BACK

⇪

The Trade Desk Edge

⇪

2. DATA-DRIVEN PLANNING - PROGRAMMATIC ADVERTISING

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Course 1 - DSP Basics

8 Key Components of a DSP: Now and At End State (Part 1)

⇒ ⇐

8 Key Components of a DSP: Now and At End State (Part 2)

⇒ ⇐

8 Key Components of a DSP: Now and At End State (Part 3)

⇒ ⇐

8 Key Components of a DSP: Now and At End State (Part 4)

⇒ ⇐

Koa Artificial Intelligence (AI)

⇒ ⇐

Course 2 - Data and Identity

Identity (Part 1)

⇒ ⇐

Identity (Part 2)

⇒ ⇐

Identity Update

⇒ ⇐

Relying on the Right Data

⇒ ⇐

Brand Safety and Marketplace Quality (Part 1)

⇒ ⇐

Brand Safety and Marketplace Quality (Part 2)

⇒ ⇐

Course 3 - Measurement and Goal Setting

Measurement

⇒ ⇐

Core Trading Principle - Omnichannel Mapping and Goal Setting

⇒ ⇐

Course 4 - Driving Growth

Driving Growth: Tips for the Modern Marketer (Part 1)

⇒ ⇐

Driving Growth: Tips for the Modern Marketer (Part 2)

⇒ ⇐

Driving Growth: Tips for the Modern Marketer (Part 3)

⇒ ⇐

Core Trading Principles - Developing Strategies

⇒ ⇐

1. MARKETING FOUNDATIONS - PROGRAMMATIC ADVERTISING

Course 1 - Introduction to Programmatic

Beyond the Lumnascape: The History and Players

Navigating the Digital Landscape

⇒ The way to understand the world of Programmatic and Real-Tide Bidding [RTB] is to learn the convoluted industry, evolution, history, players, and motive of Ad Tech. ⇐

⇒ Before 90s: direct sales via Print, Radio, TV | In the 90s: the digital world | In 1999: the Ad Networks | In 2004: the Ad Exchange | In 2008: RTB and Programmatic. ⇐

⇒ One of the major milestone was the creation of Outdoor Advertising, which was centered around the Circus going town to town with a need to create awareness. ⇐

⇒ In 1908, the creation of mass-produced Model T created lots of roads and room for Outdoor Ads [by the time, a nation of roads and billboards was being built.] ⇐

⇒ From targeting everyone possible, by late 50s, Volkswagen Campaign with compelling and effective Print Ads avoided competition with big car manufacturers. ⇐

⇒ Through Television Ads, McDonalds became one of the most successful Marketing and Advertising Companies by appealing to the brand to the general public. ⇐

⇒ Nike combined Marketing and Advertising by making a shoe deal with Michael Jordan thinking that their brand can be associated with what Jordan Accomplishes. ⇐

⇒ In a history of advertising, it had been extremely difficult of measuring anything whether you are successful or not. Some marketing campaign like Jordans obviously was proven successful, but most marketing campaigns were not so obviously successful to be accurately assessed of their performance. What happened in digital is that everything became targetable and everything became measureable. What had surfaced from this is that marketing was more effective than ever when directed and measured correctly. ⇐

Dot-Com Era:

⇒ Milestones of Digital: In the 90s: The Internet came out with the Dot-Com Era | In 1993: AOL pioneered ad-funded internet and popularized Cost Per Mille [CPM] | In 1995: Doubleclick started a concept of measurement of the Internet | In 1997: Ad Networks was began by the great need for connecting millions of both Advertisers and Publishers. ⇐

⇒ Definition of Liquidity: An efficiency or ease with which an asset or security can be converted into ready cash without affecting its market price. ⇐

⇒ Definition of Perishability: If somebody lands on a page, there are 100 milliseconds or less to show an audience an Ad which expires very quickly. ⇐

⇒ For a long time until early 2000s, the biggest problems in Digital Advertising space were liquidity problems. ⇐

