This is a summary of the “Data=dollars” white paper we wrote for OpenX, published a few weeks ago.

There are a number of exciting opportunities for publishers interested in using their knowledge of their audience to drive improved ad revenue. We’ll be exploring some of these possibilities, and walking through a series of best practices, in a forthcoming blog series. Stay tuned!

As advertisers become more sophisticated about using targeting to drive media buying decisions, the value of audience data has grown. This creates an opportunity for publishers to develop their own consumer data and use it to increase ad revenue.

To date, however, few publishers have done so. Audience data has primarily been developed by intermediaries including ad networks and data-specialists (like Audience Science or Blue Kai.)

In the white paper, we explain why publishers are in a unique position to develop their audience data, and why this represents a revenue opportunity. The paper goes on to describe the steps publishers should take to develop targeted audience segments based on user data, and use that data to drive up CPMs and sell-­through rates.

The white paper can be downloaded from the here, on the OpenX website. If you are a publisher interested in developing your audience data and require assistance, don’t hesitate to contact us.

Happy reading!

There was no chance of a consensus emerging amongst participants, illustrating how divided people are on this complicated issue.  It did prompt me to draw a number of conclusions:

• The key question is not where along the private – public continuum most people are likely to situate themselves.  It is what value they place on different bits of their privacy, and how much value they need in exchange for giving that privacy up
• At the moment, Facebook faces a “perfect privacy storm”:
  1. Users who previously expected their data to be kept behind a walled garden are uncomfortable finding out that that might no longer be the case. (Versus Twitter, where users from the joined expecting everything they Tweeted to be public.)
  2. Users do not understand what data Facebook collects about them and how this is used to generate value. Their privacy concerns are therefore derived from both a lack of transparency and control.
  3. Facebook is not in a position to educate users on how data will be used, because frankly, they’re still working out how to drive value from that data.
  4. Users don’t see any value for themselves in sharing their data. There is therefore very little to persuade them to give up any privacy. (This would change if Facebook found a way to drive value from that data that benefited themselves, their advertisers and their users at the same time.)
• There are some parallels with the privacy debate around behavioural targeting. Once again, there is a perfect storm where
  1. Users don’t understand what is being done with their data,
  2. have no visibility on what happens to their data,
  3. have no control over what happens to their data and
  4. don’t see any benefit in sharing that data so have little incentize to give up any of their privacy.
• However there is one key difference between the Facebook privacy debate and the debate around privacy and behavioural targeting. Namely, that the information Facebook has, although deeply personal, is broadly accepted as data which the users have voluntarily submitted to Facebook. This makes Facebook somewhat different to the behavioural targeting companies, who in many cases have been collecting data on users that they may not even be aware exists, and have not in any direct sense created.
• At the end of the day, Facebook is a multi-sided platform that has to work out a value proposition that works for it, its users and its advertisers.  The privacy storm suggests that it is out of kilter with its users – and I wonder whether advertisers view this with excitement or trepidation…

The Rubicon Project, developers of an automated platform that connects publishers to hundreds of different ad networks, and serves ads from the highest paying ad networks based on their own algorithms, published a manifesto on Friday for revolutionising the digital ad ecosystem and giving publishers more power. Quite rightly, they are worried that the way advertising inventory is traded short-changes publishers, limiting the amount of money available to fund high quality content. This obviously disadvantages publishers, but also disadvantages consumers of content and by extension society writ large. By restoring “power to the publisher”, the Rubicon Project aims at nothing short of saving society itself.

Whilst the manifesto makes for thought-provoking reading, it fails to address the key issue holding the online ad ecosystem back, and hence its prescription is unlikely to deliver the kind of result hoped for.

