经典入门知识

June 02, 2008

Online Advertising Business 101, Part I - The Online Advertising Value Chain

When you spend as much time as I do examining the workings of the online ad industry, it's easy to forget that, to many people, it really is pretty opaque. Not only is it characterized by some of the most complex and scalable technology in the world, but it also has its own, pretty unique, economic model to boot.

I've lost track of the number of times I've been asked by people, even super-smart colleagues from within Microsoft, "so, how does the online ad industry actually work?" So I thought I would attempt to provide a bit of a primer through the medium of this blog. Who knows, maybe someone will read it and offer me a book deal ;-)

In this first installment, I'm going to take a look at what I call the online advertising value chain:

This is a simplistic view of the industry, but it does enable us to understand where the key players sit; on the demand side of the value chain, there are advertisers, and their agencies; and on the supply side, publishers, and ad networks (and/or ad exchanges).

 

What's the product?

Before I get onto the content of the boxes in the above diagram, though, we should be clear about what's in the arrows; that is, what's traded in this market? What's the actual product, here? 

The answer is advertising inventory. There are no very good definitions of advertising inventory out their on the Internet (Dave Chaffey offers one of the better ones), so I offer my own definition:

Advertising inventory is the supply of opportunities to display advertising in a particular medium.

Most people would use the term "ad impression" instead of "opportunity to display" - the reason I haven't is because I don't like to offer a definition of a term which contains another term that you may need to go and look up. The most common definition of ad impression is this:

An ad impression is a single viewing of a single ad by a single individual.

(Another reason I didn't use it is because it fails to capture the increasing complexity in ad inventory as online advertising evolves. For example, if you're serving video ads, and the user watched half of your 30-second pre-roll ad, was that an ad impression?)

In our value chain above, it's the Publishers who are the creators of advertising inventory. By building websites or software apps or video games or e-mails which are seen by lots of people, and inserting ads into these environments, publishers create a constant stream of ad inventory which, of course, they are looking to sell to advertisers. Agencies and Networks merely help the process along.

Online ad inventory is a very interesting type of good (to use the economics term). It has an incredibly short shelf life (measured in milliseconds as a page loads), but its supply is only indirectly under the control of publishers; external factors (such as a very newsworthy event) can dramatically impact the amount of inventory that a publisher has to offer. As a result, inventory prediction is a major task for publishers; I'll be returning to this topic in a future installment.

Calculating ad inventory

Another useful way of understanding ad inventory is to look at a simple example of how it's calculated. Imagine a pretty straightforward website (this blog, for example), showing pretty simple ads, with no fancy auto-refresh stuff going on (i.e. once a page is loaded, the ads don't change, so for each page impression, you get one batch of ads). How much ad inventory is created?

The answer to this is dependent on two variables - the number of page impressions on the site, and the average number of ads per page. So, for example, if my blog generated a million page impressions per month (I wish), and had an average of 5 ads per page, then the total ad inventory (if you're just using a simple ad impression model) is 5 x 1m = 5m ad impressions per month.

 

The Players

Now that we understand what's being traded, let's take a brief look at the major players in the value chain, and then I'll let you get back to whatever it was you were doing before you started reading this post.

 

The Publisher

We've already covered this guy. He's the one with the site, or the game, or the mobile portal, who is creating ad inventory and wants to sell it to advertisers to provide income for his business. Publishers are interested in maximizing revenues, but also at minimizing risk - they hate to have unsold inventory (that is, ad space with no ads in it) so they employ a number of tactics to ensure that at least something gets shown in an ad unit that they can get a little money for.

Larger publishers have their own sales teams who maintain direct relationships with advertisers and their agencies, cutting deals for big blocks of advertising inventory over expensive lunches in chic Greenwich Village restaurants. But this model only works for big publishers selling to big advertisers. Small publishers can't afford to maintain their own sales force, and even if they did, they'd never get through the doors of Ford, or CapitalOne, because they don't have enough inventory to be of interest on their own account. So these guys sell their ad inventory through Ad Networks.

One other kind of publisher it's worth calling out here is the search engine - i.e. Google, Yahoo and Microsoft. These search engines are the creators of huge amounts of ad inventory that is sold directly to advertisers and agencies, as well as running significant ad networks (see below).

 

The Ad Network

Ad Networks are essentially outsourced sales houses for publisher inventory. An ad network strikes deals with lots of publishers for their inventory and then aggregates this inventory and sells it on to advertisers and agencies. There are over 300 ad networks in existence today - a breathtakingly large number which is sure to fall soon.

An ad network's value proposition to publishers is that it can sell inventory that the publisher can't sell itself - either because the publisher is small (and so doesn't have its own sales force), or, in the case of larger publishers, the inventory is of too low-value to merit direct selling. This kind of inventory is called remnant inventory.

The network's value to an advertiser is that the advertiser can appear on lots of sites across the Internet (potentially thousands) without having to establish direct relationships with those publishers individually.

At bottom, the Ad Network business model is to buy inventory cheaply and sell it on at a higher price. There are a variety of ways of doing this, some of which I've covered before. One of the most promising is to add value to the ad inventory by adding targeting data (so that the impression can be sold for a higher price). I'll cover this in a future installment.

Networks come in all shapes and sizes. There are 'premium' networks which work with remnant inventory for large publishers; there are vertical networks which focus on a particular industry or technology (such as video); and, at the bottom end, there are contextual networks which provide an auction-based marketplace for selling keyword-based ads on small sites. You may have heard of the #1 network in this space - it's called Google AdSense.

 

The Advertiser

Advertisers also come in all shapes and sizes, of course. The big name advertisers - the folk we've all heard of - will have significant internal marketing departments, and will also likely retain the services of an agency to help them manage their marketing. Their marketing objectives will likely be a mix of brand marketing (raising general awareness) and direct response marketing (getting someone to actually buy something online now).

Smaller online advertisers are almost always focused on direct-response - getting someone to click and buy, or possibly call up. By and large, these folk can't afford to retain an agency to do their marketing for them, so they tend to go straight to certain ad networks or publishers to buy their ads. Again, the #1 in this space is our friend Google, with AdWords (the advertiser-facing side of the AdSense network).

Advertisers are motivated by getting the best ROI on their ad investment; but amongst larger advertisers some other curious motivations creep in, like wanting to make sure that a committed ad budget for a quarter actually gets spent (so that budget isn't cut the following quarter). This drives the behavior of ad agencies, to an extent.

 

The Agency

Last but by no means least, the media agency is an essential intermediary in the advertising value chain. Ad agencies usually do one of two things (or both, such as is the case with our own Avenue A|Razorfish): they create ads (anything from designing an animated banner to filming a 30-second TV ad) - known as the creative business - and they buy the media (i.e. the ad inventory) to display the ads (known as the media business). Whilst the creative side is cooler, the part of ad agencies that is relevant here is the media business. 

A media agency, then, is one that buys media on behalf of its advertiser client. The advertiser typically says "I have x million dollars this quarter for online, and this campaign I want to run. Buy me the best media to reach my target audience". It's then the media agency's job to plan a media buy that will deliver the best return for the advertiser.

At the small-business end of the spectrum, the 'media agency' morphs into small SEM (Search Engine Marketing) shops who are good at buying Google AdWords, and maybe have some SEO (Search Engine Optimization) skills to boot to boost a company's natural search rankings.

Media agencies' motivation is driven by getting as much media under their control as possible, since they're paid (particularly at the high-end) with a cut (usually something like 15%) of the advertiser's media budget. They also don't want to under-spend on the budget they've been given, as this can annoy their client (see above).

