Posts Tagged ‘online advertising’

The Fate of the ISP-Based Advertising Industry

May 5, 2009

The premise:  Instead of building an advertising network to collect user behavior information, work with Internet Service Providers (ISP) to get access to subscriber data (behavioral, geographic, and demographic) and enable the serving of highly targeted online advertising based on a deeper understanding of the user.  While an ad network may have a footprint of tens or even hundreds of thousands of sites, the ISP has full visibility into all aspects of user activity, including all websites visited as well as all user queries across all types of search engines. It is analogous to standing over a user’s shoulder and watching his online activity all day, every day. Now add the fact that you know where the user lives down to his specific neighborhood. Overlay the PRIZM segmentation and/or US Census data and you have a pretty darn good idea of who a given user is and what he or she is interested in at any point in time.

Companies in this space include Phorm (UK-focused), NebuAd, Adzilla, Front Porch, Kindsight (formerly Project Rialto), and Feeva.  Each of these companies has (or had) a somewhat different approach, but fundamentally they are (or were) all trying to supercharge the ad ecosystem via ISP-derived user data. NebuAd has dropped the ISP-based model and Adzilla is no longer in business. Full disclosure: I was an employee of Adzilla for a large part of 2008 before the company ceased operations late in the year.

ISP-based advertising is supposed to be a win-win-win proposition for all involved, at least in principle:

  • The ISPs finally get a share of the elusive online advertising pie.
  • The ad networks get richer user data that they can use for improved ad targeting, providing superior performance to their customers (advertisers), and thus better monetizing their inventory.
  • The advertisers get to target their products and services to more relevant groups of potential customers.
  • The ISP-based ad technology vendors take their cut for enabling this new advertising model.
  • The users get to see more relevant advertising and thus have a better user experience.

Unfortunately, it turns out that users aren’t willing to go along with this scheme for four main reasons:

  1. “Better advertising” isn’t a real value proposition for the vast majority of users.
  2. Users feel that their online behavior data belongs to them, not to the ISPs.
  3. Users see ISPs as “dumb pipes” that simply provide access to the Internet – a service for which subscribers are paying monthly fees.
  4. Subscribers expect ISPs to respect their customers’ privacy.

In short, the value-exchange is shortchanging the users.

Thus, throughout 2008, the ISP-based advertising industry effectively fell apart in the US, led by the actions of NebuAd, which generated significant controversy and even led to Congressional inquiry (more details can be found here). NebuAd, which was really pushing the envelope of user privacy, has since exited the ISP-based ad model. Similarly, my former employer, Adzilla, shut down operations later in the year, after the double-whammy of the privacy issue and the economic downturn led its investors to not continue funding the operation.  In the United Kingdom, Phorm initially got off to a good start, but the company has recently been struggling with similar privacy issues, with the European Commission and Sir Tim Berners-Lee as two of their many opponents. It does not help that Phorm was a spyware company (121media) in its previous incarnation.

So where does the industry go from here?  It is clear that using the original “opt-out” model of tracking users by default and asking them to opt out is not going to fly in most markets (most notably the US and EU). There have been comparisons to Google, a company that has vast amount of information on its users.  The argument is made that if Google can track its users (via its search engine, AdSense and DoubleClick ad networks, Gmail, etc.), why can’t the Internet Service Providers?

The problem with this line of reasoning is that Google provides a massive amount of value to its users without charging them a penny. In return for its set of free services as well as a huge amount of user trust and loyalty built up over the past 10 years, users are willing to allow Google to use their non-personally identifiable information (non-PII) for monetization. The ISPs, on the other hand, already charge users for their service.  In addition, users tend to have acrimonious relationships with their service providers thanks to decades of largely bad customer service (think about your interaction with your phone or cable company here), and loyalty to ISPs tends to be nil.

There is too much money on the table for ISPs to just walk away from the ISP-based advertising industry. And they do not want the Web giants (Google, MSFT, Yahoo!, etc.) to be the sole owners of the advertising pie and of the user. Therefore, it is inevitable that the industry will push forward with an “opt-in” model. The challenge for the industry is to figure out what to give back to the user in order to get them to opt-in to these types of “intrusive” services.

