The advertising technology industry can be confusing - especially if you're new to it and being tasked with building an ad server.
This overview provides insight into the space. If you have additional questions, feel free to send us a support note or add it via 'Suggest Edits'.
If there's a specific ad tech word you're confused about, this doc should be helpful.
For Adzerk-related terms, please visit our Glossary.
Ad tech is a catch-all word for 'advertising technology'. It refers to using technology to facilitate the buying and selling of digital advertisements ('ads').
Fair question. There are a two main reasons:
Brand Awareness - These folks care about reaching the right audience with the right message. Targeting, relevance, and high-quality ads are therefore important to them. While they will likely track clicks and conversions, they usually base success on being able to deliver a set number of impressions over a certain time period.
Direct Response - Also known as "performance marketing", these advertisers care less about impressions and more about results - such as cost-per-click or cost-per-action (a downstream conversion). To cater to them, many ad platforms - such as Google's Adwords - offer pricing models where advertisers pay for clicks or conversions, not for impressions.
While no advertiser is exactly alike, they often have similar goals. After all, they are giving you money and expect something in return. Metrics they most care about are:
Number of clicks on the ad. These clicks can go anywhere - a lead generation form, a website, a product page, an eBook, etc
Number of post-click actions (such as making a purchase, filling out a form, etc)
When an ad is displayed on the site/app
Technical term for number of impressions where at least half of the ad was shown for at least 2 seconds
How many times a video ad was viewed. Sometimes broken down by how far the user got into it, like 25%, 50%, 75%, and 100% completed.
Cost-per-action (CPA) or cost-per-conversion
Total cost (ad spend) divided by actions/conversions
Cost per click (CPC)
Total cost (ad spend) divided by clicks
Total cost (ad spend) divided by thousand impressions. 1M impressions at $1K spend would be a CPM of $1.00
Clicks divided by impressions. AKA, the rate that people click on the ad after seeing it
Conversion Rate (CR)
Conversions divided by clicks. AKA, the rate that people undergo a desired action after clicking
Advertisers generally buy in one of three ways:
- Through ad networks/demand side partners, who provide turnkey access to lots of inventory (places to show ads). While the client may appear on hundreds of different sites, they only interface with one partner, rather than having to organize hundreds of direct deals.
- Through direct deals negotiated directly with publisher, who provide special placements, custom ad units, additional targeting, etc.
- Through self-serve portals, such as those offered by Facebook, Adwords, and others.
In #1 and #2, usually Insertion Orders are involved, which are contracts that stipulate metrics such as length of the campaign (or "flight"), amount to be paid, how many clicks/impressions the advertiser gets, etc.
In #3, there's usually an option to input your Credit Card and you pay-as-you-go.
Larger brands will also usually buy through ad agencies, who will be the ones who decide where and how to buy. Large media agencies include names like Starcom, Carat, OMD, Mediacom, and many more.
The most common pricing methods are CPM, CPC, and CPA (see above).
From a publisher standpoint, CPM is the safest route and most common way that direct deals are sold. As you hopefully know how many impressions your site/app gets, you can be confident on being able to deliver a certain impression number over a certain amount of time.
Cost-per-click (CPC) is riskier for publishers, since it introduces an unknown factor: click-through-rates. If you show an advertiser's ads and nobody clicks on them, you make nothing. Many large ad platforms such as Google's Adwords employ this, though, because it appeals to long-tail performance-focused advertisers.
Cost-per-action (CPA) is less common, but loved by direct response advertisers. Here, advertisers pay only for some conversion event, such as a purchase or app download. This is even riskier for publishers, since not only is there CTR concerns, but you have to think about conversion-rates too. Even if 100% of people click, if 0% of them convert, you make nothing.
With native ads - it really is up to you! However, there are certain terms (native and non-native) that you may hear from advertisers more often, which you may or may not want to build into your platform.
For a breakdown of standard ad size units as defined by the IAB (Interactive Advertising Bureau), see here.