Media buying is a manual, labor-intensive process right now, and one I'll come back to. Improvements to technology will mean that agencies (especially larger ones) will have to do some pretty fancy footwork to continue to add value for their advertiser clients.

 

That's it for now. in future installments, I shall be looking at the key players in a bit more detail, and looking at some of the interesting economics which underpin the industry. In the meantime, if you have a comment, or something you'd like me to cover, leave a comment.

[Update 6/3/08: A little more info on Ad Networks added]

June 20, 2008

Online Advertising Business 101, Part II - How does adserving actually work?

Welcome to the second part of my Online Advertising 101 series. One of the things that I've discovered people fail to appreciate about online advertising is how much goes on behind the scenes in order to bring you an innocuous-looking banner ad. Whilst this is a relatively technical topic to cover in a series which is supposed to deliver insights about how the industry works as a whole, I think it's instructive to consider it because the technical jiggery-pokery that goes on gives rise to many of the issues that define the industry (for example, security and behavioral targeting). So here goes. 

 

Back in the olden days...

Once upon a time, when the online advertising industry was in its infancy (all of about 11 years ago), if you wanted to place an ad on a website, you had to call the publisher personally and broker your own deal, and then send the publisher the ad you wanted them to place on their site. The publisher would, essentially, include the ad as part of their site design (example), and, once it was live, users requesting the page(s) where the ad was referenced would see the ad, like any other graphic item on the page:

I've shown this basic scenario as four interactions, because the user first requests the page the ad is on, and then the HTML in the page tells the user's browser to fetch the ad, which is also on the publisher's web server.

 

Enter the publisher ad server

It didn't take long for publishers to tire of this arrangement - it meant that whenever the publisher wanted to change an ad, they had to make a change to their live website, which was time consuming and risky. So publishers started to put their ads onto a separate server - the publisher ad server - which would act as a repository for the ads.

Over time, publisher ad servers have evolved into sophisticated beasts, providing a wide range of functionality that helps the publisher manage and predict their ad inventory, handle billing to advertisers, and target ads.

But the most fundamental value that an ad server brings for a publisher is the ability to rotate multiple ads through a single ad unit (that is, a slot on a page for an ad). So if a publisher has 100,000 impressions on their home page a day, and they have a banner ad at the top of that page, they can sell 25,000 impressions to each of four advertisers, load the ads into the ad server, and let the ad server do the work of ensuring that each advertiser's ad is shown the right number of times.

Examples of publisher ad servers are DoubleClick's DART for Publishers (DFP), Atlas AdManager, 24/7 RealMedia's OpenAdstream, Yahoo's AMP and Google AdManager. Plus, there are a lot of small players who offer simpler systems which really just handle ad rotation, for small publishers.

In days of yore, all four ads in the above example would be loaded onto the publisher ad server (in fact, this is still the case for premium advertising on very popular sites like MSN.com, partly to enable advertising that is very highly customized for the target site). But this arrangement quickly proved inefficient for advertisers, who would have to send their ads to each publisher they wanted to advertise with; and would then have to re-send those ads when they changed, etc.

In the above model, the advertiser also lacks the ability to track the performance of their ads across multiple publishers, and is completely dependent on the publisher for data on how many times their ads were served (which still remains the primary way of calculating the price of a campaign); and they have no means of changing the delivery rules for a campaign (for example, by increasing the frequency of one ad creative, whilst decreasing another), except by calling the publishers and asking them individually.

 

Enter the advertiser (third-party) ad server

Advertiser ad servers are often referred to as third-party ad servers because the advertiser's ad server is a 'third party' to the ad transaction (the publisher ad server being the 'first party'). I'm going to refer to this server as the advertiser ad server in this post to save you the trouble of trying to remember who is the first and third party in this world.

When the advertiser ad server enters the picture, the ad request is not fulfilled directly by the publisher's ad server. Instead, the publisher ad server responds to the ad request with a redirect, telling the browser to fetch the content it's asked for somewhere else - in this case, the advertiser's ad server.

Note in the above diagram that I've removed the original call to the publisher's website; take that as a given. So the publisher ad server receives the ad request, and passes it on to the advertiser ad server.

Advertiser ad servers have also become very sophisticated over the years, but in the context of ad delivery, they perform just a couple of major functions:

• They manage which actual ad for a given campaign is delivered when an ad is asked for, implementing rules like frequency capping, ad rotation, and targeting 
• They help with managing the actual ad creatives (the GIFs of Flash files) which constitute the ads themselves 
• They track ad delivery across all the sites that are involved in a particular campaign, measuring the number of times ads were served, the number of times they were clicked, and (through instrumentation of the advertiser's site) the number of conversions generated by those clicks.

The benefit to using their own ad server to an advertiser is that they only have to upload their creative to one place, and they can plan and execute a campaign across multiple publishers easily (at least in theory - in practice, it's not that simple, but we'll come back to that point later).

The third-party advertiser ad server market is dominated by a couple of players: Microsoft/Atlas (with Media Console) and Google/DoubleClick (with DART for Advertisers, or DFA). But there is a host of smaller providers, particularly specializing in Rich Media (interactive Flash and Video) ad serving, which also play this role in the ad serving process. And there are a number of ad servers which are really front-ends for ad networks. But I'll cover both of these cases in a future post.

Ironically, one of the things that ad servers tend not to do these days is actually serve the ad - this grunt-work task is usually farmed out to a Content Delivery Network (CDN), such as Akamai, which maintains thousands of servers across the Internet to serve ads (and other content) on behalf of other people, as quickly as possible.So when a CDN is involved, the advertiser ad server simply returns yet another redirect, telling the browser where the content really (really, really) is.

 

Sheesh! Why so complicated?

You might still be wondering why there is all this complexity just to serve a single ad. Well, the answer comes when you think about the above pictures when you have a hundred advertisers, and a hundred publishers. The publisher and advertiser ad servers enable the one-to-many relationship that publishers and advertisers need to maintain. Or, put another way:

• The publisher's ad server helps the publisher manage their inventory across many advertisers
• The advertiser's ad server helps the advertiser manage their media buying across many publishers

I'm afraid to say, however, that the picture I've painted above represents a fairly simplistic view of the mechanics of ad serving. The picture becomes more complicated when you include ad networks, rich media, tracking & targeting, or ad exchanges in the mix. But you've probably read enough for today, so I'll pick up these topics in subsequent installments.

Feel free to add your comments/questions in the comments box.

July 18, 2008

Online Ad Business 101, Part III - Ad Networks

So far in my nascent Online Ad Business 101 series, I've covered the overall advertising value chain, and looked at a superficial level at how an ad 'call' is actually handled. This installment brings together themes from those two first posts, by taking a look at ad networks.

As I have mentioned before, ad networks are in the media representation business. Even the biggest publishers don't typically have the resources to sell every last scrap of their available inventory day in, day out, so they hand over a portion of their inventory - the remnant inventory - to ad networks. Small publishers, on the other hand, have no resources of their own to sell their inventory, so they have to go to the market via networks. The networks aggregate all the inventory that they have available and then sell this inventory to advertisers.

