It’s not unfair to say that programmatic advertising is a revolution in the world of marketing. No longer do we live in the Mad Men world of mass campaigns flung wildly around a market—nowadays, marketers buy audiences and access to individuals, not ad spaces, and the question has moved from “what do we advertise?” to “how do we deliver the right message, in the right context, at the right moment?"
Unfortunately, confusion is as much a feature as it is a bug of ad tech. To help, Digital Media Jobs is here to explain. Let’s take a deep dive into the programmatic world with Programmatic Advertising Explained In Plain English.
Programmatic means Automation: Programmatic advertising has eliminated the manual process of connecting inventory suppliers, data providers and media buyers to streamline the media buying process through automation.
While humans are still involved, gone are the days of insertion orders, faxes, spec sheets, and all manner of other bureaucracy. Programmatic simplifies the relationship between advertisers and publishers and makes it easier to get ads in front of the people most suited for them.
One of the main benefits of programmatic advertising is just how targeted advertisements can get with machine learning. By tracking past results faster than any human can, machine learning can redistribute ad budget in real time and calculate the best bids. This helps increase the chance that an ad backed by machine learning will win the auction and be shown to a user.
A particular ad for pizza could be served only on sunny days to users in offices within a 5-minute drive of a Papa Pizza’s restaurant. Did they miss the advertisement on their work computer? Machine learning can identify their phone and serve a follow-up ad as they head out to pick up lunch.
Let’s start with a few basic explanations!
The Ad Server is where all the images, videos, files, and pop-ups live. So, when a website needs to populate an ad space, it connects to the ad server and retrieves the appropriate content.
Ad servers also do the legwork of tracking impressions, monitoring performance, and integrating with brand safety or fraud detection software. Ad servers work cross-platform, delivering ads to desktops, tablets, smartphones, billboards, and televisions. They can be remote (hosted by a third party as a service) or local.
Ad servers also indicate the actual targeting for a particular ad. When an ad request comes in from the marketplace, the ad server matches the request to the associated ad. It’s extremely important to make sure this software is set up correctly, or the rest of your work as an advertiser will be wasted.
Publishers and advertisers frequently have an ad server on each side of the equation. This has a number of benefits for tracking and keeping ad information updated, but can also present billing challenges. For example, Papa Pizza wants to ensure that all the publishers they advertise with have the newest set of ads. Without an ad server, they’d need to go out and update each publisher manually. With a server, they change it once and the ad updates automatically.
The billing challenge occurs when the delays inherent to the Internet appear. Two ad servers mean two systems tracking impressions. If there’s any difference in the clicks tracked between the separate systems, the publisher and advertiser may have different numbers to base their billing on, causing an ad-serving discrepancy.
There are many reasons this can occur (different ways of tracking, different platforms using separate analytics software, even geographic differences) but it’s important to maintain contact with the publishers you advertise with to help manage these discrepancies – this is where the old fashioned “pick up the phone and call” very much comes into play.
Ad Networks are the companies that buy ad impressions and sell them to advertisers. They act as the wholesale broker of the programmatic world. Advertisers don’t have time to look through every single online ad space – ad networks do this work for them. Most networks focus on segmentation by traditional demographics (age, gender, marital status, location, online habits) and current device. Others focus on price or content type (for example, some agencies may specialize in mobile, while others may choose to provide an adult-oriented service) – every network is different. After repackaging the variety of impressions they’ve purchased from multiple sites, ad networks sell impressions to advertisers for a profit.
Confusingly, ad networks perform much of the same functions as ad exchanges. There are a few important differences.
Ad Networks vs Ad Exchanges
These are the wranglers putting in ad orders and tracking that the ads are filled and go out properly. They are often part of an online advertising agency, but they may also work in-house at a larger brand. They do everything: checking campaign performance, enacting optimizations, analyzing data and results, and even working with clients in some cases. They usually work across a variety of networks, exchanges, and other online ad providers. Their job roles can vary significantly based on experience and agency – to learn more, we’ve written a full report here.
