Monday, April 04, 2011

Understanding the Average Position metric



For today's Inside AdWords post, we have a note from our Chief Economist at Google, Hal Varian. Hal and his team spend most of their time studying the AdWords auction and finding ways to make it more efficient. Today he'd like to share some insights on the average position metric:

Many advertisers are concerned about the average position of their ads. Though this metric can be useful, it's easy to misinterpret. This blog post is intended to help advertisers better understand the average position metric, its uses, and its limitations.

To begin with, it's important to understand that there are two interpretations of the phrase “ad position.” The “page position” refers to the location on the page, such as “top ad 2” or “right-hand side ad 1.” The “auction position” is the rank of the ad in the auction that determines the order of the ads on the page. The critical point is that the reported average position metric is based on auction position, not page position.


Example showing page with top and right-hand side ads


Example showing page with only right-hand side ads

This means that an ad in auction position 1 will always be the first ad shown on the page, but it can occur in two possible page positions: as the first ad above the search results or as the first ad on the right-hand side when there are no ads above the search results. This distinction is important, since, on average, ads that appear above the search results tend to get substantially more clicks than ads that appear on the right-hand side. The difference between page and auction position can lead to some seemingly paradoxical outcomes.

An ad with a larger average position may perform better than an ad with a smaller average position

If you bid on a keyword that many other advertisers are bidding on, you may end up in a relatively low position on the page due to the competition from other bidders. However, the very fact that many people are bidding on that keyword suggests that it will perform well, even when your ad is placed lower on the page. On the other hand, if you bid on a keyword that no one else is bidding on, you may end up in position 1 by default but end up with few clicks. What matters is how the keyword performs in terms of clicks and cost, not where it ends up on the page.

Increasing a bid may improve page position but not auction position

Suppose your ad always appears in position 1 on the right-hand side, and there are no “top ads” above the search results. You now increase your bid. If your ad quality is high enough, the bid increase can move it to a page position above the search results. This improved page position will typically produce more clicks, but there will be no change in the average position metric, since the ad’s auction position hasn’t changed: it's 1 in each case.

A bid increase may move your average position lower on the page

When you increase your bid, your ad will generally show up on the first page in more auctions, particularly when using broad match. Since you weren’t bidding enough to appear in those auctions originally, your ad will tend to come in at the bottom of the first page in those new auctions.

For example, think of a situation where you're bidding on vegan dog food (broad match), and your ad consistently appears at auction position 1 when the user enters vegan dog food, but your bid is too low for the ad to show (on the first page) when the user enters dog food. Now you increase your bid and start showing up in auction position 7 for dog food queries. If there are the same number of queries for dog food and vegan dog food, your average position will be (1+7)/2 = 4. So by increasing your bid, you've moved your average position metric from 1 to 4—even though your ad never appeared in position 4!

An average is not always the best way to summarize a distribution of positions

The seemingly paradoxical effect described in the last example can even appear with exact match, since differences in bids and budgets among advertisers lead to considerable variation in auction participation. You should think about “position” as being a distribution of outcomes—sometimes you're in one position, sometimes another. The “average position” is one way to summarize this distribution, but it isn’t necessarily the best way.

It's important to keep in mind that average position measures your position in the auction, not your position on the page.

Examining bids, clicks, and costs

One very helpful tool for examining the relationship among these metrics is Bid Simulator. The Bid Simulator shows you an estimate of the "click-cost curve"—a simulation based on historical data of how much you would have had to bid, and how much it could have cost, to acquire some target number of clicks in the previous week. If you combine the click-cost curve with your value of clicks, you can find a bid that maximizes your overall profit. It also shows you an estimate of the number of top and RHS impressions an ad could have received, a metric that tends to be more meaningful for measuring ad exposure than average position.

Some advertisers have asked how clicks from different positions tend to convert. In general, we've found that conversion rates don’t vary much with the position of the ad on the page. An ad in a more prominent position on the page will tend to get both more clicks and more conversions than an ad in a lower position, but the conversion rate (conversions/clicks) will tend to be about the same for the two positions.

In summary, average position is a popular metric, but don’t rely on it alone as a measure of performance. The metrics that really matter are clicks, costs, and conversions.

Posted by Lisa Shieh, Inside AdWords crew