Ad networks make money by selling the inventory for a higher price than they buy it. They can achieve this in a number of ways, which I shall list in broad order of sophistication/difficulty (with the easiest first):

• Simple arbitrage: The network buys from the publisher at a rock-bottom price (because the publisher would literally make nothing from the inventory otherwise) and sells the inventory on in larger aggregated blocks at a slightly higher price. The "value add" is small - the network is simply allowing the advertiser to soak up some remaining part of their budget without having to go to lots of individual publishers. 
• Vertical aggregation: The network buys lots of small parcels of inventory in specific verticals (e.g. travel). It then aggregates the inventory for sale according to these segments, enabling it to charge a bit more. The advertiser is able to extend the reach of their campaign in a target audience without having to deal with lots of publishers. 
• Price model arbitrage: The network buys inventory on a CPM (cost-per-thousand impressions) basis, providing the publishers with a nice, reliable revenue stream. But it sells the inventory on a CPC (cost-per-click) or CPA (cost-per-acquisition) basis, reducing the risk of the inventory for advertisers (who are only paying for success), and absorbing the associated risk itself. The network makes money on the difference between the CPM it pays publishers and the "effective CPM" (eCPM) it charges advertisers. 
• Platform specialization: Advertising on emerging-media platforms such as video and mobile still requires quite a lot of specialized technology, forcing Rich Media vendors to build close relationships with the publishers that they deal with. Over time, many of the vendors in this space have gone the extra mile for their advertiser customers and turned themselves into networks, making it easier for advertisers to buy ads in these new formats across a range of publishers. 
• Behavioral targeting: The network buys inventory from publishers, and when the ad call is passed over to the network, it drops a third-party cookie. By doing this across all its publisher clients, the network can build up a profile of users by cookie ID - knowing, for example, that cookie ID XYZ123 has visited ten sites about watersports in the past week. The network can then use this information to add value to the inventory it's reselling, enabling advertisers to buy "active surfer dudes" and the like.

 

Can you give me some examples?

Sure. Here are some examples of ad networks which (roughly) map to the types above. In practice, of course, most ad networks employ a combination of the above techniques to maximize the margin on the media they represent.

Simple arbitrage: Advertising.com
No doubt my description of Advertising.com as a "simple arbitrage" network will generate howls of protest from AOL (Advertising.com's parent company). But one of Advertising.com's main value propositions is the breadth of sites and audience it can deliver. Because Ads.com deals with so many publishers, advertisers can almost always find some inventory that maps onto the audience they're looking for, and are happy to pay a (relatively) modest fee for the privilege.

Simple arbitrage 2: Google Content Network (AdSense)
No discussion of networks would be complete without a mention of Google AdSense. AdSense provides a way for lots of small publishers to make inventory available to the pool of advertisers that use Google Adwords - in addition to their ads appearing next to Google's search results, these ads can also appear on the small publishers' sites; the ads are matched with the sites on a contextual basis (the content of the site is crawled to extract keywords which then stand in for the keywords that advertisers normally bid against for paid search results).

A crucial feature of this system is that the publisher is paid on a cost-per-click basis, so assumes a big chunk of the risk - if no one clicks, the publisher doesn't get paid. Google makes its money on the margin between the cost-per-click they pay the publisher, and the cost-per-click they charge the advertiser. The value proposition lies in connecting lots of small (and large) advertisers to lots of small publishers who are running sites which have a really good content match to the advertiser's offering. In other words, if you manufacture Mongolian nose-flutes, AdSense allows you to get your ads onto all the Mongolian nose-flute fansites out there, with very little effort.

Vertical aggregation: Martha's Circle
Martha's Circle is the (rather winsome) name for the ad network run by Martha Stewart Omnimedia. It's a classic example of a publisher/media owner extending their brand (and saleable audience) by signing up sites in the same sector (in this case, lifestyle) and creating a niche network. For an advertiser wanting to reach thirty-something women with an interest in the home, this kind of network is a no-brainer when building a media plan. Glam.com is another good example, as is Fox Interactive Media.

Price Model Arbitrage: DRIVEpm
DRIVEpm is Microsoft's own advertising network, acquired with the acquisition of aQuantive last year. DRIVEpm styles itself as a "performance" network, meaning that it uses a variety of techniques (amongst them, price model arbitrage) to enable advertisers to buy inventory on a cost-per-performance basis, whilst still paying publishers on a cost-per-thousand basis. Scott Howe, former GM for DRIVEpm and now VP for the Microsoft Advertising business unit, wrote a great article back in 2005 about some of the dynamics in a performance network from the perspective of a media buyer looking to get the best ROI. Well worth a read.

Platform specialization: VideoEgg
VideoEgg is a video advertising network (the clue's in the name, I guess). Its offering is a classic mix of innovative ad unit technology (their latest offering is something called "AdFrames") with a network attached. Another feature of VideoEgg is that it offers advertisers a CPE (cost-per-engagement) model for buying video advertising, performing the same kind of price model arbitrage that DRIVEpm is doing. Their publisher audience is widget & app developers for social media environments such as Facebook and MySpace, ensuring that their value proposition to advertisers is further differentiated (essential as the online video market becomes more crowded).

Behavioral targeting: Tacoda
Tacoda is also part of AOL's Platform A unit, and markets itself as the world's "first" behaviorally-targeted ad network (a hard claim to substantiate, but equally hard to refute). Tacoda tracks behaviors of the visitors to its network of over 4,000 sites and uses this information to associate behavioral profiles with those users. It then sells inventory on these sites on a user-target-group basis, rather than by group of site or content area. These "audience segments" have names like "Family Chef" and "Photo Bug".

 

How does it actually work?

Understanding how ad networks actually serve their ads is essential in understanding how some of the above business models (especially targeting) work. I'll cover two scenarios - a small publisher/small advertiser scenario, and a large publisher/large advertiser scenario.

Small publisher/small advertiser
A small publisher will insert their ad network's ad code directly onto their site - in many cases, this is the only ad code the publisher is using, and is serving 100% of that publisher's ads. On the other side of the fence, the ad network may provide a web UI to enable advertisers to create or upload ads, and (possibly) allow the advertiser to choose which sites (or groups of site) those ads will appear on. The diagram below summarizes this (thanks to Right Media for the advertiser & publisher people icons):

Examples of this kind of system are Google AdSense and the Yahoo Publisher Network (these are often called "self-service" ad networks). The actual ad delivery model is pretty simple - the same ad server (the network ad server) functions as both publisher and advertiser ad server (the ad call path is on the left side of the diagram above).

Large publisher/large advertiser
When it comes to large publishers using ad networks to deliver inventory to large advertisers, things get more complicated. In this scenario, both the publisher and the advertiser will likely have their own ad servers. The publisher will configure its ad server to "hand off" a certain block of inventory to the network, whilst on the advertiser side, the advertiser ad server will be configured to buy a certain portion of a campaign from a network (or networks). So the ad call has to be passed from the publisher to the advertiser via the network:

 

It's the point at which the ad call passes through the network ad server when the network is able to drop a cookie on the user's machine, enabling behavioral tracking and targeting (assuming, of course, that the users don't delete their cookies in the meantime).

Of course, hybrids of the two models above also exist: large publishers will sometimes hand over some of their inventory to a self-serve network, in particular, in which case the publisher's ad server calls the network ad server, which serves the ad itself.

This picture also becomes more complicated when you consider that many ad networks will pass the ad request on to another ad network if they themselves can't fulfil it (or fulfil it economically). So, for example, a targeted ad network may receive an ad call from a user it has no information about. Rather than serve an ad for that user at a low cost (and thereby preventing that ad impression from being served to another user at a higher cost), the ad network passes the ad call on to a "value" (read: cheap) network. So in the picture above you can have two, or even three or four, ad networks passing the ad call around like a hot potato.

This game of pass-the-parcel isn't really very good for the user, who has to wait a long time to see the ad (which really hurts the advertiser most, since a slow-loading ad might as well not render at all); and it's also not great from a security point of view, because the publisher is ceding control of a portion of their site's screen real-estate to an unknown network and an even more unknown advertiser. Which is why ad exchanges are emerging which provide a centralized clearing-house for inventory, thus dispensing with the round-robin approach described above.

