Running an Ad Network (The Mechanics of Arbitrage)

Someone once asked Warren Buffett what the secret to stock-picking was. "Easy-- just buy the undervalued stocks," he said. The same can be said of affiliate marketing, which also about buying low and selling high. Abritrage is about being the middleman between the buyer and seller-- maximizing that difference and finding ways to do it again and again. Understanding that basic concept has been the key to my success so far, whether it be selling diamonds on ebay (what I did a few years ago) to running enterprise PPC campaigns for Fortune 500 clients to doing small business lead generation. This post will go into the basics of understanding arbitrage, focusing mainly on a view from the inside on what it's like managing a network. I hope it is both simple to understand and also provides some value to you. If you apply this and are able to make some money because of it, you can say thanks by linking to my blog. THREE VIEWS (Advertiser, Publisher, Network) Most people in affiliate marketing only understand one half of the equation-- as an advertiser or a publisher. Thus, they pay money for ads or they have a website and want to get paid. For example, there is AdWords (buying ads) and AdSense (showing ads to get paid)-- two sides of the same coin.

• The advertiser want to get as many leads/sales, given a performance target, typically CPA. Assuming lead quality is the same, they'll apply the same CPA to all traffic sources.
• The publisher wants to maximize total earnings on his inventory. So if he's rational (and in my first semester econ class, assuming rationality is a HUGE leap of faith), he's looking at maximizing eCPM (we'll talk about that in a second)--- not clicks, CPC or CTR.
• The network (defined as anyone who sits in the middle and takes a cut) is balancing between both advertisers and publishers--- delivering enough ROI to keep advertisers, high enough eCPM to keep publishers, and having a decent margin left over.

There are 3 methods of paying for ads: by impressions, clicks, or conversions:

• Impressions: They're paid for based on CPM (cost per thousand-- M is "mille"-- or latin for thousand). I would have priced it per hundred, but that's the industry standard. Anyway, CPM buys are great for the publisher--- they are paid the same for their inventory regardless of performance. All the risk is on the advertiser, who has to deal with potentially low CTR and bad lead quality. The publisher can claim that they shouldn't be penalized if the advertiser has bad creatives or a crappy product. But it's still a lopsided deal in favor of the publisher, who can just reserve the crappy inventory at the advertiser, since they get paid no matter what. If you're a smart advertiser, then you already know how that traffic will perform and are willing to take the risk. And you won't allow the publisher to do tricks like show 10 ads on the same page, which will give him 10 times the earnings, all else equal. So if an advertiser is paying a $10 CPM, they are paying a thousand pennies to show the ad a thousand times--- thus, it's a penny per impression, regardless of whether the user clicked.
• Leads: Here, the risk is flipped-- it's all on the publisher now. The site owner (or whoever owns the traffic) is paid only on a conversion. Maybe it's $4 for a US dating lead to true.com, a $70 payout for a successful credit application, or 6% of sales at an online retail store. Whatever the conversion, the advertiser has no risk-- they have complete control over their ROI, but not the volume. In fact, most advertisers and affiliate networks employ "scrubbing", which is a systematic way of not counting conversions. There is a legitimate scrub, such as when leads are incomplete or have invalid social security numbers. The dirty scrubbing, which is done by pretty much all affiliate networks, is to not always fire the conversion pixel. Maybe they scrub one out of every 10 leads at first and then get greedy and increase it to 1 in 3 leads. When I first got started in affiliate marketing, I didn't understand the many ways that scrubbing can occur-- even still I don't. That's probably a topic for another post.
• Clicks: Paying a CPC balances risk between advertisers and publishers. The advertiser has to pay only upon a click, so they are incented to write good ads and target the right inventory. The publisher doesn't have to worry about a scrub or the offer not performing, but they have to deal with low CTR risk. By the way, if you're a PPC marketer and are looking for PPC affiliate tracking software (link to tracking202.com), I highly recommend Wes, Larby, Steve, and Rob-- they are a class act, and their software incorporates many of the concepts discussed here.

