Simply Ad Tech

Duopoloy

Duopoloy

noun

noun: duopoly; plural noun: duopolies

a situation in which two suppliers dominate the market for a commodity or service.

In the world of digital media and ad tech, the duopoly refers to the two advertising behemoths Google and Facebook. Because they are “walled gardens” who have a vast library of user data, Google and Facebook have a strong competitive advantage over other companies who don’t have access to such a large compilation of users at their disposal. Other companies are making strides to take back parts of the ever increasing pie, but it is a slow process.

According to a report from eMarketer, the total market share of Facebook and Google will drop from 57.9% to 55.6% during the period of 2016-2020. Most of this dropoff can be attributed towards Google as Facebook has actually increased its market share since 2016.

According to eMarketer, here are the current and predicted ad revenues of notable players in the ad tech industry. Note that from 2016 to 2019 total ad revenue is predicted to jump from $71.6B to $105.44B, a 47.3% increase.

Why are Google and Facebook so Dominant?

Easy to use. Using Facebook and Google advertising platforms bears very little barriers to entry, and there is no required minimum spend. Whereas most advertising platforms cater to large corporations, Facebook and Google provide solutions for all of the small businesses out there.
High CPMs. Facebook CPMs can run as much as 8x as much traditional cpms on other advertising platforms using the same type of ad. Because of Facebook’s unique ad tech capabilities, many buyers are more than willing to pay.
Walled Gardens. Facebook and Google both have significant amounts restricted data that cannot be accessed through any other advertising solution provider.
- These walled gardens can force users to use their entire ad stack if they want to utilize their data and services.
- They provide limited report information so as to stop data leakage. Additionally, it is hard to challenge performance when these companies are “grading their own homework”.
- Everytime that advertisers place tracking devices, such as the Facebook Pixel and the Google Analytics Tag on their websites, they are providing even more information and data to the duopoly. This creates a snowball effect that makes them even bigger.

What are other Companies doing about the Duopoly?

Facebook and Google are predicted to drop a few percentage points in total market share over the next few years. That isn’t to say that their digital ad revenues are decreasing – in fact they are actually increasing. However, the digital advertising pie is becoming significantly larger year after year, and other well known companies are growing their ad revenues faster YOY than the duopoly as a percentage. The result is that other companies are slowly increasing their market share and equalizing the playing field.

Amazon

Amazon is slated to grow faster than other other company in the industry. One JPMorgan report estimated that Amazon’s ad business would grow from $2.8B in 2017 to $6.6B in 2019. That’s a staggering 236% increase!

Snapchat

Snapchat has been increasing market share at a fast pace. Currently, around 90% of ads sold on the Snapchat platform are done in an automated manner.

Microsoft

Microsoft is on track to make over $4B in advertising revenue this year. This growth in ad revenue can be attributed to its acquisition of professional networking site LinkedIn as well as the search engine Bing.

Vox Media

Vox Media has created a digital advertising marketplace called Concert, where many publishers can join forces to fight back against the duopoly.

Sources

I don’t mind rude people. I want people that I can make money with, so if their executional abilities are good, and they’re arrogant and rude, I don’t care.
— Kevin O’Leary

Pricing Models

____________________ William Chen

After all is said and done, the buyers still have to pay the publishers for having the privilege of serving ads on their sites. This brings up the point of payment models, frequency of payment, and time alloted for payment after each sale.

Within the world of ad tech, there are many pricing models in which a transaction can be completed. Buyers and sellers use these models depending on which ones are most effective for their strategies. There are distinct advantages and disadvantages of each pricing model.

Flat Fee

Flat fees are one time fees that are paid to the publisher by the buyer as according to a contract such as an IO. As long as the performance fits within the contract terms, the buyer is responsible only for the flat fee (no hidden fees)

CPM

Cost Per Mille. This is the price that the advertiser has to pay the seller for every 1,000 impressions served on behalf of the advertiser. This is by far the most commonly used pricing model in ad tech today.