August 18, 2008

Online Advertising Business 101, Part IV - Publishers

Welcome to the latest installment of my Online Advertising Business 101 series. So far we've looked at the overall value chain, at the mechanics of ad serving, and the role that networks play in creating the market. In this post, I'm going to look at the source of advertising inventory - publishers - in a bit more detail. There's a lot of complexity to monetizing a website (especially a large and popular one) effectively, and this side of the industry is often poorly understood.

 

What's a publisher?

A Publisher is the original source of ad inventory. By creating a web site or other digital media, and getting people to come and look at it, the publisher creates ad inventory by placing ads alongside the editorial content of their site.

The term 'publisher' is the best we have for this party in the value chain, but it's getting a little stretched now. For example, someone who creates an online productivity app (like Picnik) falls into this category, though you might not think of them as a publisher in the traditional sense. And Google Search falls into this group too, despite the fact that it creates no original content of its own. The only other term that's used for this kind of organization is Media Owner, but that's associated with traditional content businesses (things like the New York Times), and tends not to be used more broadly.

Publishers, like all businesses, are driven by profits. In terms of the advertising side of a publisher's business (bear in mind that some publishers generate subscription revenues from their content too), profit is driven by the following factors:

• The volume of ads that a publisher can sell alongside their content 
• The price the ads can be sold at 
• The cost of selling those ads

Each of these has an impact on the business decisions a publisher makes. If you don't understand all these drivers, you won't understand publishers. Let's look at these in turn:

Volume: In an ideal publisher world, each word of editorial content would be accompanied by a hundred ads (you may have seen sites which look like they are close to this ideal). In the real world, there is a limit to the amount of advertising you can place around editorial before you start driving users away. Publishers are always looking to balance ad opportunities with user satisfaction.

Price: Generally speaking, the higher the price that a publisher can get for its ad inventory, the better. Maximizing this over time and across all the inventory that the publisher has is a huge challenge for publishers - an area known as yield management.

Cost: Publishers can incur various costs in selling ad inventory. One is the cost of sale - paying people to get on the phone to advertisers and agencies to sell the inventory, or paying a network to sell a block of inventory on the publisher's behalf. But another cost is opportunity cost. Any inventory that a publisher sells today can't, by definition, be sold again. So if a better offer for the same inventory comes in tomorrow, the difference between the two offers is the opportunity cost incurred.

I'm not going to spend any more time looking at the volume driver in this post - that really depends on the publisher's ability to promote their site and innovate on layout & ad formats to maximize monetization opportunities. What I'm going to look at instead is the relationship between price & cost - also known as the yield achieved by the ad inventory.

 

Why newspapers are like airlines

It may not be immediately obvious to you what newspapers and airlines have in common (except that, right now, both seem to begoing out of business). But actually there are striking similarities between an airline's inventory of seats and the ad inventory sold by publishers/media companies:

• Their inventory is perishable - both seats and ad inventory slots become completely worthless after a particular moment in time
• There is a finite supply of inventory - neither airlines nor publishers can easily mint new inventory to capitalize on higher demand (unlike, say, an ice-cream manufacturer)
• It's possible to charge different customers different amounts for the same inventory (known as price discrimination)
• Both types of company have high fixed costs, meaning increasing revenue by (say) 10% can have a disproportion effect on profits

What these qualities mean is that publishers and airline must both maximize their revenues by selling as much of their inventory as they can at the highest average price possible.

In the 1980's, American Airlines led a transformation of the US airline industry when it introduced its "Ultimate Super Saver Fares" in 1985. AA's key insight was that people would be prepared to pay more for the same airline ticket when purchased at short notice - which enabled the airline to sell and advertise tickets at a much lower price for advance purchases. Thus was the business of yield management born.

 

Yield management

Now, we're not talking about the road sign type of yield here (besides, as a Brit, that sign should say "Give Way"). Yield management refers to the practice of maximizing the average price received for inventory through a number of techniques, amongst them price discrimination (charging different customers different prices for the same thing).

Yield management in the online ad business does differ a little from the business of airline seats, principally because a publisher does have some flexible control over the cost of sale; additionally, volume discounts play a significant part in the sale of online ad inventory. What this means is that online ad inventory yield management is really about maximizing gross profit rather than top-line revenue.

 

Packaging & pricing

The first yield management decision that a publisher makes is how it's going to package its inventory together for sale. The larger the publisher, and the more diverse the content of its site, the more complex this process is.

The approach that most publishers take is to divide their site into a number of "page groups" or "channels", where each channel contains a group of pages which are in the same content area. For example, MSN has a number of channels including Money, Autos and Finance. So if an auto advertiser wants to buy some ad inventory, the first place they look is the autos channel, knowing that their ads will appear in a contextually relevant setting (it's just like taking ads out in the Autos supplement of a newspaper - you know that the folk opening that supplement have some kind of interest in cars, which is a good start).

What makes this process more complex is that publishers increasingly want to divide their inventory up in multiple, overlapping ways, depending on other characteristics of the site or the audience, such as "regular readers", or "18-24 year-olds". The ad inventory in these audience-based page groups can potentially be sold for a higher price, but managing the total pool of available inventory becomes a lot tougher.

The more specific to an advertiser's needs that a publisher can make its inventory packages, the higher price those packages can be sold for; but there is a practical limit to how fine-grained a publisher can be - an inventory package that is "18-24 year old regular visitors looking at the auto channel" may end up being just too few impressions (or unique users) to be worth the advertiser's while buying (because every ad buy the advertiser or their agency makes has a fixed cost of administration).

Smaller and less-sophisticated publishers tend to fall back upon a default page group that is known as "Run of site" (ROS), or sometimes "Run of network" (RON). This basically means that the publisher can run the ad anywhere they haven't sold ads in a more specific page group. As a result, ROS/RON prices tend to be close to the bottom of the pile.

 

Price discrimination, volume discounts and bundling

Online ad sales (in fact, ad sales in general) is all about price discrimination. Every ad sales person in the world knows that the first thing you do when you call an advertiser prospect is to pitch an (unrealistically) high opening price (usually known as "rate card") and wait for the advertiser to negotiate down (often by as much as 60 - 70%). But exactly how far the price comes down for online ad inventory depends on a number of factors:

Time: If the advertiser wants to get the media booked well ahead of time, they'll have to pay a premium, because the salesperson still has a long window in which to sell this inventory, and also because the advertiser benefits from the "pick of the crop" of inventory, making it easier for them to achieve their campaign goals.

Volume: The more inventory an advertiser is prepared to buy, the better price (generally) they can get. Selling out big blocks of inventory reduces the risk for the publisher that some of its inventory will remain unsold - and unsold ad space, like empty airline seats, is what keeps publisher ad sales managers up at night.

Another way of thinking of this is in terms of opportunity cost. In the diagram below, on the left, the publisher makes one great deal for a relatively small amount of inventory at $10 CPM (cost per thousand ad impressions) but ends up offloading the rest of his inventory at $1 CPM; whereas on the right, the publisher makes a lower-price deal for more inventory, ad $5 CPM, and ends up making more money overall.

So a lot of what a publisher is concerned with is risk management, and many of the decisions publishers make about their business are as concerned with reducing risk as they are with maximizing return.

