Real Time Bidding also known as an RTB:
Unless you have been living under a rock….
…You’ve heard this term several times in your marketing career.
What is a RTB?
A Real-time bidding platform allows advertising inventory to be bought and sold pragmatically. The RTB uses algorithms to help the publishers get the most money per impression. While helping the advertisers get the most relevant traffic for the best price. Each impression is dynamically bid on based on different criteria such as: placement, category, device, gender, etc.
This all sounds great and dandy…
…however as an advertiser, this puts everyone on an equal playing field.
To be successful on an RTB, most marketers assume their only way to get more traffic is to raise their CPC(Cost Per Click). – Which YES is a factor, but there is more at play here.
This universal untruth causes RTB’s to be very unstable for the advertiser, as marketers we’re all competing by inflating click prices that causes a “race to the bottom.”
The Good News…
We can lower our competition by understanding how an RTB works.
How An RTB Really Works
No matter what paid traffic source you’re working with, the backend technology is an ad server.
The ad server is in charge of calculating which ad is to be served milliseconds before a page load. Once a winner is calculated the ad is served on the publishers web site or app and the advertiser is charged for the impression.
One thing I want you to take note of, is how the ad server is calculating all of this based on the price of an impression.
You may be asking yourself how does this work if I am paying per click?
Even if you pay per click the most ad servers still rank you based on what’s called your eCPM (Estimated Cost Per Thousand impressions).
Say an advertiser was bidding $1 / CPM – then their eCPM would be $1.
With a CPC bid it gets a bit more tricky.
Say you are bidding $.15/click and you have a average CTR(Click thru rate) is 1%.
That means that out of 1,000 impressions you would get on average 10 clicks at a price of $.15/click.
This would make your eCPM $1.50. The formula is:
1000 * CTR * CPC = eCPM so in this case: 1000 *.01 *.15 = $1.5
So in the example above, if everything else was equal you would win the auction bidding $.15/click.
But here’s the kicker:
If your CTR dropped down to 0.5% you would lose the auction 1000 *.005 *.15 = $0.75
Most marketers would simply raise their bid. Others would add a few new creatives.
A few would write off the campaign as a loss and move on…
To be honest, once your campaign gets to this point on most RTBs it’s DEAD.
Once your eCPM has dropped and you lose your ad position, shit just hit the fan.
The reason for this is very simple:
If your CTR was already dropping while you were the very first ad shown, just imagine your CTR at position 2, 3, 4, etc. *Hint not very good.
So it is an endless cycle and your campaign will eventually die off.
You are almost to the point where you can understand the reasoning for what I am about to share.
When you first setup a new marketing campaign on an RTB, it has no way of knowing what your eCPM is.
Here’s the deal:
Each ad servers algorithm is different….
….but they all do one thing well.
Figure out your eCPM as quickly as possible to determine if your campaign will perform better or worse than everyone else.
They do this by the use what is called a multi-armed bandit algorithm.
Without getting to technical, multi-armed bandits send the most “turns” or in our case “traffic” to the currently know best option.
It does this by giving the majority of traffic to the current winner known as exploitation.
While giving a much smaller portion to help find new winners, called exploration.
Lower CPC & More Traffic
A simple example of this would be 90% of the traffic is sent to the current winner while 10% is sent to the exploration phase.
So if there was 3 advertisers A, B & C. All with one creative and A being the current winner the inventory distribution would be as follows:
However, with A still being the winner. If advertiser B had 9 creatives and advertiser C only had 1 our numbers would look something like this:
The above example is true for most RTB platforms because they base the current eCPM to the creative themselves.
This is crazy:
The ad server has no idea if a creative is the same or not.
So put simply the more ads you put in your campaign, the better chance your ad will be chosen in the exploration phase.
Want to know the best part?
Chances are by pure luck, one of your creatives will get a higher CTR even if they’re all exactly the same.
It is even possible that one of your creatives to get a much higher CTR than anything else currently in the ad chain and your campaign would be shot to the #1 position even if your CPC was much lower.
I get asked this questions A LOT:
“Brent what should I do, my ad rep told me the top CPC is $X.XX which is to expensive for me!”
Take everything that your ad rep tells you with a grain of salt.
They have one job, to sell ad inventory at the highest price possible.
Never base your entire bidding strategy on beating the current top bid.
That top bid could be on a single placement, certain device type or even a brander who doesn’t care about performance.
Point is trying to run an unoptimized campaign at the same CPC as the top bid, in most cases is a death sentence to your bank account.
We already learned more ads = better chance of success.
What is also just as important is splitting up everything to better your chances of finding a winner.
Let’s say you have a budget of $1,000 a day and you want to buy US mobile traffic.
Instead of setting up one campaign with $1,000 daily cap targeting US mobile traffic.
Break up the campaigns like so:
US – iPhone – 300×250 : $250/day @ $.05 CPC
US – iPhone – 320×50 : $250/day @ $.05 CPC
US – Android – 300×250 : $250/day @ $.05 CPC
US – Android – 320×50 : $250/day @ $.05 CPC
Each campaign would have 10 of the SAME banner.
Want to test another creative for the same ad size?
Your campaign break up would then look something like this:
US – iPhone – 300×250-1 : $250/day @ $.05 CPC
US – iPhone – 300×250-2 : $250/day @ $.05 CPC
Also, I like to always start a campaign with the minimum bid possible.
This allows me to see how much traffic I can get at that rate & test the quality of traffic.
If for whatever reason, I can’t get traffic at that rate only then will I raise the bid.
Never start a the highest CPC and work your way down.
Once you finally start getting traffic write down your CTR & CPC.
After a while your campaign will naturally die off from banner blindness.
When this happens do not raise your bid let the campaign run & if possible clone the campaign.
The new campaign will go through the same exploration phase and chances are you’ll get traffic again.
One of the great things about RTB’s is you can be very selective on what you bid on.
Because of this….
….Most marketers screwup their campaigns.
If you find a particle KPI (Key Performance Indicator) doing well, it’s natural to want to cut anything not doing as well.
Making optimizations to a campaign while it’s running seems good in theory.
However, chances are your changes will also change your campaigns eCPM making you lose traffic.
Completely changing your position in the ad server which can result in showing up on different placements, etc.
What’s the bottom line?
If you find a particular KPI that is doing well for you, clone the campaign and make your targeting changes.
This will allow you to see if making the modification actually will improve your campaign or not. If it is successful then you can pause your original campaign. However, if it fails at least you still are collecting data from the original.
More ads the better your chances are to have a high eCPM.
Break up everything to better your chances of finding a winner.
Once a campaign is running do not touch it.
Now obviously every RTB & ad server has it’s own kinks.
Also, every marketer will build their own strategy over time.
These are the rules I follow on most RTB’s and often find success.
If this post has helped you & you think someone else would benefit please help me out by sharing it.