Note: CPM is calculated on a per 1,000 basis, but the other models are calculated on a 1:1 basis.

CPM = Revenue/Impressions *1,000

CPC

Cost Per Click. Advertisers use this metric to determine whether or not their ads are performing well. A lower CPC means that fewer marketing dollars were spent to generate a click. The lower the CPC, the the happier the advertiser will be. However, it is important to be wary of bots and other bad actors that may create fraudulent clicks. A promising CPC could actually be the result of a click farm.

CPC = Revenue/# of Clicks

CPL

Cost Per Lead. This is the cost of generating a “lead” for the advertiser. What the lead entails is determined by the advertiser, but it generally means an opportunity creation, such as an email subscription.

CPL = Revenue/ # of Leads

CPA

Cost Per Action. This is what brands ultimately want their buyers to do. It can mean purchasing a certain product or subscribing to a certain service. Because CPA is based off of actual results, it is generally the most expensive of the pricing models.

CPA = Revenue/ # of Actions

Transfer of Risk

As we move along the different pricing models, we not only see the prices get more and more expensive, but we also see a gradual transfer of risk to the publisher. Although the publisher is getting more revenue for each successful click, lead, or action, the publisher could be left delivering lots of impressions for free if it is unable to generate the given metric.

CPM > CPC > CPL > CPA

Least Expensive > Most Expensive

Least Risk for Pub > Most Risk for Pub

The Painfully Slow Ad Tech Payment Times

Ever since the economic crash of 2008, there is an unwritten rule in the ad tech industry that the payment times have been slowed. These days, it is standard to see net-120 payment terms.

This slow payment is becoming more and more of an issue. In essence, the buyer receives a distinct advantage because it is able to generate a float off the 120 days that it can use the “borrowed” money.

Sellers, on the other hand, suffer because they may not receive the capital they need to stay afloat in the given time. Sellers are paying an actual price in interest because of this.

This issue stems because there needs to be payments done in every step of the ad tech process. SSPs need to pay Publishers before they get paid themselves. Advertisers need to ensure that the buys met their targeting needs through ad verification partners before they make a payment to the DSP.

Blockchain has been a point of discussion for payments at the point of service but it is far from being complete. For now, ad tech companies will have to deal with the slow payment speeds.

Sources

Price is what you pay. Value is what you get.
— Warren Buffett

Native

____________________ William Chen

The best ads are the ones that you don’t even know are there. With native advertising (also known as sponsored content), the creatives are designed and placed in a way that they mimic the actual publisher content.

There are 2 main components to classifying a creative to be a native ad.

Form

The ad must match the visual design of the surrounding content. It must look and feel like actual publisher content.

Function

The ad must behave just like the publisher content. This means user experiences such as clicking and interactivity options must be the same.

Examples of Native Ads

A classic example of a native search ad is the Google search ads that have been around since the dawn of time (or at least it seems like that way). The Google ad fits the form category because the Asphalt Green Ad is visually formatted just like the Wikipedia basketball link. The function of the ad is also the same as the same as the Wikipedia basketball link because both links click the their respective pages as indicated in the green URLs.

Other Examples Include

Amazon sponsored content
In feed ads on Facebook or other sites
Facebook messenger ads
Recommendation wiges
Paid search ads
Sponsored social media posts

How are Native Ads Placed?

Here is a great infographic on how Native ads are served in a RTB environment, courtesy of Sharethrough.

Why Native Advertising?

Consumers are becoming more and more aware of ads and are trained to avoid them like the plague. For ads that consumers do see, they generally see them as just a cash grab and treat them with an ounce of sault.

Because native content looks just like the publisher content, it will be likely that consumers don’t even realize that they are looking at an ad. This can lead to several important effects:

The consumer is more like to see the ad and also interact with it.
For consumers that like a certain publisher, they are more likely to develop trust with the brand as it appears to be part of the publisher’s actual content.