The Rubicon Project’s diagnosis of what is wrong with the online ad ecosystem

The digital ad ecosystem has been built around the needs of advertisers and media buyers rather than around the needs of publishers.  As a result, the structure of the industry, and indeed its fundamental technological components (including the ad server itself) all serve to lower the CPMs received by publishers.  It’s worth giving some specific examples to illustrate the manifesto’s point:

• By selling inventory through multiple channels (direct and via different ad networks), publishers enable advertisers to “shop around” for the same inventory at the lowest prices, eroding CPMs
• Because advertiser demand is fragmented across multiple channels, the cost associated with finding the “best paying” advertiser is enormous, meaning publishers generally have to settle for less well paying advertisers
• The way inventory is traded by third parties that sit between publishers and advertisers (e.g. ad networks, media buying agencies) serves both to commoditise inventory and increase the number of agents taking a %, both eroding CPMs.  Worse, by “owning” relationships with buyers and the data on which that industry is valued, and not sharing them with publishers, those 3rd party intermediaries can take a higher % than perhaps they deserve

The Rubicon Project believes that now is an opportune moment to change the ecosystem, because the development of tracking technologies and behavioural targetting specifically mean that suddenly advertisers are more interested in “who” their ads are shown to rather than “where” their ads are shown.  The manifesto proposes to:

• Increase transparency between advertisers and publishers, making it easier for all players to identify where value is being generated and allocate revenue accordingly
• Better connect publishers to multiple demand sources.  (Isn’t this what Rubicon has been doing to date?)
• Give publishers their own data platform, giving them a better understanding of what drives value in their inventory, and better equipping them to capture more of that value
• Provide publishers with better yield management, channel management and forecasting

What the manifesto misses

There are three approach to drive up display ad CPMs.  The manifesto touches on the first two but misses the third, key approach:

1. Increase the efficiency of the display ad market by increasing transparency and reducing fragmentation. It is a fact of economics that the most efficient markets, for both buyers and sellers (if not arbiters) are transparent markets.  Increasing transparency and reducing fragmentation should drive up the CPMs publishers receive by making it easier for them to identify the highest paying advertisers and transact with them.  For advertisers too this is great, because they should be able to identify the most cost effective places to buy suitable ad inventory
2. Enable new sources of demand.  Another fact of economics is that price is a function of the demand / supply balance.  By opening up their platform using a self-service interface to advertisers who were previously too small (or spent too little) to participate in the display ad ecosystem, the Rubicon Project should be able to drive up total demand.  Indeed, the presence of a self-service interface for advertisers is one (of several) drivers of high search CPMs.
3. Make display advertising a more effective branding media.  The more effective display ads are as a branding media, the more advertisers will be willing to spend on display ads.  In spite of a lot of industry commentary to the contrary, many advertisers are skeptical about the effectiveness of display ads, and as we have explained before they are right to be.

By missing the third approach, The Rubicon Project appears blind to the biggest challenge facing the display ad space today.

Why most display advertising is not an effective branding media

The traditional model of monetising content through advertising, where people have to consume advertising they are not interested in in order that they can access content that they are, doesn’t work online where consumers can happily switch between many “channels” (i.e. websites) and are able simply to ignore (or even physically block) adverts that appear alongside content.

Online advertising works well when the advert adds value for the consumer as well as the advertiser, hence benefiting the publisher not just by providing a revenue stream, but by improving the consumer experience.  Search advertising is a great example: sponsored results provided on search results directly benefit the consumer by making available companies that offer products and services the consumer is actively seeking.   This can work for brand advertising as well as performance advertising: on websites that attract visitors that want to be entertained for example, display ads need to “add” to that entertainment. Today’s FHM website, for example, has a full screen takeover advertising the DVD of the film “Jennifer’s body”, including clips of the film and visuals of Megan Fox, something most FHM readers will be eager to consume.

For advertisers, then, it is not enough to know “who” an individual is who is being shown an ad and “where” the ad is being shown. The advertiser also needs to know in what mindframe is the individual: are they in “entertainment mode”, wasting time or looking for distraction, as I imagine most FHM readers are. Or are they in “accomplishment mode”, researching specific things with a view to making a purchase, e.g. to buy a present for a specific someone. Depending on the mindset of the individual, AND the context in which they are, the advertiser then needs to deliver content that ads to the user experience AND makes it more likely they will respond positively to their brand and / or make a purchase.

This is a tricky but solvable problem for an advertiser looking at a specific website:  after all, the context of a website and the way a visitor has come to it would provide an advertiser with a good idea of the kind of mindset that the consumer has.  (No one goes to the FHM website to be productive.)  For an advertiser buying across a 3rd party platform (like Rubicon’s for example), it is incredibly difficult to ascertain the users’ mindset.  And even if they can, it is tricky to work out what format of ad would be most effective.  (Note how the “Jennifer’s body” ad on the FHM homepage is a whole screen takeover.)  Simple display ads circumscribed in standard banner or skyscraper dimensions offer little scope for advertisers to actively engage website visitors. Standardisation (sadly) leads to commoditisation.

