springload

(Originally published in iMediaConnection, December 2011) by Eric Picard

Typically, online publishers make their money through the sale of online display advertising, with a few making a lot of money from paid search. But the way that publishers monetize their sites has evolved over the last 10 years to a point where a lot of energy is expended on work that doesn’t pay off very well.

The one thing to keep in mind as we progress through this discussion is yield. From a publisher point of view, this is essentially the profit on inventory sold. It’s always important for publishers to consider yield in any discussion of revenue, because while they may sell inventory at a higher CPM through the human premium sales channel, the cost of sales is always going to be very high there. So when you strip away cost of sales and technology costs, what revenue the publisher keeps is its yield.

Many publishers began selling their remnant inventory off at wholesale prices to ad networks and an ever-changing and evolving ecosystem of other vendors who buy cheap remnant inventory, apply some “special sauce,” and resell the same inventory for a higher price. This arbitrage has evolved as many publishers saw an opportunity to liquidate their entire pool of inventory at any price and never leave any money on the table. But liquidating all inventory at “any price” is a horrible idea, and has led to many unintended consequences, namely driving a perception of “unlimited” inventory out to the market. There’s a reason that the broadcast networks limit the amount of remnant sales they do to approximately 10 percent.

I’ve stated before that most companies get addicted to “bad”‘ revenue that operates at a net loss. I would argue that almost any remnant ad sales at wholesale prices sold in bulk to resellers is “bad” revenue. While cost of sales may be much lower on this sales channel, all the value gets stripped off the inventory, and then the final clearing price of the inventory is super low — probably well under $1 CPM — probably as low as $0.30 in many cases, and often under $0.10.

Yahoo recently announced that they’re going to stop selling inventory this way, and I think this is one of the gutsiest and smartest things I’ve seen a big publisher do in a long time. Here’s why:

Publishers have gone to immense effort to build a refined product to sell. Your audience is not a “raw material.”  You’ve taken the effort of cultivating an audience to consume your content, and you’ve developed a sales force to represent this inventory. By selling wholesale to a reseller, you’ve turned that inventory into a “raw material,” and it’s up to the reseller to then refine the inventory and make it more valuable.

In a perfect world, you would sell your refined audience through direct sales channels and liquidate it all at a very high yield. Since this is simply unlikely for basic human scale factors, there needs to be secondary sales channels that don’t suffer from the same scale problems. One way that publishers have been trying to handle this problem is by applying a dedicated human sales force and yield optimization team to manage their remnant liquidation.

In my opinion they should have a dedicated team to manage selling inventory that “fell through the cracks” of their human sales force, but this is not remnant wholesale practices. They should only be selling the inventory that can be sold as a “refined good,” not a “raw material.” All publishers have high quality inventory that is not able to be sold by their sales team, even if they are “sold out” in the publisher’s sales interfaces.

Because of the way that inventory prediction works, there is generally more actual available inventory than most systems will allow to be reserved. This is because of a technical reason — since the industry has told software engineers that guarantees are being made contractually, it’s using a very high confidence interval on the prediction of avails. Confidence intervals specifically refer to how confident the engineers are that the inventory will actually exist, which is a mathematical prediction.

Since the engineers are being conservative due to the nature of the contract being guaranteed, there is actually always inventory that has sold out in the sales system due to high demand, but at delivery time there is actually more in existence. Additionally, since most publisher side sales systems allow (for very good reasons) the sales team to pull avails and reserve for a short period prior to the deal closing, hoping that their IO will be signed, this causes a lot of premium inventory to get locked up until it’s too late for another sales person to actually sell high demand inventory.

Let’s look at a hypothetical example: Samantha reserves a bunch of inventory for a December 15 start date for a big pharma client who has a proposal in front of them. On December 10, the client comes back and refuses the offer saying it will come back in January. Samantha frees the inventory up to the rest of the sales team, but with only five days left, it goes unsold (even though many other buyers would have loved to get access to it two weeks earlier.)