Native Ad Platforms

Sharethrough
Triplelift
Outbrain
Taboola
Google AdSense
Facebook Audience Network

Sources

Passion is one great force that unleashes creativity, because if you’re passionate about something, then you’re more willing to take risks.
— Yo-Yo Ma

Header Bidding

____________________ William Chen

What is Header Bidding?

Header bidding is an advanced publisher ad stack setup method that arose as an alternative to Google’s traditional “waterfall” method of ad serving. It allows publishers to call a multitude of partners on the header of the page, before the rest of the page even loads. The big benefit of header bidding is that it allows publishers to simultaneously auction off their inventory instead and move closer towards a true auction-based environment.

How is Header Bidding Implemented?

On the engineering side, the prebid.js code must be implemented onto the publisher page by the engineering team. The exact details of this goes beyond the scope of this article, but engineers can refer to this link to get started:

http://prebid.org/dev-docs/getting-started.html

Decide on price bucket granularity. You can set up price buckets with up to a penny in granularity. More granular setups take more time but they provide a more exact bid value from the bidders.

One set of line items for all bidders or a set of line items for each bidder

One set of line items for all bidders. This is a lot easier to maintain as there are only 3 keys to look after.

Default Key	Scope	Description	Example
`hb_pb`	Required	The price bucket. Used by the line item to target.	`2.10`
`hb_adid`	Required	The ad Id. Used by the ad server creative to render ad.	`234234`
`hb_bidder`	Required	The bidder code. Used for logging and reporting to learn which bidder has higher fill rate/CPM.	`rubicon`

One set of line items for each bidder. This setup requires keeping track of more default keys but it allows you to conduct line item reporting, see a bidder landscape, and to set up more keyword targeting within the ad server.

Default Key	Scope	Description	Example
`hb_pb_BIDDERCODE`	Required	The price bucket. Used by the line item to target. Case sensitive and truncated to 20 chars. The `BIDDERCODE` is documented at Bidder Params.	`hb_pb_rubicon` = `2.10`
`hb_adid_BIDDERCODE`	Required	The ad Id. Used by the ad server creative to render ad. Case sensitive and truncated to 20 chars. The `BIDDERCODE` is documented at Bidder Params.	`hb_adid_longBidderNa` = `234234`
`hb_size_BIDDERCODE`	Optional	This is not required for adops. Case sensitive and truncated to 20 chars.	`hb_size_appnexus` = `300x250`

Confirm that setup matches with your dev team. You want to especially make sure that the price buckets match up.

How does Header Bidding Work?

To understand how header bidding works, we must first understand the traditional waterfall format.

Traditional Ad Server Waterfall

Direct Orders > RTB Auction > Fallback Inventory

In a traditional waterfall, the line items with the highest priority would get first look at the inventory. These partners get priority because they pay a higher cpm for the inventory. If a partner declines the inventory, it would move down to the next stage in the waterfall. As we move the waterfall, we inevitably receive lower cpms from the partners down the chain.

Header Bidding

Header Auction > Direct Orders > RTB Auction > Fallback Inventory

A header bidding auction is conducted on the header of the page among the demand partners that have been placed there, before the rest of the page even loads. Now we have the winning bid from the header bidding auction compete with the partners from the traditional waterfall.

What are the Benefits of Header Bidding?

Publishers have increased control over who bids on their inventory and the bidding structure.
Publishers increase competition with more of a true auction. This leads to a higher winning bid and overall more revenue for publishers.
Reliance on a single SSP decreases, and publishers are able to increase fill rate with the increased bids.
Discrepancies will be reduced as there will be fewer steps in the waterfall. With fewer steps, there are fewer opportunities for drop offs, etc.
Advertisers can benefit from a higher chance to win inventory. Many times they may be at the bottom of the waterfall and not see the impression.
Header bidding increases transparency by allowing advertisers to see each impression opportunity that goes out, although some publishers may dislike this.

What are the Drawbacks of Header Bidding?

