programmatic guarenteed

How Should We Measure Media Value?

Originally published on AdExchanger

Measuring the comparative value of media inventory has been a longstanding challenge. For both sides, the relative value of media inventory is the most important measure to determine price.

At the heart of the problem is the fact that each buyer measures value differently. While that is true, the questions each buyer asks to find value are the same. Being able to answer these questions about available media inventory means that the optimal mix of targeted inventory, given current market conditions, can always be found.

If a media-buying team can answer all of these questions about all if its inventory sources, it can confidently say that it is always buying the best-performing inventory at the lowest price, given the condition of the market.

Since guaranteed deals set the price before the deal is done, the exact value of each possible deal can be compared to determine which will provide the greatest amount of value in budget.

How did the media I bought from this source perform on each metric?

In media, value is measured by the amount of performance achieved. Since performance happens along multiple measures, viewability, click-through rate and conversions, we can think of each of these as a measure in the value space.

Simply put, the best answer that a targeted inventory source can provide is that every impression is viewable, every impression results in a click and every impression converts.

The worst answer is when viewability, clicks and conversions all total zero. When everything converts, every last bit of value was captured. When nothing converts, no value was created.

How efficiently was this source’s inventory moving prospects through the funnel?

The next challenge in measuring value has to do with how steep the sides of the funnel are. In other words, how efficiently does the audience of this inventory source move from the top of the funnel to the bottom? Low viewability and a high conversion rate are just as inefficient as a high viewability and a low conversion rate. The efficiency of the funnel measures both. The wider the funnel is at the bottom, the better.

How much performance and efficiency am I getting from this inventory source at this price?

In the end, it is about effectiveness. Inventory that delivers high value may be important, but if the price is too high, it might actually be less effective than something cheaper. So when comparing different media inventory, it is the combination of value and price that drive the decision.

How much audience scale does this inventory source have?

Having fairly priced and efficient inventory is great, but there is still another piece missing: the amount that is available for sale. Being able to achieve campaign goals with the least number of sellers is important in keeping down the cost of the media buy and ensuing administrative costs.

If the media-buying team can answer all of these questions about all of its inventory sources, it can confidently say that it is always buying the best-performing inventory at the cheapest price, given the condition of the market.

Without viewability price does not measure value

While much discussion of viewability has taken place, there is still room to discuss viewability in the context of media markets. The true price at which something will sell in the market contains a very important bit of information. Pricing and market data within media are like the DNA building blocks for our understanding of the market. You need to have all the pieces to understand what is going on. True price is the most important piece of information. For buyers, not having the ability to understand the unit price of inventory, which will be viewed by real audience members matching their targeting, means there is a missing piece in the DNA that makes up that buyers’ understanding of the market. Without the knowledge of price, a significant amount of decisions cannot be made with certainty. An impression that is not known to be viewed has a price that has little information buyers and sellers can glean from. Viewability measurement is so important because without it, price cannot be used to compare the value of different media inventory. In turn, that means that a real negotiation is more difficult.

Buyers and sellers want to know that they are doing business ‘on the level’. Viewability is not about higher or lower prices, viewability is about finding the right price. For those hiding in the shadows, lack of viewability hides true quality, artificially raising effective price, and can be used as a negotiation bludgeon to artificially lower price. In the end though, all the good folk of the media market just want fairness.

In media markets, buyers know ‘a price’, the problem is that the price they know is not exactly for what they are buying. Some media buyers will read the previous statement and disagree. I argue that the only price the buyer cares about is the CPM of all the real impressions. If a buyer knows how many impressions were shown to their specified audience of real people, in a viewable manner, and the real unit price of what they’re buying, before the purchase, they can proactively select the inventory that will perform best.

Media is not a commodity, the viewability of each publisher is unique and the mix of real and bot impressions is unique. Moreover, two buyers who buy the same inventory at the same price will not achieve the same ROI. So, every publisher is different and so is every buyer.

Let’s work through an analogy. Imagine you’re a contractor building houses. You have an opportunity to build houses that you know will sell for $1000 per square foot. What should you build to maximize your profit? Well, if you consider all the materials and labor, you can mathematically figure out the most profitable size house to build. But what if you had no idea what the price of the real materials would be when you need to build the house? What if the amount of defective materials varied by store and manufacturer and you had no way to measure it? Viewability is exactly like that. Without viewability you don’t know the actual price of the product that you need to buy.

Today, the data landscape is rich with solutions that help separate the wheat from the chaff when it comes to impressions. This data powers buyers’ ability to look past the amount they paid and into the price of target audience and media.