Publishers can drive up ad revenue themselves

As we have seen, third party platforms like Rubicon work by aggregating both supply and demand and making it easy for advertisers to buy across multiple publishers and for publishers to sell across multiple advertisers.  This standardisation has disadvantages for publishers, because the context of the site and the mindset of the user get lost.  This is one of the main reasons why direct ad sales houses typically win publishers much higher CPMs than automated 3rd party platforms:  it is easier for a person to explain directly to a media buyer the possibilities presented on a particular website than it is to express them in a machine-readable format.

Publishers that want to increase ad revenue need to think creatively about the way they can work with advertisers to deliver better experiences for their users. For a community site, this might be adding transactional features (as discussed in a previous post); for editorial-led entertainment sites it might be about making it easy to sell and deliver homepage takeovers. This is a commercial problem with technical dimensions, not a technology problem with commercial dimensions, as the Rubicon Project presents it.

If you’re a publisher exploring ways to grow your revenue, get in touch to discuss what opportunities are open to you and the best ways to go about testing and realising those opportunities.

In today’s FT, Nikesh Arora, president of Google’s sales operations and business development, and David Eun, Google’s vice president of strategic partnerships, are both quoted recasting the nature of online advertising.  They make two points:

1. We should no longer distinguish online advertising from radio, print and TV advertising – as these “old” forms of content are increasingly distributed over the internet
2. The nature and format of ads online today need to “catch up” to exploit developments in the online ecosystem.  Copying TV ad breaks, for example, will not be good enough

Mr Arora and Mr Eun are right – there does need to be a fundamental shift in the way people think and treat advertising online, both to exploit the new opportunities that the internet presents, and also to ensure that advertising can continue to support content creation.  The key change necessary, however, is that people in the industry need to put consumers at the heart of online advertising.

1. The consumer needs to be at the heart of advertising.  In traditional TV, print and radio, the consumer passively consumes advertising as part of consuming content they’re interested in.  Because of a limitation in the number of channels available to consumers, advertisers were in a position to pretty much force consumers to watch whatever ads the advertiser wanted to foist on them.  As a result, what advertising consumers were interested in was not very important.  That is not true on the internet, where consumers can choose from a plethora of content unimaginable before.  There has to be a rebalancing, with a greater amount of spend on advertising that consumers actually seek out and consume of their own volition.  (Leveraging viral marketing and social media.)  Even where advertising is forced on visitors, more effort needs to be put into make sure that advertising is both appealing and engaging to the consumers who are exposed to it
2. The consumer needs to be given a more transparent choice between consuming advertising and paying for content.  Consumers understand that ads in commercial breaks on TV make TV less expensive for them.  Online, that choice isn’t so clear, because it often isn’t presented to starkly.  Where publishers can, they should offer their visitors the choice whether or not they’d like to pay for content, or look at ads as part of their content consumption.  Consumers who know they are saving a certain amount of money might be less likely to mind about more intrusive online advertising formats.  At the same time, they might more actively opt to choose to consume more ads on sites where the ads more interesting and appealing, benefiting advertisers who then have a higher level of confidence their ads are capturing attention
3. The consumer needs to be put at the heart of new technologies e.g. behavioural advertising and ad exchanges.  We’ve said it before and we’ll say it again here.  Behavioural targeting is a transformative technology because it allows advertisers to decide who they’d like to advertise to rather than where they’d like to advertise, which is fundamentally more important to them.  But to work, consumers on the internet need to be able to have say over “who” they are, where they are identifiable, the interests that are ascribed to them, and understand how advertisers use that to target them.  Relying on cookies, and trying to work out what people are interested in based on their web browsing history (vs letting them express it themselves) is inefficient, ineffective, and has already sparked a consumer backlash.
4. The metrics around online advertising need to change to capture consumer attention.  Currently, online advertising is measured in impressions.  Consider that any publisher can double, triple or quadruple the number of ad units per page and suddenly they have got twice, 3x or 4x as much inventory to sell.  What advertisers are interested in are not impressions, which are fairly arbitrary units at best, but units of consumer attention:  there are 60M people in the UK, they have a finite number of hours a day, and spend a finite percentage of those hours consuming media.  There is therefore a finite amount of consumer attention that advertisers can hope to attract, and a finite amount that each publisher can hope to offer to an advertiser.  

