What is a demand-side platform (DSP) and real-time bidding (RTB)?

Those days when advertisers would buy huge blocks of untargeted mobile advertising with little idea about where or when the advertisements run are numbered, or so we’re told. Today advertising technology is so efficient and so fast that in the time it takes someone to open a mobile page on a site or app, advertisers can identify the opportunity, based on visitor/publisher profiling, bid for and buy the ad spot, and serve the ad. This all happens automatically in real time. Welcome to the brave new world of the demand-side platform (DSP) and real-time bidding (RTB) exchange.

We asked Emmet Geaney, director of enterprise solutions at Byyd (formerly known as Adfonic) – the company behind the DSP platform Madison, to explain how it all works.

What is a mobile demand-side platform (DSP)?

A DSP enables advertisers to buy advertising inventory (i.e. where their ads are shown) across a wide range of publishers via online exchanges, offering access to huge numbers of impressions globally through a single buying point.

A mobile DSP, such as Byyd's Madison, performs this service in the mobile world. It automates the process of match-making mobile ads with the most appropriate mobile sites and apps, both addressing the unique challenges and maximizing the opportunities afforded by mobile. A mobile DSP brings together data from advertisers, publishers, ad tech companies and third parties, to give mobile ad buyers the best possible chance of buying the right inventory for their clients.

Rather than buying blocks of inventory, the DSP evaluates each impression, as the mobile user opens a page on a mobile site or app, and, if the slot is appropriate for a client’s campaign, it is purchased, so the ad is shown as the page displays on the mobile device. This all happens within a fraction of a second.

Why use a mobile DSP?

With a mobile DSP, advertisers can see the market rate for each and every ad impression they buy; and can judge which ads have the strongest performance, engagement or audience reach. They can adapt campaigns by controlling which exchanges to buy across, and how much they want to bid for inventory.

This is more transparent than buying though a mobile ad network. Ad networks operate in a 'black box' style, buying the inventory at the lowest rate and selling it to advertisers at the highest. Often advertisers may not know where the ads have been served or the real cost/value of the inventory they have purchased.

Mobile DSPs use proprietary algorithms to make media buying more efficient. These algorithms take all data available – contextual data from the publishers, data from advertisers, data from the DSP and data from third-parties – then use prediction modeling to assess which impressions are worth bidding for and which are not (all in real-time).

This process of using data and algorithms to match ads to audiences is called programmatic buying. It’s a good way to scale mobile advertising while maintaining brand safety.

Using a mobile DSP gives media buyers access to mobile real-time bidding (RTB) exchanges, which offer a smarter, more efficient way to buy inventory (see below), using programmatic buying.

What is mobile real-time bidding (RTB)?

RTB automates the process of buying and selling advertising inventory via an online auction.

The easiest way to explain RTB is to first look at what happens in the non-RTB world. When a publisher sells its available ad inventory via an ad network, to place mobile ads are passed around the network of ad routers until each ad slot is filled. This is good for maintaining fill rate (i.e. it ensures that more pages are shown to readers with ads), but doesn’t ensure that each slot is sold to the advertiser who would have paid the best price for it.

This can be improved using yield optimization platforms, which attempt to identify the buyers on the ad network that are most likely to offer the best price, based on historical data. The problem here is that each ad network requires a different yield optimization platform, which can make it costly for publishers.

The RTB exchange changes this model, by introducing an all-comers' auction. All available ad impressions are available to all bidders who use huge amounts of data to decide whether to bid for an impression and if so, how much. The process takes milliseconds, requiring data processing and algorithmic techniques that were not technologically possible until recently.

RTB exchanges are usually second-price auctions, which is similar to the method used on the retail auction site eBay. This means advertisers make their bid without knowing what the competitive bids are, so they don't bid against each other. Instead, they bid based on what they believe is a fair price. If the auction goes beyond that price, they're out. But if they win it, they only pay the second highest price. It's an elegant, efficient way of ensuring that real value emerges, rather than advertisers bidding against each other and distorting price levels. This process takes place in the milliseconds during which a mobile app or site page is being rendered on a mobile device.

Examples of mobile RTB exchanges include AdMob, AppNexus, DoubleClick Ad Exchange, MoPub, Nexage, PubMatic, Samsung AdHub, Smaato, Switch Concepts, The Rubicon Project and Yieldlab.

Publishers can plug directly into these exchanges, but if they want to make their inventory available across many RTB exchanges it’s more efficient to use a supply-side platform (SSP) such as AddApptr, DoubleClick, Improve Digital, mAdserve, Nexage and TapIt.

Similarly, media buyers can plug directly into the exchanges or use a DSP, if they wish to access many RTB exchanges (see examples below).

What is the relationship between a mobile DSP and mobile RTB?

A mobile DSP is the platform that advertisers use to bring together buying data, and access mobile RTB exchanges.

RTB is a programmatic buying mechanism whereby advertisers can bid in an auction for individual impressions.

Programmatic buying has been available for online media for five years or so, and is now increasingly available in the mobile space. It is a smart automated system of buying and selling media that does not require human intervention. It assesses each opportunity to buy ad impressions against set criteria and, by analyzing results, can adjust what it buys.

What is the relationship between a mobile DSP and mobile ad network?

A mobile DSP is the platform that advertisers use to bring together buying data, and access mobile RTB exchanges.

A mobile ad network is a platform for publishers to sell their inventory collectively. Where ad networks make their inventory available through mobile RTB exchanges, they can be purchased by DSPs. Thus they are complementary.