⇒ The first liquidity problem in digital advertising space was on how to take an Ad Inventory: how to convert the people who are visiting web pages into money. ⇐

⇒ The second liquidity problem in digital advertising space existed on the buy side: how to turn the input and output of Ads into cash before the buyers are gone. ⇐

⇒ After the Dot Bomb in the 00s, because people generally get way too cautious after economic crisis, everything got really focused on performance. CPM lost popularity as people wanted to control and measure performance. ⇐

⇒ Then in 2001, for search, overture developed the new model called a CPC (cost per click) so that advertiser only have to pay when somebody clicked on the ad. Google which had no revenue stream mimicked the CPC model and the revenue model is still tremendously successful today ⇐

⇒ Besides Search (where customers comes to the ad), in the ad types where advertisers were pushing stuff out like Display Advertising Ecosystem, CPMs were still being used. Ad Networks came out in 2006 to provide liquidity of provide a link between advertisers with publishers. (Ad Network Companies create relationships directly with advertisers and publishers and connect as many of them as possible). ⇐

⇒ Since advertisers and publishers had to use many different Ad Networks due to having different options, Ad Exchanges in 2007 came out with a goal to help connect all the Ad Networks and get rid of the liquidity problem by pooling all the ad networks together. ⇐

Price Discovery:

⇒ Price Discovery: When the buyer gives sufficient information for a seller to know what the market demand looks like and it is also when the seller gives sufficient information so the buyer knows what they are buying (price discovery process involves buyers and sellers arriving at a transaction price for a specific item at a given time) ⇐

⇒ If you can create a market place where people know what they are buying and selling, then you can create a healthy marketplace. Price Discovery is the biggest value proposition of Ad Exchanges and liquidity follows from price discovery. ⇐

⇒ Ad Exchanges make advertising like the stock markets, creating the most efficient price discovery process in advertising (There is a centralized auction that makes it possible for any company to buy advertising and participate in a market that is aggregating millions of websites) -- This was an innovation because buyers changed, from buying a specific billboard on the side of the board and specific ad slots, to massive pools of inventory with the most efficient price discovery in the history of advertising. ⇐

⇒ In the initial version of Ad Exchanges, while bigger and better liquidity was achieved, there existed a big problem where there was only one optimization algorithm by Ad Exchange Company, resulting in high concentration of power and infectivity through one person making all the buying decisions ⇐

⇒ RTB (and Programmatic) Changes Everything: Now everything can be measured scale, market pricing, unique optimization. RTB is much like the stock market in the sense that New York Stock Market and Nasdaq broke out of the model, making it really difficult to envision a world where buying decisions can be made much more optimally than it is today, data can be more considered, and milliseconds can be taken off. RTB has reached the level of efficiency where it will become the way that all advertising will be monetized in the future and the level of price discovery that TTD has created inside of programmatic advertising is going to change the way that television, radio, etc is run. ⇐

100 Milliseconds: Understanding the Mechanics of Real-Time Bidding (RTB)

Cookie-Exchange Table

⇒ The internet is ad funded and therefore consumers get access to content which are usually free. When consumers land on the web page, a publisher then immediately (within 100 milliseconds) needs to respond with an ad. ⇐

⇒ When a consumer lands on a publisher's website, consumers immediately go to publisher's adserver which is often DFP: Dart For Publishers, looking for an Ad. DFP, now Google Ad Manager enables publishers to serve ads, manage inventory, and optimize ad performance. ⇐

⇒ Then, the DFP sends a request to yield optimizer which is responsible for monetizing all the inventory for publishers, figuring out how to divide monetization to right pieces. Yield optimizers analyze the available Ad Inventory considering factors such as ad formats, placements, and user segments which help to find the highest-paying ads for available inventory. ⇐

⇒ After that, the process hits the ad exchange. Inside the Ad Exchange exist the concept of cookie mapping. When the Ad request comes to DSP, the DSP does not see the user's browser. Some technologies existed to identify users such as cookies which DSP do not have access to⇐