No one argues against the idea that there’s something wrong with the display advertising market. Not only do CPM rates for display ad inventory pale next to search CPMs, but the total advertisers’ spend on search engine marketing is higher than the total on display, in spite of the fact that internet users spend a tiny fraction of the time they spend online on search sites.

That was before the credit crunch.  In the ensuing months, display CPMs have plummetted further, whilst their search cousins have enjoyed continuing growth.  If anyone needed pursuading before that something was wrong with display, they don’t now.

Where people disagree today is why display advertising sucks and hence what the remedy is.

A common misconception: display advertising sucks because the market is fragmented and opaque

The argument runs as follows:  display advertising sucks because the  market is fragmented and opaque.  Advertisers cannot be bothered to deal with large numbers of small and medium websites, so concentrate their spend on a handful of top sites, or buy across ad networks.  They overpay for large sites which attract similar advertisers, and underpay for smaller sites / network buys, because they have no real idea what they’re buying on.  (And so are unwilling to spend a lot on it.)   This sucks for advertisers, because they either overpay, or lack visibility, and all but the largest publishers whose inventory sells at below its “true” value.

The solution?  More effective ways of matching buyers and sellers, either in the form of bigger ad networks (e.g. Platform A, AdSense), exchanges (e.g. OpenX, RightMedia) or algorithms to optimize which ad network an ad is sold to (e.g. Pubmatic, Rubicon).  Companies with this goal in mind have raised vast sums of money from VCs keen to take a percentage of the uplift if these solutions can drive a rise in CPM rates so that display looks more like search.

Sadly, there are more fundamental problems with the display ad market

To understand the real reason why display advertising sucks today, we need to go back advertising basics.

Broadly speaking, there are two types of advertising that happen on the internet.  Brand advertising, and performance advertising.  Brand advertising is the type of advertising you see on TV, in print and on the radio: the purpose is raise awareness and emotional affinity for a specific brand in consumers.  Performance advertising is advertising with the aim of driving a specific action in a short time frame, normally buy a product or service.  Before the advent of the web, performance advertising was a niche activity and represented a tiny proportion of overall ad spend: after all, advertiseers had limited opportunities to trigger a direct response e.g. using direct mail.

Before we consider how effective the web is as a brand or performance advertising medium, we need to think about what consumers are doing when they’re on the web.  Whilst the number of potential things is long (and growing), we can broadly divide activities into two categories:

1. Getting useful stuff done (e.g. sending emails, paying bills, booking holidays, buying groceries)
2. Killing time (e.g. catching up with friends, reading newspapers, watching videos)

That the web is a place where people are productive makes it a unique medium to advertise in.  (At least different to TV and most print publications.)  For advertisers, that creates a huge opportunity:  what better time to advertise a camera to someone then when they are in the process of buying a camera, or researching different cameras to buy?  Or a holiday to someone that’s exploring different holiday options on the internet?  That’s why performance advertising works so well on the web – it’s why search CPMs are constantly high, and it’s why well placed affiliate links, located on sites where people are engaged in activities that relate to the ad, make for effective advertising.

Why display advertising sucks for brands

So much for why performance advertising on the internet works.  But why isn’t the internet a good place to do brand advertising?

Part of the reason stems from the same fact that makes the internet such a natural place to do performance advertising.  Whilst people being productive are likely to respond positively to ads that help them meet their immediate goals, they are likely to respond very negatively to anything that distracts them from their goals, and that includes brand advertising.  So most of the locations on the internet that work for performance advertising just won’t work for brand.

But that still leaves plenty of online locations where people aren’t being productive.  Social networks, video sites, newspaper sites etc. should all be perfect places to do brand advertising.  And they might well be.

The trouble with the vast majority of advertising on these sites is it’s boring.  Standard banners and skyscrapers, even with a little bit of rich media thrown in, make for poor platforms for changing the way people emotionally respond to different brands. They’re small, they’re unconvincing, and they’re easy to ignore.  That’s why they’re worth so little to advertisers.