Ads from the same advertiser that appear in tandem. Often refers to display ads that appear alongside a video ad, such as on YouTube
Catch-all term for ads that change depending on who is viewing it. For instance, showing rain gear to people who are currently living somewhere where it is raining
An ad that takes over the entire screen (most relevant on mobile)
An ad that appears within the flow of content, such as Facebook's mobile in-feed ad unit
Native ads - often on social networking sites - that look and feel like organic posts. Think Twitter, Pinterest, Facebook, etc
The rectangular 'around the web' content ads you often see on media/news sites
Ads that automatically change size based on screen/browser size
Where you show the same user the same ad (or ads from the same advertiser) for their entire browsing session
Articles on news/media sites that are sponsored by a brand. Done correctly, these are not self-promotional (like H&R Block sponsoring an Onion article on taxes). However, there are also 'advertorials' - ads that promote a product while trying to look like objective editorial pieces
Used by eCommerce, marketplaces, and any company that shows "listings". Here, companies/people pay for their organic listing to be artificially high up in the search results. For instance, Amazon lets vendors pay to have their product be the first listing for a given search word
Sponsored [Insert Your Product]
A custom native ad unit sponsored by a brand. Can be anything. Tinder has "sponsored profiles"; AllRecipes as "sponsored ingredients"; it's really up to your imagination!
To counter concerns that some ads are displayed but never seen, some advertisers may ask for 'viewable ads'. The IAB considers a viewable ad to have 50%+ of its pixels visible for at least 2 seconds
Targeting is important to ensure ad relevancy and drive better performance metrics. There are many ways that ad platforms incorporate targeting. With Adzerk, you can implement all of the ones listed below.
Target by past behavior
If you know a user likes to comment on food-related products, you can target them with food-related ads
Target by page content
Show ads that are related to page content - such as showing Honda ads in a forum related to car repair
Target people across devices
Show an ad to someone when they are on their desktop computer, then the same ad when they are on mobile
Target based on user-provided information around age, gender, household income, etc
If you are a social networking site and your users have proffered their job titles, you could let advertisers target CEOs vs Directors vs etc
Limit ads based on how many times it's already been shown to a user
Prevent ad blindness by showing the same ad only once per day to a given user
Target by current location
Target just Durham, North Carolina with ads for a local restaurant
Target by interests & values
If a user has expressed an interest in health, you could show them an ad for Advil
Target people currently in the market for a specific product
Home Depot could show ads on Zillow, knowing that those users are in-the-market for home items
Show an ad related to what the user just searched for
A recipe site could show ads for Kraft cheese only when someone searches for "Nacho Recipes". This is also what Google Adwords does
Targeting users who are similar to a specific set of users
Facebook lets you upload a list of e-mails and then will target people who are similar to those users, using a proprietary modeling algorithm
Placement / Ad Unit
Target based on ad type and where the ad is
Require ads on the homepage to be video ads, but other pages can have static ads
Target someone who has already interacted with your site/app
Show a 10%-off internal coupon to someone who has placed something in their cart
Target a user only during certain days or hours
Show ads for Applebees only on Friday and Saturday
While we both know you aren't lying about the impressions and clicks you drove for your advertiser, THEY don't know it. So, they may ask to add special parameters to the click URL, or an impression pixel to the ad. This lets impartial third parties validate whether that impression or click did occur.
Adzerk offers instructions on how to add these.
Ad-Ops is the team that is setting up and managing campaigns. They often speak directly with the advertiser to determine when to start/pause, what to target, etc. Ad-Ops can be very different based on what kind of publisher or company you are, how you sell your ads, and how big your team is. Some have multi-member ad-ops teams within their company, while others outsource ad operations, or have a sales team that also handles ad operations.
RTB (or real-time-bidding) refers to the automated buying and selling of ad inventory in real time.
In OpenRTB (what most people refer to with 'RTB'), a publisher sends an ad impression to a marketplace (as the ad is loading) and hundreds, even thousands, of advertisers bid for it auction-style. The company with the highest bid wins and appears on the page. All this happens in ~200ms.
In Private Marketplaces, there's the same flow above, but the advertisers are invite-only (therefore, generally only a handful of advertisers are bidding).
In Programmatic Direct, there's no auction, just guaranteed impressions for a given advertiser, but the buying/selling is done through a RTB platform.
Major players in this space include Rubicon, Index Exchange, Mopub (Twitter), Sharethrough, AppNexus, Google, PubMatic, Nexage, and OpenX.