So what can ISPs give back to the user? Here are some suggestions that are probably worth investigating:

  • Discounted ISP service – The economics need to make sense for this to be viable (i.e. ISP needs to make more from the ad world per subscriber than the revenue it is foregoing via the discount), implying a very high level of sophistication and performance in user targeting.
    • Taken to its extreme, this could become ad-sponsored free ISP access. This was the model that Juno/NetZero used back in the 1990s, making it work in a dial-up world. The big challenge here is to make this work in the higher-value broadband world, using less obtrusive and more relevant advertising.
  • Faster level of service – This could become less and less enticing as broadband service in general gets faster and faster.
  • Security – Can you go beyond desktop based security mechanisms and put security in the network? Anti-virus, anti-phishing, firewalls, etc.
  • Parental controls– Similarly, can you put parental control functionality in the network, so the subscriber won’t have to configure it on a device by device basis?
  • Real privacy – Take the current trend of browser based privacy functionality and put it in the network. A Privacy Mode could make the user truly anonymous and un-trackable as he surfs the Web.
  • Personalization – With the user opting in to have his non-PII data and behavior available to the ISP, provide technology to personalize not just the ISP’s Web properties but also websites and content across the Web. This would require that the ISP share user data with the Web ecosystem (with user consent), allowing any participating site to personalize its content to the user regardless of whether the user has an account on or logged into the site. The possibilities for innovation here are endless!
  • Single sign-on – Provide a single user identity that can be used at any Web destination. This could be done via password management tools or ideally, via a real identity management approach such as OpenID.

These are just some of the things that ISPs could provide to users to get opt-in into their ISP-based advertising platforms. Multi-platform ISPs have many other options, ranging from IPTV based solutions to mobile applications.

I’m looking forward to seeing how this will all play out. I invite you to share your thoughts in the comments.

First Impressions of Facebook’s Ad Platform

April 15, 2009

As a long timer in the online advertising world, I have  had the opportunity to use many of the major advertising platforms, particularly the ones on the search side. I’ve even had the privilege of participating in the launch of Yahoo’s Panama platform a couple of years ago. Today, I’d like to offer my initial impressions of my recent experience with Facebook’s “performance” ad platform.

There has been lots of discussion in the media and blogosphere about social media advertising, the bulk of it decrying the poor monetization on the two major mainstream social networks — Facebook and MySpace (there doesn’t seem to be much chatter about LinkedIn, whose ad offerings I will review in a future post, and the second-tier social networks tend to rely largely on  Goolge AdSense). Last week, I decided to test out Facebook Ads, the company’s self-serve advertising platform. Here, I will share my experience to date with the system and provide suggestions on areas of improvement.

The sign-up process:

The sign-up process for Facebook Ads was relatively straightforward and intuitive to the experience online marketer. As soon as you decide to sign up, it kicks you right into the ad creation process. The first screen is divided into three areas: Design Your Ad, Targeting, and Campaigns and Pricing.

You first design the ad by filling out the ad creative (title & body text), upload a 110×80 image, and enter your destination URL. Facebook allows you to choose a traditional web page or a Facebook Group or “fan” Page as the destination URL. The last step in the ad design process is to choose whether you want to participate in “social actions”, which matches the user’s friends’ activities about your company or product (such as “became a fan of”, “commented on”, “is attending”, etc.) with your ad. This could be a bit confusing for an advertiser who’s not an active Facebook user, but it could potentially result in a big boost in the number of views & clicks your ad might get.

The second part of the ad creation page is all about user targeting, which is an area where social networks have a tremendous advantage over other forms of online advertising because of the richness of data that they have about the user. Here, you can select your geographic target by country, state/province, or city and surrounding area. Facebook allows you to type in your targeting criteria and uses “search assist” type auto-completion functionality to match and pick your targets, which could be multiple. You can also narrow down you audience by sex, age range, keywords (which matches free form user profile information such as favorite activities, books, etc), education level, workplaces (i.e. companies), relationship status, as well as language. As you pick your targeting criteria, Facebook provides an estimate of the number of people in its user base that match those criteria.