⇒ As a solution, every DSP in Ad Exchange has to do cookie mapping. Each Ad Exchange call over to DSP and says, for this User XYZ, look out the user and match them with User ABC of DSP Database. DSP store the mapping of those two Users, so that by the time a request comes through Ad Exchange and hit the server in an RTB situation, DSP can receive user XYZ and match them with User ABC, identifying the user and looking up user data. ⇐

⇒ Add that is in a cookie is just a single ID which does not have anything to do with customer PII: Personally Identifiable Information. Company A laying down a cookie is just them saying user ABC and Company B laying down a cookie is just them saying XYZ. When a company get other data points from identified users, a company can lump those events together such as what they clicked and searched, all keyed off a single simple ID. ⇐

⇒ In summary, different ad tech platforms assign users unique identifiers stored in cookies and cookie mapping share these unique identifiers between platforms (for example DSP and DMP: Data Management Platform) to recognize same users across different context, integrating user behavior from multiple sources and create more comprehensive user profiles for more enhanced ad targeting and increased campaign effectiveness. ⇐

⇒ Cookie used to be storing everything into the cookie, making it be a pretty rich piece of data with user frequencies and any kind of data on usage. Since the advent of RTB, that has all been moved to server side. Now there is just a placeholder ID and DSP do the mapping to know who is who. DSP has to collect dozens of records for every user on the internet to first, store data about them, and second, cookie mapping them to everybody else in the ecosystem. It is a really big data problem demanding huge bank of servers, but at the same time very powerful technology. ⇐

⇒ It used to be putting and managing all frequency in the cookie: Client Side Frequency Capping. This could put the storage and processing all on the consumer in their cookie on their machine. However, this makes it impossible to do lookups in real time. If anyone is doing anything other than user identification today, such as Client Side Frequency Capping, they are not operating in any sophisticated way. One of the disadvantages of using a cookie-based storage (storing all information in the user's cookie) is that cookie has a very limited storage to only store 20 or 30 ads or impressions on there. When server-side is brought in, DSP is able to store all of the ads that has been served for that user through times, enabling to do rich things like attribution or store many data points about the user. ⇐

Federated Exchange

⇒ From a privacy standpoint, nobody in the industry is using PII. As opt outs (not wanting to be tracked) are becoming more and more persistent, using no PII is important. ⇐

⇒ Looking at the cookie table and mapping up the users takes about three milliseconds because all the difficult work of identifying the users took place before which will further be explained through cookie-mapping and placing pixel. ⇐

⇒ The Federated Exchange federates the ad request out to all the hundreds of buy side players [DSPs] representing millions of advertisers. The tough challenge of building the federated exchange was opening a connection (integration between different ad exchanges or platforms to share inventory and data to create a unified marketplace) with cost-effective way. Under Federated Exchange, request for showing an Ad is sent out many times back and forth because it attempts an auction under federalized ad exchange which is costly. While the values of Ads were supposed to worth more due to RTB as they allow auction competition for highest bidding, making the process cost-effective [solving the problem of Cost to Serve: expenses associated with delivering ads to users] was the problem [Math does not work to have the cost increase by 200 times and having the effectiveness increase by 5 times]. ⇐

⇒ Graph of bid density and the effect on yield as a publisher: There is a point where it become asymptotic, a point where more number of bidders will not guarantee a higher yield (ex: a point where having six more bidders does not get you six times higher yield and the graph flattens out). ⇐

⇒ Inside of The Trade Desk DSP, they need to have their own cookie mapping and do their own look up (ex: knowing Google ID ABC is same as TTD ID XYZ) and the lookups happen for any of the data players: for first party data which is stored in the DMPs and also for third party data which is store in Data Marketplace (data providers). This means that Google ID ABC, TTD ID XYZ, and Bluekai ID DEF can be all synched and data about user on each sources can be pulled. In order for these data to be considered in milliseconds, all of DMP and Data Provider have to give all these data to the DSP in advance of the auction. As TTD is going through thousands of advertisers they represent, they determine the advertiser (brand and agency) that is going to pay the most in the auction. Then, TTD knows that that bid of the brand is going to represent TTD DSP, return that bid to the exchange, then essentially the exchange sends the ad straight to the consumer and the Ad shows up. ⇐

