Simply Ad Tech

Brand Safety

With the advent of programmatic advertising, publishers were finally able to monetize their long-tail inventory. Whereas publishers were previously only able to monetize their premium inventory through direct sales, they could now monetize their entire site through a fast real-time-bidding (RTB) environment. However, this new RTB did not come without its own set of challenges.

One of the major challenges came in the form of brand safety. Brand safety was historically of little to no concern as with direct sales, the brand knew exactly what pages it would be serving on. Because programmatic advertising allows almost anyone to buy or sell ad space, brands and publishers are susceptible to bad actors such as malicious ads or sensitive content.

According to eMarketer, brand safety is the second largest concern for programmatic buying tech owners.

Types of Brand Safety Offenses

Brand Runs Ads on Sensitive/Controversial Publisher Content. This is what we normally think of for brand safety. Brands are concerned that their brand reputation will be damaged if their ads run alongside sensitive content, effectively generating “negative reach”.

Examples:

A Hertz ad appearing in an article titled “Are Liberal Pervs Sexually Obsessed with Refugees?”
A Jeep ad appearing a controversial supreme court case about sexual misconduct
An E-Trade ad appearing in an article about murder

Brand Runs Sensitive/Controversial Ads on Publisher Content. In this scenario, it is actually the publisher’s site integrity that is at risk. If reader’s see exceedingly annoying or grotesque ads, they are more likely to stop visiting the site and eventually site viewership will dwindle.

Examples:

Viagra runs an ad promoting male endurance on ESPN
An unknown brand runs an in-banner video in a 300×250 slot on Washington Post, slowing down the browser considerably for the user

Brand Runs Ads Alongside Sensitive/Controversial Ads of Another Brand. This scenario is not the first thing that comes to mind when the topic of brand safety is brought up, but it is still important nonetheless. This is essentially an extension of the previous point, with the impact of bad ads affecting other advertisers as well as publishers.

Examples:

An Ashley Madison ad appearing next to a hot wheels ad
A toenail fungus ad appearing next to a Ben & Jerry’s ad
A McDonald’s ad running alongside an ad promoting healthy eating

Improving Brand Safety

Steps already taken and planning to be taken according to an eMarketer survey of advertising buyers.

Although brand safety is a major concern in today’s day and age, brands are still buying while knowing the risks. Programmatic buying and the open exchange is still very attractive due to the vast amount of data/targeting options available as well the efficiency of the RTB environment.

Accoring to eMarketer, only 32% of brands and 34% of agencies agree that brands will reduce digital spend if brand safety, viewability, and fraud issues are not resolved.

How Much is Too Much?

Critics of brand safety have a differing opinion, stating that brand safety is often over exaggerated. They believe that brand safety concerns are driven more by screenshots of ads next to tragic stories than by data.

Brands may become so fearful of running on inappropriate publisher content that they stop running on news sites altogether. This reduces a significant amount of buying scale for brands. Ultimately, brands should be wary of the sites that they are running on, but they should also think of the reach and frequency implications when apply filters to their buys.

Sources

Distrust and caution are the parents of security.

— Benjamin Franklin

How to Create an Ad Exchange

____________________ William Chen

There are many different players in the ad tech world, with some bigger than others. Major DSPs include MediaMath, theTradeDesk, and Google DBM. Major Ad Exchanges include Google AdX, OpenX and Index Exchange. These days these companies facilitate billions in ad spend. How were these players first created and how did they become as big as they are today? This article serves to provide a very basic overview of the moving parts and pieces that must be considered when establishing an ad exchange.

Ad Exchange Creation Methods

In general, these methods are interchangeable between different ad tech layers such as DSPs, ad exchanges, and SSPs.

Managed Service. A company posing as an ad exchange can utilize its ad operations team to pull the levers using other DSPs, Ad Exchanges, or SSPs to reach the client’s desired result. This is by far the largest stretch to be an “ad exchange” as the ad operations team would be utilizing the UI of other established platforms.
API Integration. A company can act as an ad exchange by white labeling another company’s proprietary technology. This is an easy way to get the ad exchange up and running without going through the troubles of having to create your own ad exchange. The limitation of this method is that customization goes as far as the features that the white labeled company has available.
Ad Exchange Technology. Companies such as IPONWEB provide ad exchange technology that can be licensed for the creation of an ad exchange. The benefit of this route is that companies will be able to customize a wide variety of features of both the ad exchange integrations and UI layout. However, the integrations with DSPs and SSPs will have to be set up whereas white label solutions will already have the integrations in place.
In House Built. This is a truly “made from scratch” method in which a company’s engineering/ad ops team creates the ad exchange without white labeling or licensing another company’s technology. The major downside to this method is that it takes major engineering resources and time.

