September 1, 2004
OK, you're a sophisticated search marketer and you immediately saw the need for automated bid optimization. Running thousands of keywords - on multiple engines - with varying cost per acquisition targets all competing in real-time auctions for top positions not only seemed daunting but utterly impossible to manage manually. You've made the right decision.
There are two classes of bidders: informed and uninformed... and you just became informed.
Informed traders make excess profits at the expense of uninformed traders.
The higher the proportion of informed traders, the faster bids change.
What makes an informed trader?
- Information about conversion rates
- Information about real-time ROI
- Information about click volume
- Information about data availability and data granularity
There are many technical aspects to consider when selecting an automated bidding tool or campaign optimizer. How do they track? How does their AI (artificial intelligence) function? Do they use business rules - or bidding rules, such as bid gaps? What type of metrics can they optimize on?
How often do they make bid changes? STOP! Let's take a look at the frequency of bid changes. Not that it is any more important than the other aspects, but it happens to be the topic of this article.
If you take a look at the competitive landscape, you'll find tools that have a fixed frequency that are not under your control, and others that allow you to set a universal fixed frequency for your campaign. So which is better? I'd say neither, and here's why.
The frequency of bid changes that a given keyword requires to fully optimize its results depends on the amount of traffic that keyword generates and the competitive landscape in which it operates. Obviously, if a keyword is generating a high volume of impressions and clickthrus, then it will require closer attention. It's a real-time auction; the more popular the phrase, the more bid volatility occurs. It makes perfect sense.
However, there are situations where relatively low volume keywords are competed for very aggressively. In those cases, despite the fact that they generate low volume, those keywords still require frequent attention because of the competitiveness surrounding them.
Now armed with these basic facts, you have to ask this question: How does the tool you're thinking of using deal with bid frequency? Obviously, the tools that have an inherent, fixed frequency don't deal with it at all. They have a preset frequency. Maybe once-an-hour, and you have to live with that. Well, like a broken clock, at least they'll be right twice a day.
Invariably, they will pay too much attention to words that don't need it and not enough attention to the words that do. In the first case all you did was waste resources, but in the second case, it cost you orders. The tools weren't able to act fast enough to compete and optimize those words properly.
“That's why we let you set the frequency,” say the more flexible tools. “You can lump your words under categories such as ‘slow,' ‘medium' or ‘fast,' and assign a different fixed frequency to each.” That sounds good, but it fails for a very basic reason. The frequency of bid changes and the competitiveness of keywords are constantly changing. Words increase and decrease in popularity all the time. And can you imagine deciding ahead of time how popular each of your keywords will be when you maintain thousands of them? It's impossible to do accurately.
What you need to solve this problem is a system that decides for itself how frequently to examine and change a bid for each keyword based on that word's needs. That system will give each keyword just what it requires to be fully optimized. Then keywords that are in hyper-competitive markets can get proper focused attention, while the less active ones are not checked too often.
If it sounds perfect, it is. We believe it's the best way to deal with the bid frequency problem. We call it Dynamic Bid Frequency Modulation (DBFM), and it's one of the cornerstones of our search management system. As long as you're writing an artificial intelligence bid management system, you might as well make it intelligent.
|