Programmatic buying is the automated purchase of targeted ads in an auction format for specific users based on their interests and needs. There are a number of services and platforms for this purpose.
a platform for publishers to allow advertising networks and sites to sell advertising space.
an advertising exchange. Places displays for sale to the visitors of the advertising networks and publishers' websites, and is responsible for accepting bids from DSPs and announcing the winner.
is an advertising banner network. It contains banners and manages their displays with the help of the Ad Server, which delivers advertisements to the publisher's site, counts the number of displays, clicks and manages campaign optimization.
companies that buy data from suppliers and use it in advertising. This article will mainly focus on the work of these platforms.
a tool for the automated purchase of online advertising, interacting directly with the SSP (Supply Side Platform), ad networks (Ad Network), ad exchanges (Ad Exchange) and websites (in the terminology of online advertising they are called publishers).
The DSP works with a number of advertisers, each of whom runs several campaigns (sometimes their number reaches thousands). Each campaign has a separate strategy for the purchase of ad impressions, consisting of a set of targeting. Let's take a look at the most common ones.
DSP bids only if the request from the SSP came from a particular site. This method is usually used by advertisers who want their ad to be in a certain context.
DSP bids if the request comes from a user located in a certain city or even in a certain zip code (relevant in countries like the U.S. or U.K.). Location is determined by the DSP platform by the IP address that comes with the SSP request.
Frequency capping refers to the ability to avoid showing the same banner to the user too often. The point of this restriction is that if, for example, after 20 displays the user performed no target action (a click, an order, or something else), it's senseless to spend money on buying impressions for this user.
In addition to limiting impressions for all time, DSPs usually allow you to set a limit for a minute, an hour, a day, and a week. Because it is not very reasonable to "bombard" the user with all 20 shows in the first minute.
This approach allows you to show ads to users who have visited the advertiser's site, but left without committing the target action (eg, buying, filling out forms, or call). As statistics show, such users can still be persuaded by reminding them of their existence.
RTB gave an impetus to the development of retargeting. Being connected to a huge number of SSPs and having access to billions of ad impressions, DSP has the ability to "reach" almost any visitor to the advertiser's site.
It is worth mentioning one of the extensions of retargeting - dynamic retargeting (DCO - dynamic creative optimization). In this case, the user is shown a personalized banner based on his history of visits to the advertiser's site. For example, a user who has viewed ten pairs of shoes will be shown a banner with prices.
This type of targeting allows you to show a relevant message to users who are interested in a certain topic (e.g., cars or sports), as well as people from a certain demographic (e.g., men aged 25-40).
Typically, the DSP buys DMP (Data Management Platform) data about users. The DMP regularly (for example, once a day) uploads a database of users to the DSP, which is used to process the auction request and make decisions about what advertising to offer this user.
A more advanced option is also possible when the DSP learns information from the DMP in real time, that is, it sends a request to the DMP after each request from the SSP.
All of the above targeting only defined the logic for selecting a user to show an ad. However, they do not affect the definition of bid size.
Working with the DSP, which has only the targeting described above, involves manual work - a marketer or traffic manager must himself select the rate and change the targeting depending on the results of the advertising campaign (the effective price per click or conversion). But in post-industrial society, labor is expensive, and there is a high probability of error due to human factors.
For this purpose, some DSPs implement predictive optimization algorithms (sometimes this algorithm is called a "predictor" or "predictor"). Let's consider their work on an example:
Let's imagine that a bank marketing manager has decided to launch an advertising campaign for a new product. To do this, he selected several dozen sites with a secured audience (targeting by domains), and chose major cities where the bank branches are represented (geographic targeting).
Usually predictive optimization algorithms are universal and can also predict the likelihood of another target action, such as a purchase (conversion).