Programmatic UA (also referred to as programmatic media buying or programmatic advertising) is a technology-based method used to run user acquisition campaigns. The use of the word ‘programmatic’ refers to the automation of the process, i.e, running campaigns automatically according to a set of predefined rules, or algorithms. 

The word ‘programmatic’ is a general term used to describe the automated process, unlike more specific terms such as ML-based UA. We will stand on the difference between the two as we’ll dive deeper into programmatic UA. 

Depending on the type of automation and the scope of the data, in programmatic UA, advertisers can define different audience metrics (from basic demographics to user behavior patterns) and use their technological capabilities to reach this audience and improve their UA campaigns, reaching their app growth KPIs. 

The automation of the process, that programmatic media buying allows, enables advertisers to run and manage campaigns on a scale that humans simply can not, at least not effectively. From the extent of the audience to the user-level targeting, through the speed in which impressions are bought, programmatic opened up possibilities that never existed before. 

Ad networks were a central component for advertisers before programmatic advertising became prominent. These companies aggregated ad space supply from publishers, classified it into categories, and resold it to advertisers. This process provided advertisers with somewhat of a targeted audience, and publishers with monetization opportunities. 

In order to prevent advertisers and publishers from cutting the middleman (i.e, the network) the process was anonymized – advertisers didn’t know where exactly their ads were shown.

With the growth and advancement of programmatic UA, ad networks have become obsolete and are gradually being replaced by companies with a strong programmatic infrastructure. 

RTB is a media-buying ecosystem that can only be used for programmatic campaigns (i.e, RTB cannot be used in non-programmatic campaigns, but it can be used with all types of programmatic campaigns). 

Real-time bidding is an online auction marketplace for buying and selling ad impressions in real-time. Such auction occurs in mere milliseconds and is completed before an ad is displayed in an app. 

RTB programmatic buying is beneficial for both publishers and advertisers. Marketers achieve major efficiencies by showing their ads to the right audiences, and decrease wasted impressions, while publishers enhance the value of their ad space and improve direct sales strategy and pricing.

RTB accounts for 90% of all programmatic buying (which is why it’s so vastly covered and frequently discussed). The other 10% is populated by alternative methods, most famously of them is the Private Market Place. 

In the RTB ecosystem, there are a few key players: publishers, advertisers, SSPs, and DSPs. 

• Publishers: Publishers are the ones who are offering ad space in order to monetize their app.    
• SSP (supply-side platform): An SSP is the one providing the platform in which publishers can sell their ad space.  
• DSP (demand-side platform): A DSP is the one managing the side of media-buying, connecting available ad space (on the publisher side) to the campaign (on the advertiser side). 
• Advertiser: Advertisers are the ones looking for available ad space that would serve their campaign. 

The process starts with the publishers, the ones offering up ad placements in their mobile app. This ad space is sent as a bid request in the RTB ecosystem, through the SSP which facilitates the auction. The different DSPs bid on that placement in the auction process (which we explain in detail in the Basics of RTB), and the winning bid gets to display their ad (i.e, the advertiser’s ad). 

Programmatic media buying enables advertisers to use AI (artificial intelligence) technologies, such as machine-learning algorithms, that add the ability not only to predict but also to adapt to changes in the data (by recognizing trends), and change their campaign’s settings and targeting accordingly. 

While programmatic media buying can be as simple as a set of predefined rules, ML algorithms differ by offering more dynamic and complex targeting capabilities. 

The algorithm’s ability to adapt to changes in the data is best explained by example. We’ll use two possible campaigns for an RPG mobile game to demonstrate. 

The first campaign is a programmatic UA campaign for the RPG game. The campaign is set to target who you know to be your audience: Males, 20+, who have shown interest in other RPG mobile games.

This campaign will probably do well for a while, since it is the target audience, but will plateau over time since it only considers the very generic parameters of the bidding decision-making process – like the placement, device, OS Version, ISP, time of day, and a few others.

One repeating scenario is when a dominant placement showing good performance loses its popularity and its user count starts dwindling –  the performance of the campaign takes a huge dip, as experienced by many running campaigns through self-service platforms.

The second campaign is an ML-based campaign. Initially, this campaign will be set up to target the same demographic as the programmatic campaign. The campaign will run for an exploration period where the algorithm will learn the data and recognize trends. 

After an exploration period, the algorithm will start targeting according to what it had learned. For example, targeting two groups: users who’d spent at least 3 hours in any RPG game, and users who play more on the weekends in general. The algorithm will then test out different options (testing who is more inclined to install – the 3-hours gamers or the weekend gamers) and change its targeting based on the results. As the campaign continues to run and the data changes, the algorithm will adapt according to users’ behaviors. 

The success of the campaign does not rely solely on the algorithms, but on the features that were set up for the campaign (someone needs to “tell” the algorithms that looking at an RPG session duration, or testing out the day-of-week usage is crucial) as we’ve previously covered in our Data Activation in Mobile UA entry. 

While programmatic media-buying, without the use of ML algorithms, can result in successful mobile UA campaigns, a well-set and efficiently managed ML-based campaign is more of a long-term solution. With the ability to adapt to changes in the data, a UA campaign can continue to scale and yield results in the long run.