Data Management Platforms sound scary, but at their core, they’re essentially a massive database. They store all kinds of data, from tracking and audience information to results and reports. DMPs are helpful in pulling out audience segments and more effectively advertising to those segments. Some platforms help marketers tie together the data from multiple platforms – normally an annoyingly complex task. Many DMPs also integrate with Demand-Side Platforms (explained below!), automating some of the ad optimization processes. Companies of all sizes can use DMP software to improve their advertising results. Many DMPs are charged on a per-second or per-data-processed basis, so it’s important to consider the pricing method that best fits your business. For more information, take a look at Google’s Cloud Pricing Calculator.
Publishers use DMPs for optimization. For example, if the New York Times knows what sort of readership is most likely to read a certain article, they can charge more for ads when those particular people visit the article. DMPs help them identify those customers.
It’s important to note that DMPs only work well when the data they store is complete. Be strategic about the data you collect and input into a DMP. Focus on data you and your direct partners have collected, as it tends to be the most useful. Third-party data should fill in the final gaps in your understanding of your customers, not as a strategy in itself.
With the data from a DMP, the Demand-Side Platform gets to work. This is what drives the actual buying of the online ad marketplace. These platforms automate the ad buying process. Advertisers no longer need to fax orders and negotiate with salespeople – it all happens in milliseconds, millions of times a day. DSPs make it easy to track and manage online ad purchases.
These platforms generally sell ads through a process known as Real Time Bidding. When a user arrives at a webpage connected to the programmatic world, an instant auction takes place. The DSP acts as one of the auction bidders – deciding whether to place a bid, and if so, how much of a bid to place.
The goal is to get the best audience possible for the lowest price. Interestingly, DSPs have also affected the traditional ad agency model. DSPs are designed to be used directly by advertisers. Agencies provide guidance and strategic support to operate in more of a consultancy role.
Supply-side platforms are essentially the inverse of demand-side platforms. Using similar technology, SSPs find the highest price for a potential ad space. By opening their space to far more advertisers than usual, a wider range of bids is considered and publishers can maximize their online ad revenue. Publishers can set many automated rules to help. They can introduce pricing promotions, set price floors to ensure they don’t sell spaces for too low of a price, automate service for ads sold by their human sales team, and much more. SSPs can also act as a bouncer, only allowing in bids from advertisers the publication has chosen to accept.
Line Items are how DSP and SSP software match their ads. When a user visits a page, the publisher’s ad server sends a request to their SSP. It matches the data the ad server collected about the page visitor to see if any line items in its’ inventory match. DSPs bid on these individual line items.
Some software requires each variable to come in as a separate line item, which can quickly result in complex tables and thousands of potential items. This can cause traders to spend more time moving budgets and changing maximum bids, rather than optimizing them. A recent solution to this problem comes in the form of programmatic bid factors.
Bid Factors automatically adjust the bid the DSP is authorized to make based on how important an individual characteristic is to an advertiser. For example, an advertiser could decide to increase the bid for Women 18-34, but not for men. In the same bid, they can also increase or decrease the amount they’re willing to pay by time of day, device type, or any of hundreds of other factors.
Here’s where digital marketing and the traditional sales process collide. RTB comes into play (usually) at the moment someone gets to a webpage where programmatic ad space lives. In the milliseconds between clicking the link to the page and the page itself actually loading, DSPs and SSPs navigate through an ad exchange and compare bids. The open marketplace accepts bids from everyone, and as long as the price floor is met, the highest bid wins.
Many bids occur in what’s called a “second-price” auction. This process helps the winning advertiser save some money by only paying 1 cent over the next highest bid. Publishers stand to lose here, and some are reluctant to accept this. A new feature where the price floats to a point between the two bids is growing in popularity on the publisher side. This could be an interesting tug-of-war as the programmatic world continues to grow.
As many of us know, computers are fast but there is a limit. Users will navigate away from a page if it takes too long to load. Waterfall bidding prioritizes a publisher’s preferred ad networks and exchanges to ensure bids from that network are considered first. If no ads match the publisher’s criteria, or no DSPs send in a bid, the system continues to the next preferred network. This helps reduce the strain of having to process every single possible bid before loading the page, but could result in lost revenue. Here’s a quick example:
The Berry Vine newspaper’s website has some ad space to sell. They’ve set their price floor at $3, so no bids under that price will even be considered. A user navigates to the page, and these bids come in on their preferred network:
The SSP software selects the $4 bid and displays the ad from that advertiser. No other networks are considered since the first one met the criteria set by the SSP. However, there could have been a bid higher than $4 on the next network. Waterfall bidding means the publisher misses out on that opportunity.