Lastly, you provide a campaign name for your ad (or select an existing campaign if you’ve created one previously), a daily budget for the campaign, as well as a schedule for running the ad. You then decide on a pricing model (CPM vs. CPC) and put in your bid. Facebook offers a bid range suggestion based on what others are bidding for your target demographic. This is a good guide, but I wouldn’t overly rely on it — test and tweak/optimize is the approach I would take.

Upon submission, Facebook allows you to review the ad along with the major targeting parameters and place the order by adding credit card information. You then have to wait for your ad to be approved (apparently by a human reviewer) before it can go live.

Incremental improvements I’d like to see, coming from a long time search marketer:

I created a single ad for an SEM consulting site that targeted the entire US, an age range of 25-50, and college grads. Facebook Ads estimated that my target audience was 2.2 million people. I used CPC pricing model with a max bid of $0.20, which was below the suggested range of 0.55 to 0.72 USD. I’ve had the ad running now for the past six days. I’ve received over 4,000 impressions and so far only 1 click, resulting in a CTR of 0.02%. Admittedly, I have not spent any time optimizing anything — basic vanilla creative, no targeting or bid updates — and am not leveraging Social Actions. No doubt these numbers could be significantly improved through testing and optimization.

Given my week-long experience with the Facebook Ads platform, I’ve already identified a few areas where I’d love to see Facebook augment the system:

  • Additional targeting parameters:
    • Zip code targeting — search marketers are used to identifying locations by zip code even if the 
    • Workplace “type” – e.g. enterprise software company, non-profit, CPG, etc. This would be especially valuable for B2B advertisers who want to target entire industries, and identifying all the major companies in an industry and inputting them one at a time is currently unwieldy.
    • Job position — e.g. Sales, Marketing, IT, research, engineering, etc.
    • Job level — e.g. Analyst, Manager, Director, VP, etc. This would be another indispensable tool for B2B marketers.
  • Forecasting tool — It would be nice to get a forecast of the number of impressions and  clicks you may get with the set of targeting & bid parameters you’ve selected. Search marketing platforms have this capability and marketers have become used to it.
  • Ad rotation/testing/optimization — In its current form, Facebook is tying the ad creative to the ad targeting and bidding parameters. If an advertiser wants to test different creative/offers, he would need to create a new ad (he can start by duplicate an existing one) with the new ad copy/creative and make sure to set all other targeting and pricing parameters to be identical to the original ad. It would be much easier if Facebook allowed marketers to add different variations of an ad to the same “ad group”, much like it is done in the search marketing world. They could also provide “optimization” by shifting impressions towards the better preforming (higher CTR) ads. This would not only make it easier for advertisers to more easily test and optimize their creative, but over time, it would lead to better clickthrough rates on Facebook as marketers test and optimize their creative.
  • Faster ad approval — While I only had to day one day to get my single ad approved, I would like to see Facebook move to a more automated, near real-time ad approval process. Again, as a search marketer who once dealt with Overture’s lengthy approval cycles, I’m now used to this instant approval process. There is the technology out there to make automated ad approval possible, even with image ads. Overall, this is a relatively minor gripe, however.
  • Day-parting — This functionality is now a standard feature on practically all online ad platforms. While it’s less important than other targeting capabilities, there are many use cases where day-parting could be a valuable targeting tool.
  • Better reporting — Admittedly, I have not spent a great deal of time working with Facebook Ads reports, given my ~3K impressions and 1 click! 🙂 However, here are a few things that I would want to see in future reports:
    • Geo-targeting reports — where did my traffic come from, where are the clicks coming from vs. impressions vs. conversions, etc.
    • Conversion metrics — this would necessarily require the advertiser to instrument his/her site with some type of Facebook or 3rd party analytics.
    • Placement report — Where/what position did my ad appear in, on average. Also would be helpful if they could show me how many of my ads missed out on impressions because my bid was too low.
    • Near real-time results — It appears that there is about one day lag time in Facebook Ads reports.  I would like to see this evolve into a near real-time reporting, ideally. At the minimum, 2-3 hours of lag time would be acceptable.
  • UI/Usability tweaks
    • In the main Ad Manager page, when I change the time period (e.g. Today, Yesterday, Ad Lifetime, etc) or the graph parameter (i.e. click, impression, or CTR), click over to a new page, then come back to the Ad Manager page, my selections are overwritten by the default values. The system should remember my previous selections and show me what I was looking at the last time I was on that page. This behavior should be maintained at least throughout my current user session on the platform.
  • Facebook Pages (I have not tested the Pages functionality and am only speculating here… this functionality may already exist.)
    • Provide “engagement” reporting. Show me the “virality” of the page – how many people came through to the page by clicking on their friends’ links vs. searching vs. directly, etc. Of those, how many joined the group or became fans of the page? What types of conversations did these actions lead to? Can I identity the influencers for my company/product, and what are their primary characteristics? I may be asking for the world here, but progress along these lines will surely be made.