This type of advertising has a lot of similarities to advertising in offline media.  It’s based on an age old formula that looks tired offline and totally dead online, namely that consumers will put up with advertising they don’t want, if it means they get to access content which they do want, for free.  TV companies have enough difficulty trying to get advertising content in front of viewers armed with PVRs and hundreds of channels.  On the web there’s no way of forcing an ad on a disinterested consumer – assuming they haven’t already installed an ad blocker, they can simply ignore it (if there’s relevant content on another part of the page) or leave the page all together if the ad really bothers them.

When brand advertisers learn how to use the web effectively

So is the web fundamentally a useless place to do brand advertising?  Obviously not…  Brand advertisers, web publishers, and everyone else in the display game need to stop thinking in terms of the old media way of doing things, and start thinking in terms of the new.

The web makes it impossible to foist a weak bit of content, in the form of an ad, on a reluctant consumer.  So brand ads are only going to work on the web if consumers want to watch them.

But advertiseres can do better than produce ads that people want to watch.  The web presents opportunities for interested consumers not just to consume ads they’re interested in, but to actively engage with them.  My favorite example is the Cadbury’s ad, featuring two school children moving their eyebrows to electro music. The campaign was not just hugely entertaining (with people watching it on Youtube countless times), but through an associated site, consumers recorded and submitted their own versions.

There are other examples too.  Countless branded, viral apps are developed for Facebook, for example, to promote film releases.  Typically games that users play, these provide consumers not only with ways to engage with brands they’re interested in, in a fun and rewarding way, but to invite their friends to participate:  a much more effective way of convincing a reluctant consumer to engage than forcing an ad down their throat.

A more familiar example of successful advertising is the type typically used on UGO.  UGO offers display inventory, but ads need not be confined to the standard IAB dimensions, giving advertisers on the site much more scope for creating something entertaining that visitors to the site enjoy watching.  (Of course, it helps that UGO is a site that people visit to be entertained rather than to perform chores.)  The recent campaign run by Valkyrie is a nice example.

Implications for the display market

If people in the ad industry were to make the switch, described above, from producing unentertaining ads to force onto large numbers of people and instead produce high quality content that a small segment of users engage with, the whole dynamics of the market would change:

1. Volumes of inventory that advertisers would be willing to pay for would decline, as advertisers focus on quality rather than quantity of inventory, to those segments of the userbase that were interested in engaging with them.  However, they would be willing to pay more for those “slots”.  (If indeed a “slot” is still the right way of thinking about the advertising platform.)
2. The whole notion of an ad network or exchange seems a bit antiquated:  these are places that match advertisers and publishers, with little room for consumers to participate and choose which advertising content they’d like to engage with
3. The whole idea of an ad server, as distinct from e.g. a content management system, may need to be rethought as consumers come to view advertising as just another form of content that they pick and choose to engage with or not. (In fact this distinction is already blurring for a different reason: major publishers are starting to route ads through their content delivery systems to circumvent users’ ad blockers.)  We may move to a world in which advertising inventory isn’t clearly demarkated on a web page the way it is today, with the whole industry moving towards different ways of thinking about the ad “volumes”
4. Advertisers and publishers are going to have to work much harder to compete for viewers attention, as advertising “escapes” out of the dimensions circumscribed by the IAB

Let’s see if people are still talking about the “display ad market” in 10 years time…

Few advertising technologies are as misrepresented as behavioural targeting. Remarkably, BT is both over- and under-hyped. There are three key reasons why this is the case:

1. The term “behavioural targeting” is hopelessly vague and ill-defined
2. The vast majority of what is called “behavioural targeting” today is limited in effectiveness (especially when compared with the promise that behavioural targeting looks to fulfill in the future)
3. A large number of challenges have to be met before the promise of BT can be delivered – challenges that the industry (or at least coverage / hype around the industry) neglects

What is “Behavioural Targeting”?

Let’s start with the easy bit, the “targeting”. Advertising on the internet can be, and often is, targeted to specific individuals, meaning that two people looking at the same website will see the same content, but different adverts.

Advertising can be targeted on a number of different criteria. It can be contextually targeted e.g. shown to people viewing websites of a specific content category, or demographically targeted e.g. shown to single, middle class males aged 25-40.