Integrations

This may be the most important aspect of any ad exchange. The integrations of an ad exchange determine the quality and safety of the marketplace for both the buyers and sellers. This is one of the very first questions that potential media buyers and sellers ask when evaluating an ad exchange.

In order to integrate different sources, the ad exchange can directly negotiate deals with desired supply partners and demand partners and implement a direct connection. Alternatively, the ad exchange can utilize aggregators such as Bidswitch to easily turn on/off demand/supply partners that are part of the aggregator infrastructure.

The integrations will determine:

Total Monthly Avails
- By Country
- By Device Type
- By Product (Display, Video, App, Native, etc.)
Availability of Exclusive Inventory
- When creating a private exchange, it is very likely that much of this inventory will be exclusive
Buyer Quality
- Select quality DSPs and buyers to minimize the chances of serving malicious or harmful ads
Seller Quality
- Select quality publishers to prevent bot traffic, spoofed sites, and other factors that may render the ad buy less effective.

Infrastructure

Some ad exchange infrastructure specifications to consider are:

Maximum QPS (Queries Per Second). The ad exchange can set up servers via a cloud or data centers. QPS allows those servers to listen to ad requests. For outside users, the more QPS the better as it provides more opportunity to purchase/monetize. For the ad exchange, more QPS comes at a cost of more expensive infrastructure, leaving some exchanges to selectively “listen” to ad requests to capitalize on the most effective ones.
Ad Verification/Measurement. Common measurement metrics include CTR, Viewability, VCR, etc. The ad exchange can evaluate and select an ad verification partner such as IAS, DV, or Moat to provide measurement reporting.
Fees and Pricing Options. At every level of the ad tech chain there are tech fees that are taken out. The ad exchange can decide the different fees that it will charge depending on its business strategy. These fees can include an integration fee and buying/selling fee (percentage or flat based). The buying/selling fee can be in the typical CPM format, a CPA format or another custom format. Another factor to consider is whether the fees will be paid out via client reporting, 3rd party reporting, or in-house reporting.
Security. Utilize companies such as GeoEdge or The Media Trust to block and prevent malicious ads. This is especially important for ad exchanges to sustain long-term relationships with publishers who will have these ads displaying on their pages.
Transparency/Hidden Fees. The exchange will also need to determine what data is shown to both the buyer and seller. This includes the number of bid requests, the number of bids, the clearing price, etc. Although it is widely frowned upon, some players in the ad tech industry can falsify clearing prices and pocket the difference between the false clearing price and the actual clearing price.
Targeting Options/Reporting. Determine what targeting and reporting options to provide. The targeting options can work in conjunction with the DSPs targeting layer, but this may result in reduced scale. On the reporting side, determine what reporting metrics/ granularity to provide. Also determine the data refresh rate and how far back data is stored both on the back end and in the UI.

White Label/Technology Providers

These companies provide services such as white labeling and/or licensing technology that companies can utilize to create their own ad exchange.

IPONWEB
Epom
Kritter
AppNexus
AppLift (Acquired Bidstalk)
SmartyAds
my6sense
Axonix
Bidsopt
AtomX

Sources

Whatever you pay attention to grows.

— Kevin O’Leary

Ad Stitching

____________________ William Chen

What is Ad Stitching?

Also known as “dynamic ad insertion” and “server-side ad insertion”, ad stitching combines the video creation and the video content into one. Although ad blocking immediately comes to mind as the reason why this was invented, there are many more benefits than first meets the eye.

Traditional vs. Ad Stitching Implementation

Again, the key difference between ad stitching and traditional video advertising and ad stitching is that the video content and ad are combined into one video. Notice that the content and ad are combined on the backend. Google Ad Manager explains the difference in a step by step scenario:

Traditional video advertising

The video player on the web page or app loads the SDK.
The SDK makes a request to the ad server and receives a VAST response.
The SDK parses VAST response.
The SDK retrieves inline video creatives (and companions, if needed).
The video player plays the ad.
The SDK loads impression pixels.
After ads are shown, the video player fetches and plays content from the CDN.