Decisioning

⇒ In the exchange side, there are mechanical challenges of how to create enough connections and how to get enough services survive. On the buy side, it is processing challenge of how to look at all the data, all the targeting information, and find the advertiser. The TTD as an aggregated platform helps the buyer only the things that are best for them, the things that are most likely to drive performance of what is available while still hitting the budget.⇐

⇒ The concept is like when a car is passing by billboard ad, what if a billboard company can show different ads to different person in the car passing through? What if for each car passing through, a billboard company can make decisions whether a driver can afford Mercedes or Toyota? If it were connected properly, it would be possible. In online advertisement, the RTB allows big or small advertisers to listen to every ad opportunity on the entire internet waiting for the right consumer to show up for their product. ⇐

⇒ Data plays a huge part in what TTD decide to bid and being the DSP, TTD is really flexible on how an advertiser bring the data to the table. Is the advertiser working with Bluekai or another DMP that TTD is integrated with or can integrate with, does the advertiser have their own in-house data to be directly integrated with TTD, is the advertiser using some of pixels to do a basic retargeting pixel? TTD have a lot of different ways of bringing the advertiser data to the table. ⇐

Marketplace Weaknesses

⇒ It is not cost-effective for the exchange to send the request to everybody. The exchange are connected to everybody, but that does not mean that they get to look at every ad opportunity from them and that leads into a lot of missed opportunities. ⇐

⇒ Who is going to represent TTD in this particular auction? TTD is looking at thousands of advertisers to figure out which one is going to do that. But some of the other DSPs are not looking at their entire pool of advertisers because it is architecturally difficult for them to do. As a result, a buyer may have an audience they really want, but do not actually get the opportunity to buy all the times they want to because they were not fully included in the auction. ⇐

⇒ It can be really expensive to store all the permutations of all the different types of users that buyer like to target. If you are trying to target left-handed race car drivers from Ohio, there is not that many of them, so it is really easy for DSP to say the betting average is too low and it is too expensive for DSP to store this permutation of its buying platform, which results at the bottom of the decision tree. This disallows one of the most important goal which is to target the really small audiences. ⇐

Data Centers

⇒ The way all of this is affected by geography: For most companies inside the RTB, the thing that is important to fulfill the 100 milliseconds is whether their data centers are close to each other. The end ad server particularly needs to be as close to the consumers as possible and also everything else needs to be closer to each other. It is important for the publishers and data center to be close together to increase the efficiency, but not the consumer and data center. The exchanges usually put their data centers 6-7 places around the world and DSPs put their data centers as close to the exchange data center as possible to increase efficiency and reduce the rate of timeouts. ⇐

⇒ While a lot of people think about Amazon Cloud as the way to get into this RTB, it is really difficult to make RTB work inside the Amazon because of performance issues, so it is important to take the jump and build some more serious and robust data centers if DSP is trying to participate in high level. ⇐

Introduction to DSPs and SSPs

What are DSPs and SSPs and How are They Different?

⇒ Market Efficiency is inevitable in a marketplace where barriers to entry are low and competitions exist. The definition of market efficiency is that people get compensated in direct proportion to the value that they add (if you price too high or too low, that will be fixed by the market efficiency). ⇐

⇒ Market Pressures: The market visualizes that the advertisers on the buy-side are on the left, publishers on the sell-side are on the right, and the exchanges in the middle. What used to happen is the Ad Networks in the middle connecting the advertisers and publishers. Now, the exchanges have replaced the Ad Networks and Ad Networks essentially had to be either one of two things: something of a DSP or something of an SSP. ⇐

