It has become the cheap word that every ad network, rich media company, technology provider, publisher, and agency drops when they attempt to sound sophisticated.
It makes its way into every media plan, and sneaks onto every vendor Web site. So what's wrong with optimization?
Optimization is like love. It is an easy thing to say, and much harder to put into practice.
Worse, everyone means something different when they say it, and most mean very little at all.
This is a shame, because if the first generation of online display advertising was about delivering reach, and the second generation was about delivering audience, the third generation will most certainly be about delivering ROI through optimization.
In order to have an effective debate, we should first agree on what we mean by optimization as it relates to online display advertising. I will offer up a simple definition: Optimization is the process of maximizing ROI by evaluating all of the relevant data points.
The goal in all cases should be really simple: profit. While advertisers have varying objectives short-term, every single advertiser hopes to earn more money on their investment long-term than they spend. The only variable is timeline. Direct-response or transactional advertisers seeks to capture the purchase instantly or near-instantly. The rest of the advertisers hope to advance the sale now, and capture the purchase eventually.
Rocks, Pebbles, Sand, and Water
The first step toward helping advertisers achieve ROI goals is to connect them with buyers, not demographic groups.
Relying solely on demographic data is a recipe for massive inefficiency. Individuals are only willing to buy a certain number of products at a given moment, so advertising to an entire wide demographic like 18- to 34-year-old males, rather than the 5% of Web site visitors willing to buy a pair of shoes in the next 10 days, is a mistake.
Add in their shopping habits, behavioral data, context of the sites they visit, spending history, and other data points and you will get closer, but you then need to know what order to use the data in.
In fifth grade science class, we all watched a teacher fill a jar with rocks until the jar was full, then add pebbles, then sand, and finally water until nothing more could be added. Each time it looked full until a smaller-sized item was added.
The same logic applies to optimization. If you want to maximize the effectiveness of your data, you need to work in order of granularity from largest to smallest.
Knowing What to Ignore
Social networks routinely tout their vast mountains of data, yet they earn a small fraction of the average CPM rate Online Publishers Association members get. One reason for this disparity: advertisers do not know how to leverage the social network data to improve their ROI.
There are more than 100 discrete types of user data, and thousands of companies that own slices of data. These data sources are far from equal.
Over the next five years, our industry will face the same challenge every Wall Street analyst has: figuring out which 90 data points to ignore, and which 10 really matter.
Targeted Delivery Without Being Creepy
Once you have the right data points, the next challenge in optimization is selecting the delivery vehicle for the message.
Data-driven ads use programming hooks (typically an API) to enable you to query a back-end database and retrieve the right ad unit for a given set of data points.
Early implementations of data-driven ads have been primarily focused on trivial elements such as swapping out buttons, photos, and color schemes. While this can have some impact on take rate, there is a much larger future for data-driven advertising.
The real value is to enable you to create a one-to-one message in real-time that offers a product to a customer who is "in-market" for that category of product at any given moment.
Get Real About Performance
Performance-based advertising is the future of our business, and if done right, everybody wins. If you can understand a user's intent, you can deliver advertising that feels like a feature. A better user experience will lead to higher clicks, and thereby increased advertiser ROI. As ROI increases, publishers have the ability to increase rates without a negative advertiser reaction. This creates a vicious cycle in the right direction.
Choosing to use optimization technology is an IQ test for advertisers. For publishers, it will rapidly become de rigueur.
Advertising execs often refer to the last mile problem: how to deliver the right creative to the right audience. There is in fact a last mile problem in advertising, but these execs got it wrong.
The last mile is optimization to help advertisers maximize ROI and does not end with marketing at all.
It ends in the CFO's office.
The day the buck stops in the CFO's office, is the day that we need to hang it up as an industry and all go work for "As Seen on TV". I agree that marketing has become increasingly performance based (and that it is a good thing), and that over time better performance relates to a better ROI.