Behavioural targeting should mean targeting based on a users behaviour, typically on specific websites or across all the websites they visit. Even this limited definition of behavioural advertising covers a huge number of practices, from advertising a book that someone has browsed on Amazon back to them when they’re on their favorite newspaper site (retargeting), to building target segments to advertise a specific campaign to (e.g.” investment savvy individuals”) based on statistical and even psychological analysis of people’s viewing habits across a wide-range of sites.

However, the “promise of behavioural targeting”, which is never fully-articulated, but underpins the way many people (esp. privacy advocates) think about behavioural targeting, encompasses an even broader vision of what behavioural targeting might be. And that is that individuals might be targeted based not only on their “behaviour” on the internet, but their offline behaviour, including where they go, what they wear, what they purchase, who they talk to, how they spend their free time etc. Many are understandably spooked out at the idea that advertisers might start to understand us better than we understand ourselves, giving them enormous power over us.

What does behavioural targeting look like today?

BT today is very limited. Basic retargeting is common: if you’ve spent a lot of time browsing digital cameras on an e-commerce site, chances are you’ll start to see camera ads with surprising frequency when you’re on unrelated sites. On large web properties e.g. portals, if you spend a lot of time on the finance section, you may find finance-related ads when you’re browsing the sports section.

What isn’t happening today, at least not on any kind of scale, is advertisers taking the massive volumes of web browsing data across large numbers of individuals, and working out things about those individuals that even they don’t know about. Nothing even close to that is happening at scale.

The challenges that need to be met before the “promise ‘of BT” can even begin to be realized

1. BT companies need to get good at using large volumes of data to making intelligent inferences about someone’s behaviour and hence their likely response to particular adverts:
   • Currently the inferences made are enormously simple e.g. “this person visited a BMW site therefore they like BMW, let’s advertise a BMW to them”
   • A huge amount of data is collected whenever individuals visit websites. (More on big data in a later post.)
   • We don’t yet know whether that data can be used to make more intelligent inferences, if it can, which bits in the data are relevant, and what kind of analysis is necessary
   • Part of the problem is not just the enormous volume of data, it’s the fact that we have a limited understanding of how different people react to / engage with different adverts, so aren’t yet in a position to tie those differences in reaction to differences in their behaviour prior to seeing the ad. (More on this in a later post.)
2. Individual consumers need to actively take part in the targeting process, rather than being passively segmented and targeted:
   • It is often easier and more reliable to find out that e.g. an individual takes fashion seriously and which brands / styles they like by asking them rather than inferring it from their online behaviour
   • For BT to really add value, it has to build on the information consumers are willing to directly share with advertisers (more on this in a later post) by saying something more, rather than revealing things we either already know, or should be asking for.
   • In order to do this, consumers need to have a voice / input into the BT process. This is actually something Google is already letting people do:

     

3. BT companies need a robust way to identify individuals across across platforms, on and offline
   • In order to tailor advertising to a particular individual, you need to know that the person you’re serving the ad to is the individual that you think he / she is
   • Today, the vast majority of targeting is done through cookies. This is a really weak solution: cookies identify a particular browser, not the person looking at the browser. As a result, it can’t differentiate when two different people use the same browser, or work out when the same person is connected to the internet using a different browser (e.g. on her mobile phone)
   • The rise of social networks demonstrates that where there’s a benefit, individuals are more than happy to identify themselves accurately online. Indeed, technologies like OpenID or Facebook connect could provide the “glue” required to realize the promise of BT. (More on this in a later post.)
4. Individuals need to understand and be able to control who has what data on them
   • There is simply no way that long-term, consumers will tolerate companies collecting large volumes of data on them, and using that data to manipulate them without any comeback
   • At Keplar we believe that people will be willing to input into the BT process, and give companies data on themselves, if there’s a benefit to them. For now, there’s no clear benefit, and when people experience current attempts at BT on the Web, they find it deeply intrusive – hence the current furore around BT

The promise of BT is huge. Advertisers will always care more about whom they’re advertising to, and how they’re likely to respond to their ad, than where an ad is placed. But until the challenges outlined above are addressed, we’re stuck with a very limited version of BT, one that falls well short of “the promise”, but is enough to spook privacy advocates out. The worst of all worlds…