Search marketers hear a lot about portfolio optimization for paid search campaigns. Manual or rules-based bidding of keywords may get you some basic optimization benefits. But for enterprise advertisers, portfolio optimization is acknowledged as the gold standard to squeeze out every last drop of return from paid search.
Describing portfolio optimization is straightforward. First, build a mathematical model for every keyword in your campaign that describes exactly what you’ll get at every bidding point: clicks, cost-per-click, position, revenue, and so on. Then, given a goal (like a monthly budget or cost-per-acquisition target), advanced mathematics can determine exactly where to invest every dollar for the maximum possible return. As you collect new data, update the models and repeat.
While easy to describe, it’s hard to execute. Adobe Media Optimizer’s algorithms have been updated and refined over several years. There’s a constant process of innovation as the SEM landscape evolves around Google product listing ads or enhanced campaigns.
So how does a marketer evaluate what’s going on inside a portfolio optimization platform? Google announced bidding features within AdWords. Other platforms also promote features that are described as portfolio optimization.
Illumination is a good metaphor to use: tungsten, halogen, and LED bulbs all might emit the same amount of lumens. Yet, LEDs are far more efficient. How can you make sure a platform is producing light instead of heat?
Like me, you probably don’t have advanced mathematical or statistical knowledge, but you still want to understand how it works.
There’s a handful of “must haves” marketers should insist on. In this post, I’ll discuss why these are important and how Adobe Media Optimizer answers each of them.
Keyword Models Available in the User Interface
If you want to optimize algorithmically, understanding the possible costs and returns for every keyword at every position is essential. Let’s be clear: this isn’t a nice-to-have, it’s table stakes. You can’t do anything else if you don’t have keyword models.
Every keyword model built by Adobe Media Optimizer is available within the user interface. We call them bid units. A bid unit is more fundamental than a keyword. If a keyword is in a portfolio on three different match types, in three different geographically targeted campaigns, and on two search engines, that results in 18 separate bid units. AMO also builds separate models for Google product listing ads, where exposure, rather than position, is more important.
For Adobe Analytics customers, the latest engagement data algorithms are also informing the revenue part of the model. It’s bringing even better model accuracy and higher ROIs. Latest engagement data? I told you LEDs were more efficient!
Any Adobe Media Optimizer customer can inspect every bid unit within the user interface and compare forecasts against actual bids placed. You can also break it down by mobile and computer/tablet, but more on that later.
If you can’t see keyword models in a platform’s UI, you can’t be sure that they are there at all. At best, it’s a black box modeling approach. More likely, the platform isn’t doing keyword modeling at all, and therefore doesn’t have the foundations to do portfolio optimization.
Anyone can make a forecast. You made a forecast if you left the house with an umbrella this morning. What’s important is whether a forecast is accurate.
Adobe Media Optimizer bares all with forecasting, and provides Model Accuracy reports right within the UI:
If you have accurate models, then you will find better performance and ROI. But accurate modeling is key: if it’s not accurate, it’s worthless. Adobe Media Optimizer exposes this in the UI for customers to verify accuracy for themselves.
Accurate simulations available on demand within the UI
The Adobe Media Optimizer UI automatically creates simulations for every portfolio once a week. Customers can run custom simulations in minutes in the UI at any time, tweaking settings to meet their needs. There’s no need to request simulations from account managers — though our support team is always on hand to help with questions and best practices.
Advertisers use these simulations to budget, plan and execute their campaigns. To do that, they need confidence in the accuracy of the simulation.
Accuracy needs to be within a single digit percentage. Large error boundaries of 20%-40% are not acceptable — that’s the difference between a resounding success and abject failure. I wouldn’t want to bet my search marketing career on that.
Optimizing adjustments for Google Enhanced Campaigns
Google AdWords has changed dramatically in the last 12 months. Optimization used to be about bids and budgets, but with Enhanced Campaigns it’s now about bids, adjustments and budgets. Setting the right bid adjustments is critical to success.
Scroll back up to the bid unit example, and you can see how we’re already modeling differences by device. Adobe Media Optimizer is making algorithmic adjustments at ad group level. We’ve seen that this level of bidding makes a winning performance difference. Later in the year we’ll be releasing algorithmic bidding for geographic, time of day and audiences adjustments too.
Note that a text box in the UI where you can enter a bid adjustment is not optimization. To really push performance with Enhanced Campaigns, you want a platform to algorithmically calculate that value for you.
There’s plenty of detail I haven’t gone into here. In future blog posts I’m going to discuss how Adobe Media Optimizer deals with seasonality, sparse data, keyword learning and more. But I hope this helps you understand what portfolio optimization really is, and how to verify the many claims you’ll hear in the industry.