Steps 2-7 are repeated every time ads need to be shown.

Ad Stitching

The web page or app loads the IMA SDK.
The IMA SDK makes request to the Ad Manager DAI cloud service for content.
The cloud service returns ads and content stitched together in a single stream.
The video player plays the stream.
The IMA SDK loads impression pixels.

What are the Benefits of Ad Stitching?

Prevent AdBlock. Because the content is stitched with the ad and the incoming domain is from “publisher.com”, ad blockers are unlikely to block the ad. If the ad were blocked, the content would inherently be blocked as well. This is one way publishers are getting around the increasing usage of adblock.
Enable Dynamic Ads. Some more advanced platforms are able to allow for targeted ads according to the viewer.
Reduce Buffering/Lag. With ad stitching, there is now 1 video to render instead of two separate videos, which often may have different elements and require additional buffering. With a combined video, there are fewer steps during the actual ad serving process.
Reduce Impression Discrepancies. In some cases, impressions may be counted as the creative was sent to be served, but it was in fact blocked.

What are the Drawbacks of Ad Stitching?

Difficult to Implement. Because ad stitching has a different format than traditional ad serving, it requires additional engineer/ad ops work to bring it up and running.
Lack of Interactivity. Currently, VPAID has little to no support. Many platforms show a simple linear video with no interactivity, but some platforms have been able to add a rich media element.
No Video Fallback. The video ad attached to the content is the only eligible ad to serve.
Must be VAST Compliant and Transcoded. Video ads must follow the VAST standard communication language and be transcoded to the same format as the content before it serves.
Difficult to Buy Programmatically. Ad stitching is best done with direct and sponsorship deals that have a preset ad in mind. Programmatic RTB environments are more difficult to implement with ad stitching due to a different layout. However, these barriers are getting reduced as time goes on and adoption increases.

Sources

Optimism is the faith that leads to achievement. Nothing can be done without hope and confidence.
— Helen Keller

Bid Shading

____________________ William Chen

What is Bid Shading?

Not to be confused with bid caching, bid shading is a compromise between the first and second-price auction systems in which the winner of an auction will pay an amount between what they would have paid in the first and second-price auctions. Bid shading was initially introduced in order to reach a middle ground for buyers that did no like the first price auctions, as they would inevitably pay more under this pricing model.

First-Price Auction

The winner of the auction will pay exactly what they bid. If the winner bids $3.00, the buyer pays $3.00.

Second-Price Auction

The winner of the auction pays a penny more than the second highest bid in the auction. If the winner bids $3.00, but the second highest bid was $2.00, the winner would pay $2.01.

Bid Shading

The winner of the auction pays an amount between the first and second-price auction. If the winner bids $3.00 and the second highest bid is $2.00, the winner would pay an amount anywhere between $2.01 and $3.00.

Bid shading is currently offered by SSPs as a free service. DSPs are switching towards bid shading as a way to avoid first price auctions and minimize costs. However, vendors may start charging fees for buyers to use the bid shading service, a move that buyers are not looking forward to.

Auction Pricing Models at a Glance

According to a report by eMarketer, first-price auction adoption by vendors has spiked a staggering 37.5% in just 3 months. This spike has definitely made an impact on the push towards bid shading.

Why the Shift to First-Price Auctions?

The shift towards transparency. With a second price auction, there is room for vendors to take a cut of the winning bid. With first price auctions, buyers know that they will be charged exactly what they bid, and there is no excess change to skim off the top.
SSPs have implemented various alternative second price auction strategies, dynamic floors, and other variations of a standard second price auction. The move to a first price auction creates a more consistent buying structure for the industry.
Header bidding. More than half of the major publishers are now using header bidding. Within header bidding, the publishers are forcing the buyers to transact in effectively a first price auction environment by squeezing bids into bid buckets as small as 1 cent.

How does Bid Shading Work?

SSPs will base the final bid shading price on several factors:
- Bid rates
- Publisher/web page
- Winning bid prices
- Losing bid prices
- Placement on the page

The Publisher Perspective

Although bid shading does reduce publisher payout as compared to a first price auction, publishers still maintain more revenue as compared to a second price auction. A benefit for publishers is that they will see a more consistent revenue stream, as buyers will be more inclined to bid their true value amount of the impression with bid shading. Compare this to a first price auction where buyers may bid less than what they value the impression in hopes that they will score the impression for cheaper.