Header bidding solves this problem by conducting a more unified auction. It allows publishers to receive bids from multiple ad exchanges at once. It’s a tradeoff that opens more revenue potential for the publisher, but processing all those bids in the time it takes a page to load can cause longer page load times. This type of bidding is also difficult to implement and can be expensive to roll out. It’s not ideal, but companies need to make a strategic business decision regarding which of these two processes is more important to them.
Google has recently attempted to solve this problem with Exchange Bidding. This process works similarly to header bidding but happens on Google’s servers instead of via a browser request. This solves the delays inherent in header bidding but results in other drawbacks, such as being unable to match an ad to a user (known as match rate).
Programmatic PMPs (private marketplaces) are invite-only. Publishers with a premium reputation sell their ads only to approved advertisers. Ads are still served programmatically, but in this case, the buyer knows exactly what website their ads will appear on. Ad exchanges are not involved, and the DSP connects directly to the publisher.
Another version of this is the closed auction. Real-time bids still come in, but only from a list of approved advertisers.
TL;DR: Private club. Tip the bouncer.
All these names basically mean the same thing – direct purchase of ad space outside of the bidding system, but still served programmatically. Guaranteed/direct sale ads can be sold in the traditional offline method, through ad exchanges, all kinds of different ways. It combines the guaranteed locked-in clicks of the “old methods” with the hyper-targeted precision of modern online advertising.
For extra confusion, preferred deals exist. In this case, a publisher locks in a price with an advertiser. They get offered ad space first but are not required to buy the spot. If they decline, the space then enters the RTB system.
For even more confusion, the Interactive Advertising Bureau calls this “automated guaranteed”.
TL;DR: Deal happens outside of RTB but they’re still served programmatically.
Deal ID is one of the most important terms in the programmatic world. Deal ID is a unique string of characters that serves as an identifier and is submitted along with a bid. Anything can be tied to it: custom data, level of transparency, priority, minimum pricing, access to ad units, and access to specific sections of the site; even specific agency and client approval that would normally be unavailable can be tied to Deal ID. Some have called deal ID the “programmatic IO” or “modern insertion order”, and it remains an interesting quirk of the programmatic process.
Deal ID combines the best of traditional advertising with the flexibility of programmatic advertising. This process gives power to publishers, but also a bit more risk – since they’re tying in more tightly with an advertiser, it’s possible their trust may be abused and their terms may not be fully respected.
It’s definitely important to know who you’re working with in the programmatic world. Some providers specialize in just one segment, while others provide a full range of programmatic services.
Programmatic TV isn’t all there yet, but it’s coming. The technology hasn’t been fully embraced, and some TV networks are likely to resist it – when you’ve got something that works, why change? However, the little networks with ad space to spare definitely have reason to take a look at this technology.
You may have experienced programmatic TV advertising at home – Hulu uses programmatic technology to fill ad space, and it’s likely even more digital TV and connected TV networks will make the switch. You’ll also hear a similar term called addressable TV advertising – this just means targeted TV advertising within the same program (So if you’re watching the X Games, you may not see the same advertisements as your neighbor). Learn more here!
Ad viewability will be a central concern here – advertisers will want to measure how many people can both see and hear their ads.
You’ll see this referenced in a variety of ways: Digital OOH, Digital Out of Home, DOOH. They all mean the same thing: Internet-connected video billboards. This is already starting to become a reality.
Digital Out of Home ads can already be served programmatically and update to take advantage of the latest geo-targeted ad data. In this case, advertisers are currently buying “expected audiences” rather than individuals. For example, they can buy ads based on who drives on a particular part of the 405 freeway at certain times of the day. Not many players have quite yet entered this space since most billboards are owned by a handful of companies.
Now that you’ve learned more about the programmatic ad process, why not look into programmatic careers? You can find these kinds of jobs and more on our Jobs page!