As the mobile advertising ecosystem has evolved, and RTB comes to the fore, some ad networks have reassessed their business models. Some have move more towards the advertiser side and become a DSP (as was the case with Byyd), some have move to the publisher side by becoming an SSP.

What is the difference between a mobile DSP and a desktop DSP?

While they both offer access to large amounts of inventory and give advertisers more transparency, control and efficiency, mobile DSPs need to overcome some issues that desktop DSPs do not.

The major difference is that there is no universal cookie for mobile as there is for online, so the principal means by which online DSPs track and target users is not available. To put this in perspective, the default setting for mobile devices running Apple's iOS is not to allow third-party tracking cookies. So a DSP that relies on cookies is unable to recognize a substantial section of the mobile audience.

The mobile DSP uses different methods to recognize the user through the mobile ecosystem. This means it can manage campaign delivery for an advertiser across mobile apps and mobile Web to such a level that it is possible for an advertiser, for example, to specify the number of times a potential customer sees an ad in one day.

The holy grail of digital advertising, generally, is to understand who the customer is on any digital platform – Web, desktop, smart TV and so on – so that the advertiser can have a conversation with the customer throughout the day, whatever device they are using, targeting a campaign that is relevant to the context of the consumer and therefore non-intrusive. This needs to be done in a way that ensures anonymity, i.e. without storing any data that could be used to identify them personally. It’s a problem that many mobile advertising companies are working to crack.

Being able to recognize the visitor in two environments, mobile Web and mobile apps, is a good start towards this holy grail of user recognition across multiple platforms.

How does a mobile DSP help targeting?

A mobile DSP brings together the data from various sources, which are then analyzed using smart algorithms to help the advertiser target the right customers through the most appropriate mobile media.

There are four main types of data: 1) contextual data supplied by the publisher (i.e. about the nature of the site and audience); 2) first-party data that characterizes the advertiser's desired target audience; 3) second-party data that companies such as Byyd provide based on huge amounts of processing and insight; 4) third-party data bought in from independent sources to target specific segments.

Typical criteria for ad targeting include country, city, zip/postcode, geographical radius, platform, device type and model, time of day, day of week, with positive and negative retargeting (retargeting is the practice of continuing to show ads to a user who has shown interest in your product).

Who are the main players in the mobile DSP space?

The main mobile DSPs are Byyd, Human Demand, mMedia and SiteScout.

How to choose a DSP?

Integration: the best mobile DSPs will be integrated with the greatest number of mobile ad exchanges to offer the widest and deepest access to inventory, which can run to billions of impressions per day. They will also readily integrate different mobile ad formats, particularly rich media.

Data: the effective running of the DSP/RTB ecosystem is dependent on data, so the more first-party data it can access from the advertiser, combined with third-party data bought in to complement this, the better.

Algorithms: this is the main point of differentiation for many mobile DSPs. The algorithms that a mobile DSP uses to combine all the data, analyze each impression, decide whether to bid for that impression and then decide how much to bid, lie at the heart of the mobile DSP's effectiveness. Algorithms are the 'brains' of the operation, and are critical to identifying value. The mobile DSPs with the most sophisticated algorithms, will win the most bids at the most competitive prices, and will pass on higher click-through and conversion rates, as well as efficiency savings.

How much does it cost? Does it make it cheaper?

The cost of mobile DSP services varies between vendors and license agreements. They are usually charged on a cost-plus model, so the advertiser pays a certain cost per thousand impressions (CPM) for ad space and the DSP’s fee is a percentage of that CPM.

Regarding whether it is 'cheaper', studies by Nexage and Rubicon all suggest that the RTB model, through which a DSP enables an advertiser to bid, can improve click-through rates and win rates, and reduce costs for advertisers, while also increasing effective earnings per thousand for publishers.

However, this is all part of the bigger theme around what an advertiser regards as 'value'. If, through RTB, an advertiser buys more of the right audience and less of the wrong audience then the media buying is more efficient and therefore better value. It's the opposite of 'spray and pray', where ad networks buy a lot of inventory at low rates with the hope that some ads will attract clicks.

Meanwhile, the mobile DSP gives total transparency for the advertisers. They can see exactly where their ads are appearing, and how much they paid for them. This gives them the control to adapt throughout the lifetime of a campaign and zero in on what is working well for them to maximize value, rather than advertising blind or semi-blind on an ad network.

Does it mean less control over where ads are placed?

The advertising model with the greatest control over specific site and app placement is the premium model, where the advertiser selects individual sites that are suitable for its ads and places them only on those sites.

However, premium doesn't necessarily yield the best returns as it tends to focus on buying inventory in blocks and focuses on targeting people in broad proxies such as the devices they use or the sites they visit.

Conversely, mobile DSPs enable the individual ad slot to be matched to the individual user. This ensures brand safety simply because the right people are being displayed the right ads in the right inventory. There is no assumption that the audiences are right simply because the publisher says so. The rest of the data brought to bear on the decision to bid makes it a much more accountable system.

Another model that offers control is the private marketplace. If RTB is an all-comers' auction, then a private marketplace is auction by invitation only, where a mobile publisher only allows a select group of pre-vetted advertisers to bid for its available inventory.

DSPs can add value to this private marketplace, as its knowledge about each visitor and therefore the relevance of each impression to the advertising, enables the advertiser to make a more informed choice as to whether to purchase and what to bid.

Updated: December 2016