• by James Green , Op-Ed Contributor, March 31, 2014

RTB is growing even faster that we expected. eMarketer predicts that U.S. advertisers will spend $8.69 billion on RTB ads by 2017. RTB’s growth isn’t just from display advertising -- it’s now rapidly expanding into mobile and video, and some even predict that it will eventually move to TV.

There are many ins and outs of buying media over RTB, which can be confusing for brands that are just beginning to enter the marketplace. The variables in pricing, management systems and types of campaign deliveries can be daunting. Here’s what we’ve learned along the road to successfully delivered ad impressions:

Valuing a Bid

There is a level of sophistication that brands must take into account when it comes to bid management, and I’m not convinced it is as prevalent in the market as many may think. It’s pretty obvious that flat-bid pricing exists, since many bids are won around “round numbers” like $1 or $1.50. If the market was truly using really variable bid pricing, it would be highly unlikely for those bid numbers to remain as static as they do. Although publishers setting (round number) floors on their pricing is one reason that prices don’t fall further.

Sophisticated players value every bid differently as they let the data inform/determine the market value of the person you are bidding for. Data such as context, time of day and ad sequence all add up to give a specific value at any given moment in time. The only way to do this is to have optimization engines that can calculate bids based on outcomes you are looking for (CPA, CPC, or other metrics).

Price has a direct correlation to winning rate, so it can be worth bidding high when your algorithm says you will be targeting the person you are hoping to reach. For example, companies may sell an ad campaign for a flat $4 CPM but buy some inventory way above that price, often in the tens of dollars, because they know that a certain target has a high probability to convert. There are, of course, exceptions to this rule.

Another key point is that bidding high does not always mean paying a high CPM. A majority of RTB exchanges do play by rules of second-price auction. As bid prices increase, the higher the spread between bid price and paid price. 

Another must-have investment is fraud-detection software that can help you avoid non-human traffic, saving you a small fortune and providing peace of mind that your ad is being delivered to a real person. One important part of a bid calculation formula is accounting for quality inventory (if you’re an optimist) and fraud (for you pessimists). Make sure to also use external signals as well as home-grown detection systems to ensure that you’re serving the ad to a real person in a clean, well-lit environment. These quality algorithms should be required for value-based bidding.

RTB for Branding vs. Direct Response Campaigns

Although CPA direct-response campaigns are popular for marketers, they are usually not a very fair trade for the publisher/ad sales organization. For CPA campaigns, the most prevalent model is by far “post-click attribution.” Given the near-unanimous understanding that click-through rate (CTR) is not correlated with sales -- or put more bluntly, the number of people who click on an ad are a small fraction of the people who actually buy the product -- this means that the ad sales organization is really making more sales (or, at minimum, influencing more people) than they are getting credit for. 

CPC campaigns can be big problems for marketers, yet they are far less risky for the publisher/ad sales organizations. The inside joke among ad technology companies is that any optimization algorithm that optimizes for CTR rate is synonymous with a (primitive) fraud detection device. BotNets always have good CTRs, so if your algorithm isn’t paired with a good fraud detection service, you are likely to get clicks from non-human traffic. Again, to put it more bluntly: if the click-through rate looks too good to be true, it is. Do not rely on exchanges for fraud detection, and instead implement some sort of pre-bid fraud detection technology.

For branding campaigns, defining the target metric can be much more difficult.  Typically “interaction” or the dreaded CTR is suggested as a metric. The better solution is to go back to the future, and pay for a control segment, and then measure the increase in brand awareness, propensity to purchase, or whatever the goal is for the branding campaign. It can be simple for vendors to game the system and artificially make a control group look worse than the branding campaign, so make sure you (the advertiser) control which creative is shown, so you know that your control group is statistically relevant and that you’re able to measure your results.

As we, and the technology gets smarter, we will continue to face new hurdles for RTB. The more transparent the value chain becomes, from publishers to the service providers, the more we can help our customers reach their customers with efficient and effective marketing messages.