MAXIMIZING NETWORK PROFITS Ok, now that the basics are out of the way, let's discuss something that's rarely talked about-- how networks manage between supply (publishers) and demand (advertisers). There is something I call a two-part profit that the affiliate network (insert your favorite affiliate network here) has to manage against. The network is collecting a certain amount of revenue for the traffic they sold and they pay out something less to the publishers. What they pay to the publisher plus the amount they keep as a gross margin adds up to the gross revenue. Thus, if the inventory actually earns a $1 eCPM, they might actually pay out 60 cents and keep 40 cents for themselves. The amount the network takes is usually a matter of secrecy. Most networks set a particular cut that they take from a publisher, tiered based on volume (you do more leads, they take less of a cut). In the case of someone like a Commision Junction, the network takes about 25% off the top-- a fee that an advertiser gladly pays to not have to deal with managing and paying a bunch of affiliates. So usually the network sets a particular percentage cut for pubs and sometimes sets a flat eCPM. AdSense, for comparison, may take 20-30% of your earnings, which can go down to single digits if you're on a premium tier-- if you're big enough, you get to negotiate this number. So as an ad network, you want to be as big as you can, but also have the largest cut you can. If you drop your cut, then publishers are making more money and, all else equal, your network will grow. The simple equation to describe this is: Volume x Margin = Profit. It's the shape of a demand curve-- so for those of you who are aspiring affiliates and haven't taken economics, go pick up Economics in One Lesson or Freakonomics. I'm reading the Wealth of Nations by Adam Smith for my econ class right now and it's teaching me concepts that I already learned doing affiliate marketing. But anyway, there is a trade-off between volume and margin. You find the right balance where the increase in your cut is only partially offset by how the percentage of much your network volume declines by. For math people out there, it's where the slope of the elasticity curve is equal to minus one. I've run a private network for a couple years now and just this month have opened it to the public. I know other network owners and can tell you what a shady affiliate network will do upon launch. They'll start with high payouts, because their cut is low to perhaps even negative (if they're VC-funded and can weather a loss). Once they have sufficient volume, they can start to increase their cut, but using the scrub to balance out the perceived effects. For example, they could raise the payout per lead from $4 to $6, but have a 50% scrub. In other words, that $6 payout is really only $3 were there not a 50% scrub. So while the payout looks like it went up, really, the network increased their earnings by a buck a lead--- as payout dropped from $4 to $3. Ever been to the mall, where some department store has everything 50% off? You know they just doubled the price and said everything was half off. BEING A SMART PUBLISHER Are you with me so far? Let's talk about some other ways that you can be street-smart as a publisher. And by the way, the reason I'm telling you this is that I believe my social ad network (link to leaderclicks) will be the eventual winner in a truly competitive market. And if you're not a Facebook application developer, then you can still follow along here and hopefully gain some insight that will help you with your traffic optimization. For example, if you use AdWords, Google will show the ads that make them the most money-- based on eCPM, not based necessarily on what makes you more money. So back to the topic of what behavior to watch for in a network:

• Priority or boost factors: When an ad network determines delivery priority strictly based on eCPM (which ads are earning the most revenue per thousand impressions), then you have a fair competition. The ads with the highest eCPM get shown first and on down the line until the crappy ads are shown. In this case, "crappy" is defined as ads that don't monetize well. In the context of social network traffic, branded ads are typically bottom of the barrel-- the CTR's are low and the revenue per click is also bad. So if you're an ad network trying to masquerade as a promoter of big brands, you have to give the brand ads a multiplier--- maybe give it a 4x or 5x boost, else those ads would never get shown. Your branded advertisers are happy, but this comes at the expense of the publisher.
• Frequency capping: If you see the ad network running the same ads again and again--- then so are your users. If you aren't interested in what the ad has to offer the first few times it was shown, it's not likely that you'll be interested the 184th time around that day. The Power Law (or 80/20 rule) says that 20% of your users consumer 80% of your impressions. Thus, there are a small group of users who comprise the bulk of your impressions. If you frequency cap at 4, for instance, then that user who has 200 impressions per day (quite common in dating sites and social networks, where some folks "live" all day), then you'd be able to show 50 different offers, instead of just one offer 200 times. The net effect of smart frequency capping can be effectively doubling or tripling your inventory, because those heavy users are the bulk of your inventory. So where you see the same ads over and over, you know your ad network is wasting your inventory. Google AdSense recently implemented frequency capping,in fact--- curious to hear of anyone's early experiences.
• Extravagant marketing: Are you impressed by the size of their booth at AdTech New York (yes, I think I'm going to get a booth this year- so come see me). Or maybe one of the executives is bragging about how lavishly they are living, how fancy their offices are, or how many employees they have around the world? Remember that it's your earnings that are paying for this. And if they are VC-funded, then remember that the firms bankrolling the company are expecting a pretty hefty return on their investment, too. If you are fortunate enough to be near super-affiliate status and have an ad network executive take you out on a nice trip or buy you some expensive toys--- you should ask for a better payout instead. That $2,500 plasma TV might actually be costing you $10k. I can't tell you how many affiliate marketers act more like rappers than shrewed businessmen. Take that higher payout and buy the fancy watch or whatever item yourself. But some folks are more concerned about the perception of making money than actually making money--- yeah, it makes no sense. But you can probably supply your own example of friends who are like that, eating steak dinners on maxed out credit cards.