⇒ The publishers always want the middle to provide enough value or else they will replace them and go directly to advertisers. The publishers always try to keep things as direct as possible and try to put as much pressure to the middle, so they keep as much of advertiser's money as they possibly can. The advertisers feel the same way where publishers only get a few from spending. Therefore, the publishers and advertisers both put pressure to the people in the middle to add more value than takeaway. ⇐

⇒ The epicenter of power today: There will be more demand (the buy-side) than supply (the sell-side). It is much easier for publishers to throw another tag on their page and increase the impression of their website [consider a tag is worth 1 billion impressions, a publisher can easily add 2 more tags and earn 3 billion impressions] [this is how they historically increased their revenue] whereas it is much more difficult for advertisers like Nike to increase Ad Spending from $1 Million to $1.5 Million. As a result, the epicenter of power will always be on the buy-side as long as the market is heading towards efficiency. Also, considering market efficiency where people get compensated in direct proportion to the value, the complexity of buy-side dealing with data and dealing with targeting are giving much more pressure to the buy-side to innovate than sell-side. ⇐

⇒ The pressured middle which is Ad Exchange are putting value to help with decisioning, helping people to make better purchasing decisions, which is where all the value has been historically in such as the stock market. Whether that is on the sell-side to help them better understand what to sell, what price to sell it at, what inventory they have, or whether that is on the buy-side with big opportunities. Understanding programmatic is important because programmatic is a better way to transact all of it and the brands need help understanding what decisions need to be made. ⇐

Course 2 - Advertising Concepts

The Purchase Funnel

Purchase Funnel and Attribution

⇒ ⇐

Measurable Events

⇒ ⇐

Bid Factors: Line Items and Bid Adjustments

Understanding the Difference

⇒ ⇐

Focusing on What Works

⇒ ⇐

Defining Bid Factor

⇒ ⇐

Device ID/Cookie Mapping Explained: How It Works and Why It Matters

Digital Identities

⇒ ⇐

Mapping Across Domains

⇒ ⇐

Cookie Tables

⇒ ⇐

Pixels and Data Collection

Whiteboard Session: Actionable Data

⇒ ⇐

Whiteboard Session: Mechanics of Collecting Data

⇒ ⇐

Setting Goals and Achieving Success

Defining Campaign Goals

⇒ ⇐

Key Misconceptions

⇒ ⇐

Retargeting Misconceptions

⇒ ⇐

Crowded Media Plans

⇒ ⇐

Course 3 - Audience Activation

Recency and Frequency

Key Tactics

⇒ ⇐

Data Analysis

⇒ ⇐

Data 101: Understanding and Applying Data to Improve Performance

Data, Data Everywhere

⇒ ⇐

Understanding Location Data

Ingredients of location Data

⇒ ⇐

Precision and Scale

⇒ ⇐

Course 4 - Campaign Success

Marketplace Quality

3 Steps to Setting the Standard

⇒ ⇐

Viewability: What it is, Why it Matters, and How it Works

Embracing Viewability

⇒ ⇐

Whiteboard Session: In-View Rate

⇒ ⇐

Viewability Probability

⇒ ⇐

Demystifying Gross Rating Points (GRPs) and On-Target Percentage (OTP)

Trust

⇒ ⇐

Digital GRPs

⇒ ⇐

OTP Tips

⇒ ⇐

Gold Standard Challengers

⇒ ⇐

Cross Device: How to Get in the Game and Not Leave Effectiveness on the Table

Macro Look at Cross Device

⇒ ⇐

The Unified Journey - Targeting

⇒ ⇐

Incremental Lift - Attribution

⇒ ⇐

Private Markets: History of Publisher-Side Monetization

Mechanics of Publisher-Side Monetization

⇒ ⇐

Deep Dive of the Stack

⇒ ⇐

The Power of Decision

⇒ ⇐

Private Buying

Private Buying 1.0

⇒ ⇐

Private Buying 1.0

⇒ ⇐

Private Buying 3.0

⇒ ⇐

Data-Driven Transactions

⇒ ⇐

Spot Market Over Programmatic Guaranteed: When and How to Buy

⇒ ⇐

Page updated

Report abuse