That is it for today. I will have more to say as I get additional experience with the platform. But overall, I’m impressed by what Facebook has accomplished in a relatively short period of time, and the potential for user targeting on Facebook is huge and I’m excited to see where it goes over time!

In an upcoming post, I will share some ideas about how Facebook could better monetize their platform. Stay tuned.

The Future of Online Advertising

March 31, 2009

TechCrunch is hosting a Steel Cage Debate on the future of the online advertising industry between Danny Sullivan, the search guru and editor of SearchEngineLand, and Eric Clemons, Professor of Operations and Information Management at Wharton (UPenn). Clemons started a blogstorm on TechCrunch last weekend when he put up a post titled “Why Advertising is Failing on the Internet.” I had major disagreements with Clemons’ article and posted my response on TechCrunch, debunking many of the assertions he had made in his article.

My thoughts on the Steel Cage Debate follow:
Overall, I’m squarely on Danny Sullivan’s side. While he doesn’t necessarily address all of Prof. Clemons’ arguments in the original post and largely focuses on the search advertising area, he makes convincing arguments to debunk Prof. Clemons’ general premise using largely qualitative arguments based on his 15 years of experience in the field.
Thoughts on a few specifics:
  • The idea that search advertising is “misdirection” is overblown and shows a total lack of experience in this area on Clemons’ part. He focuses on the topic of competitive bidding on brand term searches and mentions that he’s been “hearing this complaint from senior vice presidents in travel companies for years, and this year the chorus has been joined by retailers and manufacturers” and implies that Mr. Sullivan is out of touch. Being deeply involved in search marketing in a variety of roles during the last five years, I have come across this argument on a few occasions, but this is far from being serious enough of a problem to dissuade advertisers from using search advertising as a medium. Moreover, the vast majority of searches do no involve brand terms. In short, search marketing brings value to all constituents in the equation: Users get advertising/offers/content that is highly relevant to their searches; Advertisers get to show their messages to highly targeted and engaged consumers; and publishers (i.e. Google/Yahoo/MSN and their search partners) get to monetize the free search service that they provide to the vast masses. This 3-way value exchange is what makes search marketing a unique and very effective advertising/marketing medium. Sullivan deftly takes apart this misdirection argument using his 15 years of experience in the search space (vs. Clemons’ non-experience).
  • In fact, Google (and Yahoo and MSN to a lesser degree) are in a constant battle to ensure that there is as little misdirection as possible, primarily through the constant updates of their Quality Score algorithm. Their goal is to make sure that the most relevant advertising shows up for any given search, resulting in a win-win-win scenario for all three parties in the value-exchange (user-advertiser-Google). While one can argue that Google is self-serving and does these updates to continually increase their RPS (revenue-per-search), one cannot deny the associated benefits to the user experience as well as to the “right” subset of advertisers.
  • Additionally, search engines realize that brand searches could potentially present potential “redirection” problems to brand owners and have tried to address it in a variety of ways. For the last few years, Yahoo! did not allow non-brand-owners to bid on popular brand terms.  More recently, Yahoo! has developed functionality to provide better “ownership” of these searches to brand owners, with the recent launch of the Rich Ads In Search (RAIS) product, of which I was an early contributor and proponent while working at the company. No doubt Google and MSN will follow suit with similar offerings.
Beyond the search misdirection argument, Prof. Clemons’ main argument is as follows:
  • Users don’t trust ads.
  • Users don’t want to view ads.
  • Users don’t need ads.
  • Ads cannot be the sole source of funding for the Internet.