Using that data to power media buying decisions, the ability to measure substitutability begins to emerge. Figuring out “what to buy instead” is a very important function of the buy side. Houses are not commodities, but we all know that when you have to choose a place to live you figure out how to balance the good, bad, and price. The second choice at a lower price can quickly become your first choice. Without viewability, you cannot accurately measure price, and without price you can’t make good decisions to balance the good, bad, and price.

The Programmatic Catalog

As the process of media transacting becomes increasingly streamlined, the number of buyers a publisher interacts with continues to increase. While many are hidden behind the real-time technology stack, publishers have many more counterparties buying their inventory than ever before. The increasing presence of technology and complexity of the media transaction process  has created a new need – product information management. The recent acquisition of Yieldex by AppNexus for its forecasting capabilities and programmatic direct platform is a clear indicator that the big players are beginning to recognize this need. Ad Ops and Sales teams need an electronic map of the inventory landscape. What is on the map? How much of that inventory do they have? At what price should it be sold? In which channel? Publishers need an “app” that can do all of that by automatically collecting and analyzing data. Remember Programmatic Inventory Has A 'Yellow Pages Problem'? Folks have been talking about this since 2012

With that in mind here is a simple question: do publishers manually update their audience data? Are there analysts that update cookie profiles? We all know that the answer to that is no. So why should publishers have to manually create all their placements, ad units, products, and guaranteed targeting in the ad server. Publishers should have tools to scale up that process using technology.

Today, solutions construct a single view of the future. In reality, there is not one future, there are a bunch of possible futures each of which has a different probability of becoming reality. The future is uncertain but understanding the landscape of things that will very likely happen, will probably happen, and might happen can be clearly defined.

Obviously there are too many ways to sell and too many prices for today’s manual processes to address. The problem is that you can’t just update the current process. Almost all of a publisher’s new buyers have a very specific set of buying criteria for media. None of these criteria exist on the rate card and none of them has a price someone can just look up. It’s just too complex to sell that way. In the real time environment, data companies have filled the need for near instant classification of impressions for sale and pricing is solved for via auction. What has not changed is how publishers classify forward inventory.

Fortunately, these problems have been faced by other industries and has been solved for them by some of the best technology companies in the world, Oracle and IBM to name a few. It’s called a Product Information Management System (PIMS):

Centrally managing information about products, with a focus on the data required to market and sell the products through one or more distribution channels. A central set of product data can be used to feed consistent, accurate and up-to-date information to multiple channels such as sales teams, marketplaces, and direct deals. - Modified, from Wikipedia

Some will contend that they have such a system. While that may seem true, almost all the information in their system is created, maintained, and updated manually. Again, do publishers manually update their audience data?

A product information management system for publishers should be designed to be integrated with all sales channels, provide an accurate and constantly updated catalog of all the things that you can sell, how much inventory is available for each, and its price. Synchronized systems allow publishers to increase prices as inventory is selling out or decrease it if it is going unsold. It keeps sales teams, marketplace offers, technology partners, and ad ops teams all on the same page. Publishers can present any package they could possibly sell, with an appropriate price, to any sales channel, at any time through a single synchronized platform.

We Need To Rethink Marketplace Fees If We Want Better Liquidity

Originally posted in AdExchanger We can thank the long struggle between advertisers and agencies for the fee structure used throughout today’s online advertising industry. Agencies have always wanted to pass marketplace fees on to advertisers, but advertisers only want to pay media costs.

As a result, a significant portion of the industry hides transaction costs in the media price. Transacting environments nearly always charge the publisher by deducting a fee from the buyers bid. It is unclear if any transaction business in the industry has been able to layer its fee on top of the bid.

This struggle between agencies and advertisers has hamstrung the whole industry by forcing the use of a fee structure that fails to yield the best result for anyone. I believe that moving to a make-or-take model can fix this and improve market liquidity for everyone.

In the end, everyone seeks to get the most value out of any transaction, even when that transaction is buying services from a marketplace. The most obvious strategy is to maximize our own outcomes. Given that we know we will all act this way, does every fee structure for the services of a marketplace have the same result? No. The best model is one in which everyone’s interest are aligned, including buyers, sellers and the marketplace itself.

‘Going First’

There are two types of participants in a market, and I’m not talking about buyers and sellers. We need to look at the participants of a transaction in a different way.

In every transaction, there is someone who “goes first.” More specifically, they are the first party – it could be either the buyer or seller – to state the price at which they are willing to do the deal. That information creates liquidity. The marketplaces in which lots of buyers and sellers are willing to broadcast their price need far fewer transactions to be liquid. In contrast, the markets where the price at which buyers and sellers are willing to make a deal are secret need many more buyers, sellers and transactions to be liquid.