This high quality inventory is dropped on the butcher shop floor like some sad porterhouse that is washed off and then ground up for hamburger. Across the industry, there are tons of great inventory going to remnant sales that could be sold by the publisher as New York strip, filet, and rib eye rather than ground into hamburger. But the remnant sales channel doesn’t allow for this — and everything looks like hamburger.

Finding a way to offer this inventory to the market through a secondary sales channel and selling it with as much of a possible premium is a critical issue. The inventory should be sold not in a wholesale way with all data stripped away. It should be sold in a channel that avoids conflict and that drives highest possible yield. I’ll give some ideas on what these channels can look like below — but first, let’s discuss the most basic approach: Simply stop selling remnant.

If publishers would simply reinvest the resources they spend selling the 40 to 60 percent of inventory that is currently sold “wholesale” remnant, and put the same headcount to use (maybe different people) focused on selling premium, they only need to increase premium sales a very small percentage would completely offset all their wholesale remnant sales. When 10 percent of your revenue comes from half your inventory — there’s a problem. Better to stop selling it wholesale and bolster your average CPM, and protect your user experience. Ideally remove ad units when there isn’t a sold impression (rather than always delivering ads.) This does require some design changes, but could be well worth it. At the very least, if you can increase your sell-through on premium sales just a tiny percent, you will more than make up for all that unsold remnant.

Yahoo certainly has done this math and determined that it doesn’t want to feed the remnant piece of the pie chart above any longer. Its decision is that it’s going to push the inventory into the purple piece of pie I’ve labeled “audience-based sales.” It’s doing this through Right Media Exchange (RMX), which it owns. It’s requiring that advertisers have a seat on the exchange, and it’s creating programmatic mechanisms for those advertisers (through its partner agencies, trading desks, and DSPs) to purchase Yahoo’s premium inventory in an automated way without going through the premium sales channel. And it looks to me like it’s going to do this in a non-blind way, meaning the buyer will not be buying hamburger — it’ll buy porterhouse or New York strips.

This is a bit of a gamble, but a super smart gamble. Publishers can create “tunnels” through any of the major exchanges, where they can set agency- or advertiser-specific rules about who can buy inventory at what price, with what discounts, and in what ways. This technology has been around for a while, and some publishers have opened “private exchanges” using it. But Yahoo’s taking this to the next level, where it’s only selling its “remnant” inventory this way now. A move I applaud (whoever made this call and got it pushed through, let me know — I want to buy you a drink!)

My sincere hope is that Yahoo sets hard price floors on what it sells through this sales channel, and that it doesn’t liquidate the inventory at any cost. And when it does let retargeting companies or its other former “wholesale” customers of remnant purchase the inventory — it should ensure that it’s getting paid what the inventory (based on the audience they’ve attracted) is actually worth — and force the reseller to identify who the buyer is, and what the closing price of the inventory is. But baby steps are fine with me!

(Originally published in iMediaConnection, January 2011) by Eric Picard

Real-time bidding (RTB) is a hot topic in the online ad industry these days. (Personally I wish the industry would talk about real-time buying and real-time selling — because bidding is not really a requirement to get the benefits. But that’s perhaps the topic of another article.)

There are a lot of misunderstandings about the issues in the RTB space, particularly from publishers. Many publishers are jumping in with both feet, but a large number are still in wait-and-see mode, and have big concerns that they want addressed before they enter the fray. Some of these concerns are not areas that should cause concern, and there are other issues that they should probably care about — but that aren’t even on their radars.

Years ago, publishers were very concerned about sales channel conflict with ad networks. This concern ultimately was addressed by publishers only allowing resale of their inventory by networks if it was sold blind, meaning that the network could not identify the publisher as the inventory source. Over time, ad networks have become deeply ingrained in the industry ecosystem, and for the most part, the channel conflict issues have been addressed through this blind resale mechanism.