The reality, however, is that the path to ROI performance is long and arduous, and before turning consumers into customers we as marketers first need to generate awareness and consideration. If we relegate everything that we focus on to an immediate direct response (as measured by a 30, 60-, 90-day CPA basis), we will provide short shrift to our customers, and not do anything to help them build the brands necessary to have a long and happy life. Maybe the dirtiest of all words is performance? Is it about an immediate ROI on each campaign element or a positive ROI over time?
A co-worker of mine wrote the following; I think it is appropriate here.
Amen -- nice article. It's still a long road to educate agencies and advertisers that it's not about buying predetermined buckets, stereoptyped profiles or inventory blocks based on the usual suspects -- where they think their buyers will be looking, or shoehorning advertisers into pre-determined generic audiences. Our customers always want to know our reach in a particular demo or behavioral segment -- we tell them how much reach do you need and we'll go get it. It's not about buying segments -- it's about having the technology to go find those customers and "activating" them anywhere they are on the web. If a person is on a cooking site, but is an avid runner, and you know that, wouldn't it make sense to show them a running shoe ad? It's not just about any one type of data -- behavioral, geo, demo or social targeting -- it's about building a system that is data agnostic or "hybrid" -- one that uses every type of data available to build the best picture of the Web surfer to serve that ultimate "personal" ad. And yes -- all data is not created equal.
@Grant, I agree with you that not everything can result in an immediate purchase (I think I said this, when I said that the variable is the timeline).
That said, once a sufficient supply of quality display inventory is available on a performance pricing basis, the rest of the market will fall in this direction. It's a tough train to derail once it gains momentum.
Google has often argued that its text ads have a brand influence as well, but advertisers will tell you that this value is baked into the ultimate CPC cost.
Good one David! I agree-selecting the delivery vehicle for the message is key with online display. I wonder how the Mobile and D-OOH folks will apply optimization techniques in the coming years? - Tony Anderson - Gen-Y Media Inc. - email@example.com
DK, interesting article as always.
@DK "That said, once a sufficient supply of quality display inventory is available on a performance pricing basis, the rest of the market will fall in this direction. It's a tough train to derail once it gains momentum."
Not sure I totally agree with the above statement for a few reasons. There are currently many places to get 'quality' inventory at little or no upfront cost and this hasn't helped the advertisers or the publishers to improve their respective metrics. 'Quality" is subjective of course. The advertisers have only improved their cash outlay, not their conversion metrics. The publishers have decreased their cash flow positions, which in the long run doesn't help anyone. I do agree that more 'buckets' are needed to help improve display ROI. Semantic Technology is slow and hasn't shown that it has scale, at least yet. IMHO, behavioral is the only way, at least right now, but who knows how that is going to limited in the coming years with pending legislation. The days of the 'awareness' campaign have come and gone.
So, if the 'train' needs momentum the 'fuel' is going to be a model that mutually benefits all parties not just the advertisers. This means everyone needs to be on board with more testing to create win/ win models that achieve everyone's objectives.
@Michael, it absolutely is going to require publisher and advertiser cooperation. This has been hindered in my opinion by an overly complex ecosystem that has too many seats at the table.
As for the market moving in the direction of performance-based, I think it is already happening. There is tremendous pricing pressure on premium publishers that are charging the $10+ CPMs. Some smart guys in our industry like Dave Morgan believe the CPM prices will never recover.
I believe the only path to higher CPM's is going to be performance-based advertising that actually delivers for advertisers.
Inherently all advertising is performance based. However, performance cannot always be directly correlated to a click. That is why measuring brand ROI is difficult. Obviously, we cannot target only those in 'buying' mode unless we implanted a cookie in every consumers brain to measure intent. So we optimize, target demographics is one of the techniques. Is it perfect, no. Is it wasteful, no. Not if you can measure brand awareness and equity growth. Just like a click these metric also are a road to measuring ROI.
@Jonas, actually we believe you can target those in buying mode. We are doing a lot of work at Adventive in this exact area.
That would truly be ground breaking. I'd love to see it.