Frequency capping: This topic almost deserves a post of its own, but simply put, frequency capping is a technique used by advertisers to ensure that their ads are not seen over and over and over again by the same people. It's an essential element of online ad buying because otherwise a publisher could in theory sell a million ad impressions and deliver those impressions all to the same person (that would be a very loyal reader, to say the least). So advertisers usually specify a "frequency cap" of something like 4 or 5 ad impressions per unique user, to stop this happening.

What this means for the publisher is that if they have some very active users consuming, say, 20 or 30 pages per site visit, it will be very hard to sell out all the ad inventory that these people will see because they'll be 'capped out' before they've stopped looking at the site. So publishers are prepared to sell inventory at a lower cost to advertisers who are prepared to accept a higher frequency cap.

Guaranteed vs 'discretionary' inventory: When a major advertiser is planning a big product launch, they're focused on ensuring that they get a certain amount of exposure, or reach - that is, that their message ends up in front of a particular number of people over a particular period. So publishers are called upon to guarantee that advertisers' ads will be shown a certain number of times, to a certain number of users.

Because, in practice, a publisher doesn't know exactly how much site traffic or how many users they'll have visiting on a given day, it is absorbing some risk by doing this. So so-called 'guaranteed' inventory commands a premium compared to 'discretionary' inventory, which is often passed off to ad networks and sold on a auction/spot-price basis (see below).

Retargeting: As I've mentioned, a publisher can use their own data about their audience to create audience-focused ad groups and sell these for a higher price. But retargeting refers to the case where the advertiser brings some of their own data to the table, particularly data to identify users who've had some kind of interaction with the advertiser in the past. So the advertiser may say "I want to reach this list of 10,000 users [identified by a cookie] who came to my site this week but didn't buy anything". Because of the much higher value to the advertiser of reaching these users in particular, the publisher can sell that inventory at a higher price. 

 

Remnant inventory & spot pricing

Selling online ad inventory is a bit like doing one of those puzzles where you have to fit a number of differently-sized pieces together to make a square; except the puzzle has to be done again and again, day after day, month after month:

On any given day, the ad deals that the sales team has sold get fitted into the 'box' of available inventory. Whatever space is left over is known as remnant inventory (outlined in red below):

(As a side note, sometimes the pre-sold inventory won't fit inside the box of available inventory, forcing the publisher to actually buy inventory from other sites and re-sell it to make up the shortfall. This is why some publishers are starting to look more and more like ad networks).

The remnant inventory is usually passed off to an ad network where it is either sold for a pre-determined 'floor price' (like, say, $.10 CPM), or it is auctioned off in real-time. In this sense the inventory is being sold in a similar fashion to spot-priced commodities, where the guaranteed deals are more like forward prices. Even when the remnant inventory is sold at auction, most publishers have a good idea of what the effective floor price for remnant inventory. Woe betide the ad salesperson who sells guaranteed inventory below this floor price.

What the ad network/remnant model means is that almost all inventory gets sold at some kind of price, though 'premium' publishers will only stoop so low to fill inventory spots. Even though an airline would make more money selling an empty first-class seat to London for $50 if it would otherwise be empty, if you've paid $8,000 for your seat and some stinky back-packer turns up next to you, you'll not be happy (even if the back-packer won't tell you how much he paid for his seat). Advertisers are the same - if Ford is paying top dollar for ads on the homepage of MSN, it doesn't expect to see cheapo debt consolidation ads in the same slots, even if MSN's salesfolk haven't managed to sell out all of that inventory.

That ad networks have supplied/soaked up low-value, 'filler' inventory has been an article of faith in the industry for the past 10 years, but this particular apple cart may be upset in the future as networks become more sophisticated in the targeting they're able to offer as they aggregate the behavior of users across the many sites in the network they visit and roll this data up to create targeting profiles.The upshot of this is that the floor price for a publisher is moving up, with some network prices starting to approach the lower end of the 'premium' deals that the sales team are striking. This is creating a demand from larger publishers for an 'all-up' inventory & revenue management system that can treat the publisher's available inventory as a single, homogenous block, rather than creating the arbitrary distinction of premium & remnant inventory.

 

So I think I have detained you long enough today. Please feel free to use the comments box to ask questions, add your own observations, or point out where I am grievously and embarrassingly wrong about something. I reserve the right to edit this post if I discover I have been a total dumb-ass about something.

October 08, 2008

Online Advertising Business 101, Part V – Rich Media

Welcome back to my Online Advertising Business 101 series. In my previous posts in this series, I’ve painted a somewhat simplified picture of the major players in the online advertising market, and how they interact; and we’ve looked at some of the technology that these folks use, specifically ad serving technology for advertisers and publishers.

This post looks at a significant, but hard to define, aspect of online advertising: “rich media”. I’ll attempt to explain how rich media fits into the overall picture, but also why it kind of doesn’t. You see, whilst some of the more mainstream parts of the online advertising world are starting to settle down a little, with familiar patterns of interaction emerging, and some fairly well-established standards, out at the bleeding edge of rich media, things are still evolving quickly. So the rich media ad industry operates as a kind of overlay with mainstream online advertising, with its own set of specialized vendors, and its own set of rules.

 

What is Rich Media?

“Rich Media” is an umbrella term for ads which are more interactive, or use sound and/or video to get their point across. The following are all examples of rich media ads:

• An expandable banner with embedded video 
• An ‘advergame’ (a mini-game embedded inside an ad unit) 
• A video ad served before, after or during a video clip
• A ‘page takeover’ ad (an ad which covers the page it’s on, or interacts with that page or the other ads on it in some way) 

Another more demotic way of defining rich media is “fancy Flash and video ads”. That sums it up to a reasonable degree of accuracy – there are funky ads built in Flash, and then there are video ads, either delivered through their own dedicated Flash units (in which case, they look a lot like regular Flash ads to ad servers and publishers), or embedded into the stream of a video clip. And it’s also reasonably accurate to say that in-stream video is the more challenging to get done, for a variety of reasons (which I list below).

Although rich media currently makes up only a fraction of the ads delivered on the Internet, these ads are the best-paying for publishers, and their share of online display ad spend is set to grow to around 50% in 2009, according to Jupiter, with video being the biggest contributor to this growth. In fact, video is often broken out in such predictions as its own media category – so this post could have been entitled “Rich Media and Video”.

Rich media’s very attraction – the fact that it’s something other than bog-standard banners, buttons & links – is also its main challenge. Each fancy new rich media idea from an agency brings with it its own set of implementation challenges. So a whole industry has grown up around bridging the gap between publishers – who want a simple life, by and large – and advertisers & agencies who want to push back the boundaries of what is possible with online advertising.

 

Rich Media Vendors

Because of the specialized nature of rich media advertising, and the fact that there are few to no standards in the rich media world, a collection of specialist companies – known as Rich Media Vendors (RMVs) – has sprung up to help advertisers & agencies get these kinds of ads onto publisher sites.

An (incomplete) list of some of the more common rich media vendors is as follows:

• Atlas Rich Media (now part of Microsoft) 
• DoubleClick Rich Media (now part of Google) 
• Yahoo! Rich Media
• Eyeblaster
• EyeWonder
• Pointroll
• Interpolls
• Rovion
• Unicast
• CheckM8

These vendors provide a combination of technology and services. Much of the technology side comprises proprietary ad templates and delivery and measurement technology, whilst the services focus on authoring ads and help with the ad trafficking.

Because of the vertically-integrated nature of the RMV’s services, they need to interface on the one hand with advertisers’ creative agencies, and on the other hand with the publishers (and, increasingly, networks) that will be running the ads. So a good RMV will be on the “approved” list of lots of agencies and publishers/networks.

 

What do they do?