Header bidding can be both time consuming and complicated to set up.
Header bidding may increase latency with auctions happening on the header of every page.
Infrastructure costs are increased for SSPs and DSPs as they are now able to see the same impression twice.
Publishers enjoy a first price auction between all partners, but they are limited to the second price auction mechanics by all the bids. For example, if one DSP passes a $10.01 value with a highest bid of $20, another DSP that passes a $12.01 value with a highest bid of $15 would win. In this situation, the DSP that had the $20 bid should have won, but this is not the case.

Server Side Header Bidding

Server side header bidding is a new alternative to the traditional waterfall method that aims to correct some of the issues of traditional header bidding. This means that the publisher will only call one SSP on the header which will integrate with all of the other SSPs on the server, outside of the user browser. Benefits of server side header bidding include:

Eliminates the fragmented win scenario where the highest bid may not always win.
Reduces page latency.

Google Exchange Bidding Dynamic Allocation (EBDA)

As a response to the ever increasing adoption of header bidding and to provider their own server side header bidding solution, Google has rolled out their EBDA product.

Sources

We’ve always believed in fair trade.
— Rajeev Suri

Cross-Device Targeting

____________________ William Chen

What is cross-device targeting?

Cross device targeting is the practice of linking a user’s devices together so that advertisers know that they are targeting the same person through different devices.

As device switching becomes more and more common, advertisers are forced to be able to both match and target a user’s multiple devices in order to drive home a sale. Imagine a user who sees an ad for furniture on his computer and adds a couch to his cart, but does not check out. He leaves the house and does not have access to his computer for the rest of the day. As a marketer, I would be extremely interested in being able to match that user’s computer to his mobile device so that I would still be able to target him throughout the day. If I can do this, I will be able to show him a retargeting ad which brings him back to his cart. This can be the difference between failing to convert or making a sale.

A user can be matched to her laptop, tablet, mobile and desktop, and have all these devices linked to each other.

What are the benefits of cross-device targeting?

Advertisers can utilize sequential targeting to first make a customer aware of a product or service and eventually drive a conversion. Because of the omni-channel state of the current ad industry, sequential ads are more easier to implement if advertisers can link together a user’s multiple devices.
Advertisers can implement multiple touchpoints and A/B test for which one is the most effective for ad spend. Another form of this is to
Advertisers can appeal to the same person in different ways in accordance to what device he is browsing from. For example, a user may be targeted with a standard display ad that takes them to a product page on desktop, but they may be driven to the app on a mobile device.

How are user devices matched to one another?

Deterministic Data

This is any purposeful action where the user declares information about themselves. Using deterministic data, we can be almost 100% sure that the user who has logged in to one device is the owner of another device. This is a lot more accurate than probabilistic data but scale is not as high.

Login information. We can match a user’s desktop to his tablet and smartphone if he is logged into Facebook on all 3 devices. The same holds true for any other types of logins, such as emails, Youtube accounts, etc.
Surveys and forms. Users can fill out forms and give marketers their information

Probabilistic Data

Using probabilistic data, we can infer from user behavior with a likely certainty that the owner of one device is the owner of another. This can be applied to a much broader range of devices that deterministic data but is not as accurate.

Location (IP addresses)
Device IDs
Conversion type
Date
Landing page
Interests and web browsing history

Putting it all together

Cross device companies feed all of these different data points into their algorithms to create their cross-device graphs. From these cross-device graphs, they can paint a pretty good picture of who the consumers are and what kind of ad stimulus that they would be most affected by.

Note that even deterministic data is anonymized. This means that although we won’t know the email of the person or the person’s name, we will know that we are targeting the exact same person on her desktop and her mobile if we were somehow able to match her two devices deterministically.

Sources

Imagination is not only the uniquely human capacity to envision that which is not, and therefore the fount of all invention and innovation. In its arguably most transformative and revelatory capacity, it is the power to that enables us to empathize with humans whose experiences we have never shared.
— J.K. Rowling