When RTB companies first entered the market, publishers treated them just as if they were ad networks (some of the RTB companies are ad networks, by the way), and they limited the inventory relationship to remnant inventory that was sold blind to match the way it had worked in the ad network case. Most publishers at this point are over focused on the issue of channel conflict here, because the drivers of RTB are different than many in the industry realize.

The RTB space has two faces — one that is focused on acquiring good inventory at a steep discount, and another that is focused on enabling programmatic media buying and selling. The latter focuses on removing the human negotiation process in order to increase efficiency in the purchase and sales processes. While the first group of buyers is discount focused, the second is focused on more efficient and effective media buying. And publishers should not only enable them — they should encourage them.

When publishers sell inventory using RTB, they not only reduce their cost of sales, but they also remove a huge amount of their ad operations costs. The buyer only picks up the inventory if it matches the buyer’s goals, and there is no guarantee involved. No manual hunting for placements that have enough traffic to cover the campaign goals. No fire drills to handle at the end of the month.

While publishers should be concerned about advertisers trying to get media discounts by playing the human and automated sales channels against each other, that’s easily overcome. Almost any mechanism publishers would use to expose their inventory to RTB — such as an exchange or a supply-side platform — would also enable them to set floor pricing. This means that the inventory would be protected from RTB prices undercutting their sales force.

I actually suggest that publishers expose their entire premium portfolio to RTB, and that they should not sell it blind. They should think of the RTB channel as part of their sales forces — a non-human sales team that lets the buyer achieve its goals more efficiently and reduces the costs of sales.

One issue that many publishers (rightfully) fear with RTB is data leakage. But this is an endemic problem that has been floating in the industry for years. Let me start with a more fundamental description of the issue:

Let’s say an advertiser makes a premium buy from a publisher for 50,000 impressions of a behavioral targeting segment called “auto shoppers,” starting on Jan. 1. The buy is frequency capped at a frequency of one, and the advertiser is paying a $45 CPM for the privilege. The ad is being served using third-party ad serving, and the advertiser also drops a tracking code from its demand-side platform into the creative — which they don’t to disclose to the publisher.

Now that the campaign has gone live, the advertiser is reaching an average of about 1,500 unique web browsers per day, setting a cookie on those browsers. Immediately, the advertiser pulls the trigger on a RTB campaign that looks for those same cookies on the various ad exchanges, and it pays a $1.50 CPM to reach them. To make matters worse (from the publisher’s perspective), if that publisher is exposing remnant to RTB in an exchange, the same advertiser might well deliver ads during the run of its $45 CPM campaign on that same publisher for $1.50 CPM to users it has already reached.

This is a real issue — something that happens every day. From the publishers’ perspective, these are data that they have worked hard to collect, and they believe they are their proprietary data. And they feel strongly that nobody should be able to steal those data from them.

The reality is that every ad network out there has been doing this since the beginning of the industry. If ad networks buy targeted inventory, they can cookie users quite easily, and then resell them as part of their targeted network sales.

But ultimately, there is a bigger conversation to have about the value of publisher targeting data:

• Publishers only know what a specific browser is exposed to when on that publisher’s site, which can be a very slim aperture of activity viewing.

• Most data that advertisers want access to are pretty much a commodity. How many times does an advertiser need to know what kind of car someone drives, and who has access to that data? A lot of parties know this information, and many have the right to sell it. Only data that are either fairly unique (people who follow their finances on Yahoo Finance probably aren’t going to MSN Money) or that are processed and matched with other activity data (search engine results or shopping activity) have proprietary value that a publisher should protect.

• Given the industry discussions about privacy, it’s very possible that third-party tracking of data for targeting could run into some issues over the next few years. This will set publishers up with many more unique data opportunities than they currently have.

Ultimately publishers need to grapple with the issue of RTB through the lens of their own business challenges. However, the two biggest issues that I usually hear about from publishers — channel conflict and data leakage — are ones that shouldn’t stop a publisher from participating in RTB. In fact, these issues should probably bring a new view to the conversation.

RTB ultimately has an opportunity to change the game for the better for all ecosystem participants — but only if enough valuable inventory is made available to the channel.