The challenges of getting a rich-media campaign up and running are as follows:

• Creating the actual ads themselves (the creative) 
• Trafficking the ads (getting them actually placed onto publisher sites) 
• Managing and serving the creatives themselves, and measuring delivery and response

Creating the ads

Creating a good rich media ad requires a lot of skill – even a simple expanding ad requires knowledge of Flash and JavaScript, together with decent graphic design & copywriting skills. Creating video ads is a whole different ball of wax, especially if you’re aiming for something more compelling than just repurposing your 30-second TV ad for the web. The small screen-sizes available and things like bandwidth considerations further complicate matters.

The RMV will work with the advertiser’s agency to code up the ads they need – the agency may provide a relatively complete ad that has already been coded using the RMV’s template, or the RMV will take some raw creative – say, a video – and code it in their template.

Trafficking the ads

This is the really fun part. The complexity of trafficking a rich media ad is dependent on the kind of ad it is. For ads which are more or less “fancy banners” (i.e. IAB standard-sized units with interactive capabilities or video), the ad can often be served through the advertiser’s and publisher’s respective servers just like a static ad, making trafficking relatively simple. As these kinds of ads grow more sophisticated, however – expanding over the page, for example, or needing to interact with the page (such as the recent Apple ads on NYTimes.com, which talk to one another), special work needs to be done by the publisher to insert the ads onto the page.

This picture becomes even more complicated when you consider in-stream video ads. The burgeoning crop of online video sites (YouTube, MSN Video, Hulu, YuMe, Vimeo, Veoh, Joost, Google Video, DailyMotion, Blinx and more than 200 others) all have slightly different players (though many are Flash-based) which use an array of different video sizes and encoding formats. So anyone looking to serve video ads has to be able to transcode their video into the various required formats and work with the video sites to insert these ads into the video stream. 

Furthermore, many online video sites are now enabling new kinds of ads within their players, such as overlay ads; our own adCenter Labs is even looking at video hyperlinks – making parts of the video itself clickable. RMVs have to keep up with all these developments.

Unsurprisingly, no standards exist for the way in which these richer kinds of ads are implemented, so publishers that want to host rich media or video ads end up working directly with a handful of Rich Media vendors (here’s the list that AOL supports, for example) to define a set of formats that they will support. The RMVs end up acting as a kind of gateway to the publisher, ensuring that some kind of consistency and reliability is maintained.

Delivering & tracking the ads

Many rich media vendors (for example, Eyeblaster) will also take a hand in actually delivering the ad, providing their own campaign management and ad serving technology. In situations where this is the case, the advertiser can either use the RMV’s ad management system in parallel with any other ad server they already use (such as DFA, or Atlas Enterprise), or they can use their “primary” ad server to manage the campaign and then hand off the ad calls to the rich media ad server when necessary, as in the diagram below:

A third scenario is that the advertiser adopts the RMV’s ad server as their principal third-party ad serving solution – this, unsurprisingly, is the tack recommended by the RMVs themselves.

One other key reason that RMVs provide their own technology is that this makes it easier for them to offer detailed and relevant metrics about ad delivery and interaction. Because rich media is designed to be interacted with without leaving the ad (anything from a simple whack-a-mole-type interaction to something like spec’ing a new car), measuring rich media (and therefore charging for it) is a very different proposition to measuring static display or text ads, where the click (and, to a lesser extent, subsequent conversion behavior) is king.

You’ll be getting a little bored by now of hearing me say that there are no standards for rich media measurement, but they are pretty thin on the ground. Impressions is a fairly useless metric if the unit in question is a ‘peel-back’ overlay – the kind where you have to mouse-over to see any of the ad copy at all – or where the ad is a 15-second pre-roll. Likewise, there’s a world of difference between an ad where a single click signs the user up for an e-mail newsletter and an advergame where the user may click hundreds of times on the ad in an effort to whack that pesky mole.

So another function that RMVs (and the creative agencies they work with) perform is to demonstrate and quantify the value of the work they have done to their clients – coming up with engagement metrics that can be defended in front of the CMO.

 

Where next?

The vertically-integrated nature of the rich media ‘industry’ is sure to change over the next few years, as the industry looks to grow out past its home base of deep-pocketed advertisers and large, sophisticated publishers. Smaller advertisers will want to be able to create richer ads, and to be able to serve ads into richer environments. Similarly, smaller publishers would like to be able to benefit from the higher eCPMs of rich media ads. Larger advertisers, on the other hand, would like to be able to serve their fancy rich media ads across a broader range of sites.

In order to enable these scenarios, advertisers need access to systems that enable them to create or upload one set of creative and have free choice of a wide range of sites to advertise on – without having to implement the creative separately on each site. And publishers need to be able to create standardized rich media ad units which can form the supply side of this larger, more liquid market.

There are some moves afoot to bring about this kind of world. Google AdSense now supports video ads (enabling publishers to have video ads appear on their site), and AdWords text ads will appear as overlay ads on YouTube videos, enabling AdWords advertisers to appear in this context. We’re doing the same with our Silverlight Streaming for Windows Live service. And some RMVs are turning their publisher base into specialized ad networks – VideoEgg’s network being a good example – whilst traditional ad networks such as Advertising.com are branching out into video.

Industry growth won’t get past a certain point, however, without more agreement on standards. The IAB is pretty active in this space (though, bless it, it does move rather slowly), having agreed a set of rich media ad guidelines this year, and there is also a useful set of measurement guidelines available (with some nice diagrams on tracking & serving). But a set of widely-agreed measurement standards seems a little further off for the time being.

November 13, 2008

Online Ad Business 101, Part VI – Ad Exchanges

Yes, it’s time for another Online Ad Business 101 post. This post deals with one of the players that I left out of my first post about the online advertising value chain: Ad Exchanges. If you don’t know what an ad exchange is, now’s the time to learn, since these little-known companies (most of which are now owned by some extremely well-known companies, such as Microsoft, Yahoo and Google) are set to have a major impact on the way the industry works over the next few years. Our own AdECN recently announced the launch of its new “federated” ad exchange, making this post especially timely (in truth, the timing is no coincidence – the announcement was the gentle kick I needed to get this post out the door).

 

The problem with ad networks

Ad networks are great. They must be – there are enough of ‘em (300 the last time I counted). As you’ll know from reading my post about them, ad networks make money by creating a marketplace between publishers (providers of ad inventory) and advertisers (consumers of ad inventory). By connecting many publishers to many advertisers, they create some efficiency and add some value into the bargain (by providing targeting capabilities, for example). Networks sit at the center of their own little web of advertisers and publishers (apologies to Right Media for, ahem, borrowing their little people graphics for this post):

The key challenge with being an ad network is that you have to grow the supply side of your business (the publishers) in parallel with the demand side (the advertisers) – there’s no point signing up a huge batch of new publishers if you’ve no one to sell their inventory to. But doing this in practice is extremely hard.

To solve this problem, the ad networks have brokered relationships with one another over the years so that, if a network has an impression that it needs to sell, but doesn’t have an advertiser to sell it to, it can sell that impression to another network. Similarly, in the reverse case, if a network has an opportunity to sell an ad, but doesn’t have the inventory to fulfill the sale, it can buy the inventory from another network. So the picture (with each network’s advertisers and publishers collapsed to one of each) looks like this:

In the diagram above, Network 2 can sell Publisher 2’s inventory to Network 1, who sells it on to Advertiser 1. Similarly, Network 2 can buy inventory from Publisher 1 (via Network 1) and sell it to Advertiser 2. In the real world of ad delivery, the ad call is redirected from the publisher to Network 1, and then to Network 2, before finally being redirected to the advertiser’s ad server.

