- The Search Agents - http://www.thesearchagents.com -

Position 1 Does Not Necessarily Imply Maximum Impressions

Google doesn’t shout it from the mountaintops, but not every search results in an impression for a qualified ad, even if that ad averages placement in the first position.  That is, if you bid such that your ad always appears in position 1 and there are 1000 searches each day for which that ad could be shown, that ad still might only be shown a fraction of the time; perhaps 1000 times, but perhaps only 800 or 600 or fewer.  This fraction, the ratio of the number of times an ad is actually shown to the number of times it is eligible to be shown, is called the Impression Share.

Google has revealed very little about the rules it uses to determine whether an ad that is eligible for an impression will actually be shown or not.  But, AdWords does report (at the campaign level only) the portion of each campaign’s Impression Share lost due to lack of budget and lost due to low rank.  The impact of a budget cap on Impression Share is obvious: after a campaign has run out of money, then none of the ads in it will be shown whenever a new search occurs.  (The implication for advertisers is to not use campaign-level budget caps at all, if possible.  If an ad is profitable per impression, then you would want that ad to be shown as many times as possible.  If it’s losing money per impression, then you should decrease its bid or shut it off completely.  But you shouldn’t shut off an entire campaign worth of ads, some of which might deserve more impressions, simply because that campaign has reached some collective spending level.)

The impact of rank on impressions is not as clear.  Obviously, an ad whose position averages 1.0 (the top of the page) would probably be shown more often than one whose position averages 9.0 (the bottom of the page), because the one in lower position might get bumped periodically onto the second page, where the chance of getting an impression is very low for any given search.  But what’s not as obvious is that an ad whose position averages 1.0 might be shown less often than that same ad (still in position 1.0) but with a higher bid.  In other words, even an ad which is in position 1.0 and which has no campaign-level budget cap hindering it might still have its impressions curtailed due solely to its bid.

To show that this is the case, I looked at the estimates of impression volume offered by the Google Bid Simulator (GBS).  For those who are not familiar with this feature, Google provides an estimate at the keyword level, for 4-7 various possible bids, of the number of impressions (among other things) each word could have gotten in the past 7 days.  For keywords which get a sufficient number of clicks per week (where “sufficient” seems to be about 25-30), Google also estimates clicks and cost for several different possible bids.  It’s important to note that the estimates that the GBS provides are not predictions, in the sense that they are not intended to project what will happen in the coming week.  Instead, they are retrospective tests: They assess, based on the traffic that actually did occur, what would have happened in your account (with all other factors assumed unchanged) had you bid at different levels.

In a previous AdWords interface, the GBS also provided estimates of the average position at various bid levels.  So I sifted through the numbers for 8 high-volume accounts and found the keywords which had multiple estimates which were at an average position of 1.0, but with different bids.  (Though I’ve previously blogged about the fact that average position can be a misleading metric [1], the case where the average position is 1.0 is a special one, since to receive that average position, all (or almost all) of the impressions which occurred must have been in that position.)

For all of the exact-match keywords, the estimated number of impressions per week at the highest bid that had an average position of 1.0 was exactly the same as (or almost exactly the same as) the number of impressions at the lowest bid that had an average position of 1.0.  (I refer to the ratio of estimated number of impressions at any bid higher than what achieves position 1 to the number of impressions at the lowest bid that achieves position 1 as the Impression Uplift.) In other words, when the bid for an exact-match ad is at the lowest level necessary to reach average position 1, the GBS estimates that increasing the bid further will not bring any additional uplift in impressions.

For the phrase-match keywords, the average estimated number of impressions at the highest bid with average position 1.0 was a little over 4% greater than the estimated number of impressions at the lowest bid with an average position of 1.0.  About 40% of the phrase-match words saw less than a 1% uplift in impressions.  Only two words were found where the number of impressions increased by 10% or more, with the larger one seeing a 26% uplift.  In this latter case, the bid to get to first position was about $10/click, but the bid to get those 26% more impressions was over $30/click.

For the broad-match keywords, the average estimated number of impressions at the highest bid with average position 1.0 was almost 8% greater than the estimated number of impressions at the lowest bid with an average position of 1.0.  About 1/3 of those words saw less than a 1% uplift.  About 1/4 saw 10% or greater.  Two words saw an increase of nearly 60%, one of which is shown in the figure below.


At a bid of $3.25 the Google Bid Simulator estimates this word would receive 685 impressions per week.  (The older version of the GBS indicated that this bid and all those above it would result in an average position of 1.0.)  Increasing the bid to over $5, $7, or $10 resulted in up to about 64% more impressions.  In the 7 days on which these estimates are based, this word actually had a bid of $3.47, was in first position, and received just over 700 impressions and over 20 clicks.


Interestingly, of the phrase- and broad-match words that saw at least a 5% uplift in impressions, only one was a brand term and only one was a competitor’s brand term.  Most were just generic terms relevant to each account’s business.

The take-home lessons for the average search marketer are:

* that the Google Bid Simulator says that increasing bids on exact-match terms which are already averaging position 1.0 will not bring additional impressions; and,

* that increasing bids on phrase- and broad-match terms that are already in position 1.0 could bring in an average of about 4% and 8% more impressions (respectively), with the largest portion of words seeing no increase, the next largest portion seeing up to a 10% or 20% uplift (respectively), and a small handful seeing a 20-64% increase.

In short, just because your phrase- or broad-match ad has reached position 1.0 does not mean that its impression volume is maximized, but when your exact-match ad hits position 1.0, the Google Bid Simulator says that there’s no need to increase its bid further.

About Bradd Libby

Bradd easily becomes obsessed with small details. He can barely make it through a conversation without drawing a graph of something. He lives in the woods on the side of a hill in Norway. Bradd doesn't willingly wear shoes. He makes maps of places that don't exist. He picks up small shiny things he finds on the ground. He devolves into violent shouting matches with himself. He takes things apart to find out how they work but then can't figure out how to put them back together again. He also writes at Search Engine Land [6], Search Engine Journal [7] and his personal blog [8].