A Word of Caution for Buyers

By utilizing the bid shading technology, buyers must inherently trust their vendors. Currently there is no way to measure and make sure that the SSPs are not taking an extra cut for themselves. In the $3.00 and $2.00 bid example, a correct price with bid shading may be $2.50, but the SSP may inflate this number to $2.75 and take home the extra $0.25. In order to provide full transparency (what our industry has been trying to achieve), SSPs can consider providing the calculation methods for their final bid shading prices to buyers.

Sources

Well done is better than well said.

— Benjamin Franklin

Bid Caching

____________________ William Chen

What is Bid Caching?

Bid caching happens when a lost bid from one auction is used to fill a subsequent auction — but without the buyer knowing, and with slightly different ad targeting information. For example, if a demand-side platform bids for a specific impression on a publisher’s homepage to appear at a certain time, and it loses the bid, the exchange will roll the bid into another auction where the impression characteristics don’t quite match up. So instead of bidding on a homepage, the buyer could end up with the ad appearing on an article page. The risk: The buyer might not get what it thought it was paying for.

How are the benefits of Bid Caching?

Reducing latency. By using cached bids, a publisher’s SSP doesn’t have to listen to as many new bids that come in. This will speed up the page load time as it reduces the number of signals coming through, especially on mobile pages.
Second chance to win. A buyer that may not have another chance to win with the initial bid is given another chance to win inventory that they may not otherwise have a chance to bid for. This is especially true for latency heavy environments.

Why is it frowned upon?

Buyers are inherently not getting what they paid for. The buyers would submit a bid for with one set of targeting parameters, but they would win the bid and come out with an impression that doesn’t quite match up to the initial set of targeting criteria. Yes, there are inherent advantages of allowing a bid that wouldn’t have had a another opportunity to win to be given a second chance, but the buyer should definitely know about this. With the constant aim of the industry to move towards a transparent model on all parts of the ad tech ecosystem, implementing this without notifying the buyer seems to be a move in the opposite direction. Ultimately, bid caching has advantages for both the buyer and publisher, but there should definitely be transparency to the buyer.

The Index Exchange Situation

Index Exchange has been utilizing bid caching technology under the radar for over a year, on around 50% of their impressions. This has enabled them to deliver higher CPMs to publishers by generating overall more bids from buyers, but the buyer complaint is that they are not getting what they originally paid for. The Index exchange response to the community can be seen below:

“In the context of our exchange, we call the caching we leverage bid caching — a feature built and designed in accordance to industry guidelines (OpenRTB 2.5, section 7.2). Bid caching is used in latency constrained environments to allow a buyer to win a bid they otherwise would have little chance to participate in. The most common examples are video (at the mid-roll, and post-roll points as consumers will not tolerate wait time for ad decisioning), responsive design (increasingly ads load as a user navigates a page, something that should never “block” the content that follows), mobile environments leveraging prefetched content, or swiftly traversed slide galleries. We dynamically identify and optimize the exchange in these circumstances to reduce latency wherever possible. This ensures the best user experience, while also guaranteeing that buyers have maximum opportunities available to increase win rate and purchase impressions across the audiences they highly covet.

The exchange intelligently activates this feature where it is determined to be necessary or of value. Bid caching is therefore not used on every bid and is only used dynamically up to a unit of time measured in seconds. We believe this achieves the best balance between user experience (faster page speeds, lower latency) and buyer experience (recent intent, higher win rates, honoring frequency caps). We’ve optimized caching to ensure buyers are bidding on the same inventory, for the same users, on a very tight time window to maximize performance for their advertisers.”

Caching the Mona Lisa

Below is an amusing comic from AdExchanger that brings up an analogy to Bid Caching. Imagine you are an art collector bidding on the legendary Mona Lisa painting. You bid an exorbitant amount for the art piece, but you lose the auction to a higher bidder. A second replica Mona Lisa is now up for sale, and the auctioneer automatically rolls up your first bid for the real deal to this inferior replica painting. You obviously win the auction as your bid was intended for the much more expensive authentic Mona Lisa, and you are left in hysterics. This is how some digital media buyers feel when their bids for premium inventory get reused for inferior inventory.

Sources

Of all the things I’ve done, the most vital is coordinating those who work with me and aiming their efforts at a certain goal.
— Walt Disney