That being the case, it doesn’t matter if you are a buyer or a seller, “going first” is clearly a benefit to everyone in the market. So, the side creating the benefit should be rewarded and the side consuming that benefit should pay. In a marketplace, a fee structure that accomplishes that objective is called a make-or-take fee structure. In other words, if a buyer puts in a buy order before there exists a sell order to match with it, the buyer gets the reward. It is also true the other way: If the seller puts in a sell order that must wait for a buy order to match, the seller gets the reward.

This way you can always choose to go second and keep your desired price secret until you see someone on the other side that wants to make a deal at your price.


This strategy makes for a business model where buyers and sellers are rewarded for participating in the creation of a crowdsourced “book” of supply and demand that all participants have access to. In stock markets, the side that is “going first” is referred to as making liquidity, while the side that is “going second” is said to be taking liquidity. Hence the name: make-or-take model.

In real terms, this means the current financial exchanges, such as the New York Stock Exchange or NASDAQ, charge market participants who “go second” about $0.30 for every 100 units, of which they pay the participant who chose to “go first” about $0.27. The difference, $0.03, is the exchange’s revenue. The implication of this model is that those who choose to “go first” in the market actually get paid to trade.

For a long time, folks in the industry have dreamed about a central repository that enables buyers and sellers to understand at what price inventory will clear and how much is available. But if buyers and sellers receive no benefit from listing their intentions in this repository – meaning there is no payment for “going first” – no one will do it.

That is why the dream of a giant catalog of available inventory has not materialized. There is no mechanism in the market to create that incentive.

For Guaranteed, 2-Sided is better than 2nd Price

Hal Varian, Chief Economist at Google once said “All of the major search engines use auctions to price ads. The reason is simple: there are millions of keywords that need to be priced and it would be impossible to set all those prices by hand.” Hal’s thesis was the underlying rationale for using the same auction model for search and display advertising. That is why today’s exchange technologies all operate using a second price auction. The problem is that you can’t use second price auctions for guaranteed. Which is why no one is doing it. There is a better way to trade guaranteed media. We can’t leave the best guaranteed inventory to sell through technologies which ecommerce figured out 20 years ago.  We need a market driven, technology powered, negotiated marketplace to buy and sell guaranteed media.

At first, it is easy to think that the focus of Hal’s statement is about labor. But if you look more deeply, you find that the key action in Hal’s quote was ‘set all those prices.’ If you can’t calculate and set prices as a seller you need the buyers to set them for you. Or in Google’s case, if ‘can’t’ is because it is too complex or expensive, you let the auction do the heavy lifting for free.

What if the problem of ‘set too many prices’ is already solved? What if there was already a way to automate it too? If that were true, the seller would want to have control over setting the ask price of their sell orders. The nature of guaranteed is that we have much more time to buy and sell it. If the seller has no bids, they can wait. If a buyer sees no inventory they can leave a bid. This type of trading works much better in two-sided auctions.

The fundamental difference is that two-sided auction transaction prices are set by both sides. Neither side has to do the deal right now. Guaranteed media is a negotiated media transaction, there is no negotiation in second price auctions. The auction decides. In a negotiated environment, each side has the power to change the price at which they are willing to do the deal. Negotiation is all about change.

Two-sided is different:

-Auctions match bids and asks, they don’t determine clearing price.

-Resting orders. Buy and sell orders stay open.

-A continuous auction where a buyer describes demand so it can be understood by any seller, to identify if they have matching inventory they may want to offer in reply, and vice versa.

- Depth of forward supply and demand liquidity can be measured. - Supply competes for demand and demand competes for supply in the same auction.

- Buyers can search for supply and Sellers can search for demand. Buyers can query sell orders and sellers can query buy orders; each can query the price and amount required to fulfill their order.

- Market data is deterministic.

For this, orders have to be kept in a ‘book’ and every market participant needs to be able to see this book. If you know what the supply and demand in the market is, you don’t need to guess what it will be. Negotiation of all those prices is possible because the problem of ‘set too many prices’ is already solved.

As an industry we need to have a new conversation about technology for guaranteed. Current guaranteed models that use the ad server as if it was designed to be the back-end for an ecommerce platform for guaranteed media buying don’t meet the needs of buyers or sellers. There is a much better way than ‘add to cart’ and ‘proceed to checkout.’

We can’t rely on auction models designed for substitutable real-time keyword searches to trade inventory that is not substitutable. We need innovation.