  • Ad revenue will diminish because of oversupply of inventory and will be replaced by other revenue models such as micropayments and subscription payments for content.
  • There are many other business models that will work on the net.
Danny Sullivan does not directly address these points, so I will do so here:
  • Users don’t trust ads.  I would argue that it does not matter. Users know that advertising is biased, but there have been countless studies that show that advertising has a significant impact on user behavior in favor of the product/service being advertised.  As a matter of fact, the entire offline advertising industry is based on this premise and its astronomical growth over the past century is proof that trust is not necessary for advertising to be effective.
  • Users don’t want to view ads.  This is naive.  I offer two counter-arguments to this. One, consumers do not need to like advertising for advertising to work. They merely have to learn to live with it, understanding that the high quality free content that are viewing is only possible because of advertising. In general, no one likes TV advertising but that has not prevented the continual growth of TV broadcast and cable industries as well as the massive television advertising industry. Two, the Internet has the potential to make advertising a medium that users may actually like and demand. With personalized offers for products, services, and information, the advertising could be much more targeted than what’s been possible in traditional offline media. Search advertising is by far the best example of this and all indications point to similar developments in other online ad channels. Danny Sullivan makes a good argument for online advertising in this regard, pointing out that online ads are much more targeted than offline advertising and are much less of an interruption to the viewer.
  • Users don’t need ads.  My response: So what? People don’t need big screen TVs or $5 Starbucks lattes, but those things still exist and thrive in our economy (well, at least till very recently 🙂 ).
  • Ads cannot be the sole source of funding for the Internet.  I absolutely agree with this (as would anyone else I’m guessing), but this is not much of an argument to argue that online advertising will fail.
  • Ad revenue will diminish because of oversupply of inventory.  I think that non-search ad pricing will diminish because of oversupply of inventory, but overall ad revenues will continue to increase once economic conditions improve. The structural changes in the media consumption and advertising industry will necessitate this — specifically, massive flow of consumer attention to online and mobile channels, followed by the flow advertising budgets to these media. This is already taking place in large ways and will continue to take place as soon as the economy improves. Additionally, innovation such as behavioral targeting, vertical content aggregators such as Adify, and dynamic ad formats such as Yahoo! Smart Ads and Tumri will help maintain premium pricing for highly targeted advertising, counteracting the impact of oversupply of inventory.
  • There are numerous other business models that will work on the net.  I absolutely agree with this, but again, it does not necessarily imply the downfall of the online ad industry. There can be multiple monetization models in existence concurrently (such as ad-supported, subscription, micropayments, virtual currencies, hybrid, etc). However, at this point, there hasn’t been a non-ad-based model that is poised to dominate, yet. We shall see. Danny Sullivan points out that some of these have been tried (e.g. access to newspaper content) but haven’t been very successful, leading to ad based models.
In retrospect, Prof. Clemons’ original premise wasn’t clearly defined. What did he originally mean by online advertising will fail? That online ad revenues have peaked and will continue to decline for the foreseeable future? He wasn’t specific enough. In his final rebuttal, Prof. Clemens says that online ad revenues will end up being less than 20% of overall online (non-ecommerce) revenues, and as such online advertising would have failed. No one knows what the breakdown will look like in 5-10 years, but I am willing to bet that overall advertising spending will be significantly higher in five years than it is today.