ECPM IS KING We talked earlier about how economics is based upon people making rational decisions. And in this case--- advertisers, publishers, and networks "should" be boiling things down to eCPM. But not always. And where they don't, you have a great opportunity to jump in and profit off that ignorance. For example, consider a publisher who has a fascination with what he's earning per click. Say he's earning 6 cents per click and getting a 1% CTR--- that's a 60 cent eCPM (6 cents x 1% x 1000). So you tell him you have an offer that earns 8 cents per click-- and knowing his preference to manage by CPC, he goes for it. Perhaps that offer is 8 cents CPC but gets only a 1/2% CTR-- for an eCPM of only 40 cents (8 cents x 1/2% x 1000). He's happy, since he's getting 2 cents more per click. You're happy, since now you're paying 40 cents per thousand impressions versus 60 cents. How about the publisher who has tunnel vision about the CPA? They think that the dating site offer that pays $6 is better than the one that pays $5. But how well does it convert? As a publisher, you should be ranking what offers you run based on eCPM (earnings per thousand impressions). Something that pays out 10 times more per lead (like certain financial offers) might have 1/20th the conversion rate. So you still want to look at how many cents you're getting per thousand impressions. Anything other metric, and there is a distortion where others can take advantage of your ignorance. What about the advertiser? A smart advertiser will calculate, by traffic source, the cost of their traffic (a cost eCPM) and the revenue ( a revenue eCPM based on a CPA), with the difference being the margin. So perhaps that Google content network traffic is earning you a $2 eCPM (because that's what you calculate for your traffic being sent to Webfetti toolbars). And maybe you're buying it at 80 cents eCPM (though the method you paid for it was based on CPC). In this case, you made a profit of $1.20 per thousand impressions. Now you could look at intermediate metrics (eCPC, conversion rate, landing page bounce rates), but the final word is eCPM. And what if you're an ad network? You should be agnostic to how advertisers pay or how publishers are paid. Some advertisers want to pay based on CPM, others have a fascination for CPC or CPA. You don't care-- since you back it out to eCPM anyway. Same is true for publishers--- they want to be paid based on whatever metric preference? Fine-- don't try to convince them that eCPM is what matters (which, ironically, is what I'm trying to do here). Let them have their way and profit on their misunderstanding. Believe me, I've tried this many a time--- people either get tangled up in the math or they just stubbornly stick to that one metric they like for whatever reason. I've been to Vegas a few times now-- and though I can't gamble legally (in the United States) for another 5 years, I have to believe that casinos and the lottery are just a tax on people who are bad at math. Oh, and for the math crowd, here's how the formula breaks down: eCPM= earnings per thousand impressions = CTR x eCPC x 1000 = (impressions/clicks) x eCPC x 1000 = Iimpressions/click) x (CPA/clicks per conversion) x 1000 The eCPC is how much you are earning per click. So if you're promoting Netflix DVD's by mail and get paid $20 per signup and it take you 40 clicks on average to generate a signup--- then your eCPC is 50 cents. In other words, 50 cents x 40 clicks = $20 for one signup. CONCLUSION If you've suffered this far with me, then my hat's off to you-- yes, I did get a new hat -- do you like it? If you've been doing affiliate marketing for a long time or are lucky to be an uber affiliate then this is probably tediously boring stuff. I hope you've enjoyed it. And I appreciate any feedback you have for me--- although please keep it constructive and make sure it adds value for the readers here. I've received a number of article requests for my affiliate blog, so I will keep writing so long as people are finding it worthwhile. If I made an error here, my apologies-- just let me know.