The DSP and why bid optimization is only half the battle
So this is my first work-related post. Apologies to those who aren’t interested in online marketing.
There’s been a lot of back and forth in the media lately about the imminent arrival of so-called Demand Side Platforms, or DSPs. (If you aren’t familiar, the gist is that DSPs will maximize online advertising ROI through sophisticated bid optimization, combined with features such as third-party data integration, real-time bidding, global frequency management, etc.)
The one thing that I haven’t seen discussed explicitly yet is the following:
The bid optimization found in DSPs needs creative optimization to function correctly. It sounds counter-intuitive, but is actually straightforward if you follow a simple line of reasoning.
Many studies have shown that variations in creative can affect the performance of intent-to-purchase (or whatever your metric du jour) by as much as 10x.
If you accept the premise that performance is tied to value, then this means that the right creative can improve the average value of an impression by as much as 10x, or vice versa. (Or 2x, or 3x – whatever.)
But that doesn’t mean that every impression’s value will improve by 10x if paired with the right creative – just the average. Some will improve by 5x, others will improve by 15x. A range of values will emerge around the mean, and the true value of each impression must be determined separately.
What that means for bid optimization then is that for a bid to be truly optimized, it has to be made with a view or prediction of how that impression will perform, when paired with an optimized creative.
For example:
Impressions #232334342: Male, 34, San Francisco, with interests in football, cars, and looking for a first mortgage. Good credit. College graduate.
Bid #1: $3.00 eCPM -based upon average value of all past creative variations
Bid #2: $5.00 eCPM – based upon the prediction that ad variant 6, with background color 8, message 4, and call-to-action 10 is expected to be worth a $5.00 eCPM when paried with this particular impression.
Who do you think wins? Clearly the bidder with better information, Bidder 2.
Bidder 1 on the other hand – because he’s constantly bidding the average – will overbid the less valuable impressions, and underbid the more valuable impressions. Lack of information yields poorly optimized bid, leads to lack of performance, value, and ROI.