If you add another ad network to the mix, then each ad network can forge relationships with the other two, and trade impressions in much the same way:

If there were just two or three ad networks in the world, this might not be a problem. But of course there aren’t – there are three hundred. But each ad network can’t have a relationship with every other ad network; each network would have to maintain 299 relationships, which comes to (299 + 298 + … + 2 + 1) = 44,850 relationships!

So instead, the networks form a kind of ‘daisy chain’ – each network passes off some portion of its inventory to one or more others, which in turn pass some of this off to their own partner networks, and so on. So a single ad impression can pass through half a dozen (or more) networks before finally being fulfilled:

Of course, the diagram above dramatically over-simplifies the picture; each of the networks in the chain will have multiple relationships with other networks, and so inventory can take a series of routes from a particular publisher to an advertiser.

This daisy-chaining sucks, big time, for the following reasons (and others):

• Each network has to maintain multiple bilateral arrangements with other networks, sucking up time and technical resource 
• The more networks there are in the chain, the longer the ad takes to serve 
• Each network wants to take a cut of the cost of the inventory,cutting into the publishers’ margins 
• In addition to the margin problem, it’s very difficult for publishers to get the best price for their inventory, since each network in the chain has to make the best guess around the network it thinks will deliver the best price 
• The publisher has very little or no control over the quality of the ad that ends up being displayed; it’s very easy to insert poor-quality or even malicious ads into the system 
• If the ad is clicked, the click path is very convoluted, and is liable to hijacking (another security vulnerability) 
• The system is completely opaque to the publisher (no network can provide a comprehensive list of what actual ads they served, or had a hand in serving, on a publisher’s site) 

 

Enter the Ad Exchange [fanfare]

By now I can almost hear you crying, “But surely there must be a better way!” Well, you’ll be glad to know, there is. Rather than the ad networks all dealing with each other directly, we need some kind of impartial intermediary which can act as a central hub through which the networks can trade. An ad exchange, if you will. Of course, exchanges (especially commodity exchanges) have been around for a long time – as I noted recently on this blog, everything from pork bellies to weather futures are traded today on exchanges around the world today.

The Chicago Mercantile Exchange (one of the biggest exchanges in the world) is even launching an exchange for credit default swaps – those bad-boy financial instruments the opaque trading of which (in a manner which is alarmingly reminiscent of the ad network relationships above) have had such a hand in getting us into the mess we’re all in now.

So if even something as evanescent as a CDS can be traded on an exchange, why not ad inventory? Well, it turns out there are some fairly interesting technical challenges, since the volume of transactions is extremely high and transactions must be completed within a few milliseconds; but those are surmountable. By adding an ad exchange into the picture, the trading relationships look as follows:

Now each ad network has just one trading relationship – with the exchange. So if there are 300 networks, there are 300 relationships, and every network is just one ‘hop’ away from every other network.

What this is means is that for a given ad impression on a publisher site, the network that owns that impression can say to the exchange, “what am I bid on this impression (one careful owner, full service history, nice neighborhood, good references, etc)?”. The exchange can then hawk that impression to all the other networks and solicit bids. Depending on the data that is attached to the impression (or a cookie that one or more of the other networks may recognize and be able to attach data to), the various networks may be able to sell that impression for a greater or lesser amount. So the bids come in, the winning bid is selected, and passed back to the originating network; and if that bid is better than what the network could get from its own advertisers, it wins, and the ad is served.

Crucially, there are only ever two networks (plus the exchange) in this transaction. So each network will take a cut of the impression price, and the exchange will charge a flat transaction fee (this is essential to maintain the exchange’s impartiality – taking a cut would introduce bias). Just having two networks in the transaction means more money for the networks and the publisher, and possibly better pricing for the advertiser. So everyone wins.

 

There’s more…

The benefits of moving to an exchange-centric model for ad inventory trading don’t end there. As I said at the beginning of this post, one of the irritating things about running an ad network is having to match demand to supply – as networks grow, they have to recruit both advertisers and publishers. The network model allows one-sided participants to flourish, dramatically increasing the range of ways in which businesses can participate in this market.

In the above example, Network 2 doesn’t actually source any inventory direct from publishers – it gets it all from the network, and focuses on being great at selling that inventory to advertisers. Another (perhaps better) name for the kind of company that does this is a Media Agency. Havas has just announced its intention to do something similar to this. You could also easily imagine the likes of Amazon and eBay – both of which have huge rosters of small advertisers, but no corresponding publisher base – to participate in this fashion.

Similarly, Network 3 above has decided to do away with its advertiser customer base and just sell all its inventory to the exchange. In this sense it becomes a bit more like an ad sales house or publisher aggregator than a true network. A company in this mold might be Six Apart, creators of the TypePad platform, which has lots of publisher relationships (including with me), but no advertiser relationships to speak of.

And there’s a third scenario which is even more interesting, which is that the exchange model makes it possible to add value (and make money) without trading any inventory at all. A company like Nielsen might choose to sell the data it has on internet users to the exchange, helping to drive up inventory value, and taking a cut of transactions that use its data.

Building a true ad exchange is non-trivial, mind you, which is why the most significant efforts in this space are courtesy of the Big Three of Google (most visibly via the DoubleClick Advertising Exchange), Microsoft (with AdECN) and Yahoo (via the RightMedia Exchange). And a big question-mark still hangs over how exchanges will earn money – the revenue model is well understood (transaction fees), but whether those fees will be enough to support the exchange’s costs remains to be seen. That’s why it’s the companies named above which are most active in this space, because they all have ad networks of their own that they want to add value and liquidity to, so that they can recruit more advertisers and publishers, and ultimately take over the world (bwwahahaha).

 

Impact

Ad exchanges are poised to have a transformative effect on the online advertising business. Given the current economic climate, you can probably expect these creatures to fly under most people’s radar for the time being – probably until late 2009, I’d say – but their influence will be felt as advertisers find it easier to reach the audience they need (via greater liquidity in the marketplace), and publishers are able to hang onto a bigger chunk of the price of their inventory.

In the latter case, Exchanges could end up changing the balance of power between direct-sold and network-sold inventory – if a publisher can get a better margin (taking into account sales costs) by sending some inventory that was direct-sold to the exchange, they will do so.

But what about networks? They will likely see better margins by going through an exchange for inventory they can’t clear themselves; but exchanges will level the playing field in terms of inventory access, meaning that networks will have to ad value  over and above simple aggregation in order to survive. Competition for publishers may intensify since, if networks are backed by exchanges, there’s less incentive for publishers to have deals with multiple networks. Certainly we can look forward to lots of change – consolidation, specialization, fragmentation – in this industry in the years to come.

For more reading on this from someone who knows much more about it than I do, I’d recommend these two excellent posts on the topic from Mike Nolet, formerly of Right Media.

June 07, 2010

Online Advertising Business 101, Part VII: Demand-side Platforms (DSPs)

It turns out, alarmingly, that it’s been over a year since my last Online Advertising Business 101 post. And a year is an awfully long time in the world of online advertising. Long enough, in fact, for an entirely new kind of company to emerge and become the next big thing. I’m talking, of course, about Demand-side Platforms. You’ve heard of them, I trust? No? Then read on.

 

DSPs, RTBs, oh my

As it should happen, my last post in this series was on the subject of Ad Exchanges – a new kind of participant in the advertising value chain that acts as an intermediary between ad networks, allowing them to exchange inventory to make up for shortfalls in supply and demand among their own publishers and advertisers, and also allowing other scale players (e.g. big advertisers like eBay, or a big agency) to buy inventory dynamically across a number of different networks. All those folks had to do was implement some technology to interact with the exchange, and place bids in real-time (on of the key characteristics of an ad exchange being that it can auction inventory in real-time).

