimes retain data (e.g. to report success metrics back to our customers, or when data was supplied through an existing customer database), we stay far away from gathering or distributing any personal identifiable information such as email addresses, etc.

If you prefer not to grant BizGreet access to your information through the Facebook Graph API, that’s of course fine, too. We’ll present you with the standard video stream to watch, which is still a compelling message, but just not tailored to who you are.

We are Internet users, too – so we are all for you being in control of your data and who gets to use it!

 

Filed under Ad tech, Facebook, Personalization, Privacy, Streamvine, Technology Tagged with , , ,

Follow

Get every new post delivered to your Inbox.