It turns out that that last glib sentence conceals an awful lot of complexity. Real-time Bidding (RTB), as it’s known, is a pretty complex technical challenge to pull off. It involves building a system that can listen for inbound impression opportunities, and parse them more or less in real-time (milliseconds) to transmit back a bid for that impression. If you just wanted to send the same bid back for every impression, or vary the bid on the basis of some very simple variables (e.g. the size of the ad unit), then you could probably hack something together reasonably easily, but of course it’s not that simple.

Once you get into the business of real-time bidding, you want to take into consideration a wealth of data (the more the better, in fact) about the impression, including (but not limited to):

• Ad unit size/format (e.g. Rich Media) 
• Site 
• Page content category 
• Geo 
• Time of day 
• Frequency 
• User profile 

Of course, if you’re an agency or network doing this, you want to be able to manage different bids and budgets for your different advertisers, and also to incorporate things like advertiser/publisher block lists (e.g. Wall Street Journal doesn’t want to advertise on the NYT site). And you probably want to enable advertisers to manage (or at least track) their campaigns through the third-party ad servers (Atlas Media Console, DFA, OAS) that they’re used to.

 

What’s a DSP?

The complexity of building something like this has spawned the new technology category of the Demand-side Platform (or DSP, which always makes me think of something else, showing my age). The name comes from the fact that these systems “aggregate demand” – i.e. they provide a single interface to the supply-side of the value chain for a number of advertisers.

Since many of the established participants in the advertising value chain already represent (or serve) advertisers, almost everyone – agencies, networks, exchanges - is these days calling themselves a DSP, much to the annoyance of Mike Nolet at AppNexus. Mike argues that the glib name, together with its appropriation by every man and his dog, make it all but useless to claim (or describe) something as a “DSP”.

It is true that the “Platform” in “Demand-side Platform” is a bit of a stretch, but I’m a little bit more forgiving than Mike, and don’t think that it’s an entirely useless term. As to what defines a DSP, I would offer the following definition, based upon this list of capabilities in the table below. If a company offers a well-integrated set of, say, three-quarters of these capabilities, and offers a real-time bidding capability, then you’re probably looking at a “proper” DSP:

DSP typical capabilities

Advertiser bid/campaign management UI or API

Ability to execute real-time buy bids with exchanges and other “RTB-enabled” counterparties

Multidimensional bid optimization

Ability for advertisers to provide their own data (e.g. targeting cookie data) to refine bidding

Universal user management (e.g. frequency capping) across multiple inventory sources

Integration with third-party ad servers

Consolidated reporting/billing/payments across multiple third-parties

Most of these “DSP capabilities” are nothing new. The items above can be found in other systems, such as those provided by some ad networks and third-party ad servers – which is why companies in these businesses are busily launching DSPs. What is new is the consolidation of these features into discrete offerings – and the new boundaries that are being drawn as a result.

 

Where to find a DSP

As well as stand-alone DSP providers (see the list below), DSP functionality is appearing in two other main places: ad networks and media agencies. 

Ad networks are incorporating DSPs into their advertiser-facing offerings so that they can broaden the range of media they can offer to their advertisers beyond that which they directly represent. The latest network to do this is your friend and mine, Google, which this week announced the acquisition of Invite Media, which brokers bidding on Google’s own DoubleClick Exchange, but also a range of other exchanges including our own AdECN.

Networks will also increasingly use DSPs themselves to source their inventory, gradually replacing their “in-bulk” supply-side relationships with DSP-mediated deals where the network gets to pick and choose which impressions to add to its pool.

The other place that is getting all DSP’d up is the media agency. Several large agency holding companies are establishing ‘trading desks’ which are consolidating the buying (aka demand) from across the company’s advertiser base and leveraging that demand to drive better-value deals for the agency. These trading desks are becoming increasingly sophisticated and have recently started engaging in real-time bidding. Some of the better-established trading desks (with the providing agency in parentheses alongside them) are in the list below:

• Vivaki Nerve Center (Publicis) 
• Omnicom Trading Desk 
• B3 (WPP) 
• AdNetic (Havas) 
• Cadreon (IPG) 
• ATOM (Razorfish) 

 

Who are the DSPs, then?

Well, that depends who you ask, of course. But now you’re this far down this post, I feel emboldened to provide you with a list of the best-known independent DSPs. If I’ve missed someone, please let me know, and I’ll add them in.

Company	Description
AdChemy	Provider of audience-based display and search advertising platform (the “AdChemy Experience Platform”). Focuses on generating a “relevant” experience for consumers, bundling in dynamic creative generation with other DSP-style services such as segment targeting & optimization. Typically partners with Agencies – largest partnership is with Accenture Interactive.
x+1	Offers campaign/media optimization (“Media+1”) as part of a suite that includes landing page & site optimization (straying into MVT territory). Media+1 enables real-time bidding for exchange-based inventory - x+1 has created a “Virtual Audience Network” across multiple Ad Exchange & Network inventory sources, to compete with traditional ad network offerings.
Media Math	“The new marketing OS”. Provides a DSP offering (“TerminalOne”) to agencies, through a combination of technology and services. One of the first companies to market with such an offering.
Turn	Offers both agency trading desk and full-fledged network management software – both include the full range of DSP “table stakes” features such as exchange integration, RTB, targeting & optimization. Network offering adds cross-network inventory/campaign optimization. Because of both offerings, Turn works with both agencies and ad networks.
DataXu	Fairly straightforward independent DSP – offers real-time bidding across a handful of exchanges, with optimization and tracking built in.
AppNexus	Styles itself as a cloud-based platform on which RTB/DSP-style systems can be built, rather than a DSP in itself. Partners with agencies and ad networks to enable them to build out their own DSP or trading desk capabilities & offerings.
Triggit	Another fairly vanilla DSP, with a fairly standard range of capabilities & partnerships. Fairly strong portfolio of data partners to aid with campaign targeting.
LucidMedia	Offers self-serve DSP (LucidMediaDSP). Claims to be able to reach 95% of the US online audience via partnerships with the usual suspects (Google, Yahoo/RightMedia, Rubicon, PubMatic, etc).
Invite Media	Now no longer independent after its acquisition by Google. “Bid Manager” tool provides cross-exchange buying; supports the major exchanges (Google, Yahoo! RM, AdECN) plus other supply aggregators such as PubMatic.

 

Where next for DSPs?

That’s an excellent question, and I’m glad you asked it. The short answer is, ah, not sure. But one thing I can aver: if you come back to this post in a year’s time, you’ll chuckle at the (by then) wildly out-of-date list above. Some of the companies in the list will no longer exist, either because they’ve gone out of business or been swallowed up by a larger fish, and there’ll be some new wunderkind on the block who I’ve completely missed out.

Which is all to say, there’s a lot of change going on in this area. The balance of power in the industry is shifting, and right now it seems like the DSPs are calling the shots, but don’t expect the big ad networks to sit around idle and let DSPs drain away the demand-side of their business. We may well find (as is happening in the web analytics industry) that very soon, there are no independent DSPs left, and every agency is a DSP also.

On the other hand, life is becoming much harder for small networks to survive in this new world; first their advertiser base will be wooed away by the DSPs (and by DSP-enabled network competitors); then their supply may fragment as DSP-based agency buying takes hold. If you thought the online ad industry was starting to settle down and make a little more sense, then, well, too bad. But at least it’s not boring.