43 Paid Search Signals You Need To Understand
Google and advertisers incorporate at least 43 signals that answer six fundamental questions to value and satisfy the search and searcher. Here's a review of those paid search signals.
Google and advertisers incorporate at least 43 signals that answer six fundamental questions to value and satisfy the search and searcher. Here's a review of those paid search signals.
The SEO world often talks about the idea of signals – data points used by search engines to deliver the most relevant results. But the paid search community almost never discusses them (a point Kevin Lee made two years ago). Yet, our ability to participate and profit from ad auctions is dependent on a deep understanding and manipulation of a wide variety of signals.
Google is a publisher. Their systems are designed to maximize the yield on their search results – to maximize the number of searches, clicks on paid search ads per search, and the cost of those clicks.
As advertisers, our goal is to maximize the yield on our media spend – to maximize the number of clicks, conversions per clicks, profit per click, and minimize the cost per click.
Both parties incorporate at least 43 signals that answer six fundamental questions to value and satisfy the search and searcher:
Intent is far and away the most important tool for maximizing profit. What someone says they want right now is the core signal to determine the right combination of ads to display, if any.
Google only gets paid when people click on ads. As a result, advertising on Google isn’t a direct auction where the highest bidder automatically gets the top spot, because there’s no guarantee that the person with the most money has the most compelling ad.
Instead, Google looks at a combination of quality fit and economic fit. Put another way, they calculate ad rank by multiplying your bid times quality score.
The best quality fit for Google, in most cases, is ad most likely to be clicked on. Google has disclosed some of the components of quality fit, most thoroughly documented in “Quality Score in High Resolution.”
In general, more data is better and newer data is better. The greater your reputation for getting searchers to click on your ads – historical, recent and specific to the current search and searcher – the more likely you are to have your ads displayed in the auction in a higher position at a more favorable price.
No matter how great your great quality score, there is a minimum amount you have to pay to get on the first page and, based on a competition, an increasing amount you have to pay to move up the page.
We obviously use different and more varied signals to value search traffic, including:
Not every searcher is equally likely to click on a specific set of ads, even if they’re using the same terms. And on the Display Network, where no search query can indicate intent, more signals are necessary to deliver the most relevant ad.
Google has a variety of data to determine not just what you want, but who you are (in an anonymous way, of course). This pulls from behavioral data gathered from AdSense and remarketing cookies, social connections from Google and other online profiles, topical interests gathered from sites visited in the Display Network and even demographic information from third parties.
Search queries are the greatest indicator of intent, because they are explicit and recent. In addition to, or in the absence of, search queries, recent and historical browsing behavior can help us and Google to value and target different audiences appropriately.
Last week, Google rolled out their +1 button that allows users to share pages with their network of contacts in both paid and organic results. For more detail on social signals and PPC, read my previous column, “How Social Media Affects Paid Search .”
Display network targeting works either through managed placements, where we choose specific sites, or automatic placements, where Google analyzes the keywords in the ad group to pair with thematically similar sites and pages. There is also middle ground for targeting called topics.
Some sites on the Google Display Network, such as social networking sites, collect and opt to share demographic data. Google, which it’s worth noting does not collect these data themselves, enables specific bidding on demographic data when available.
Similar to enhanced CPC, bids are adjusted in real time when the data are available and the preferred audience is available, including:
The way people search, especially the device they use, and where the network where the ads are served, search vs. search partners vs. display, can say a lot about what someone is looking for.
Mobile devices are probably the most dramatic example of this. If you’re searching for car rental from your phone at Heathrow, it’s pretty safe bet you want a rental company in London and need a car right now.
How you search provides implicit signals of relevance.
Mobile searches are different. Users are very local. For example, 53 percent of mobile searches on Bing have a local intent. Eric Schmidt from Google also said “1 in 3 queries from smartphones is about where I am.”
Mobile users are also very urgent. Seventy percent of PC “query chains” (essentially search tasks) are completed in about one week while 70 percent of mobile users do so in one hour.
Similarly, Pew Research found that:
Around one in ten Americans with annual household incomes of $75,000 or more own a tablet PC or e-book reader, while fewer than 5% of households earning less than $50,000 per year contain one of these devices.
Not surprisingly, the different sized screens, download speeds, and mindset of searchers among the devices means that maximizing yield requires different ad types and, sometimes, advertisers.
By default, Google lumps together campaigns to target both the search network, people who are actively searching on Google.com or one of their search partners, and the Display Network, the sites that monetize with AdSense.
This is a clear money grab, because the way you manage search and display network campaigns is entirely different, requiring different ad group structures, creative, and bids.
Temporal signals, the time when someone searches, can hint at your needs or value: weekends vs. weekdays, days vs. nights, work vs. home, and so on.
Not every searcher is equally likely to click on a specific set of ads, even if they’re using the same terms. Take the previous example of “paris hotels.” The auction for someone searching with that query in Texas would look very different than someone in France.
This is especially true for local searches, such as “Philadelphia gym,” which can often trigger results from local specific ad formats, such as location extensions and Google Boost.
Groups are not homogenous. Groups don’t purchase or click on ads, people do. Every signal is used at a tool to determine which specific set of ads are most likely to appeal to this particular person at this particular time.
Intent gathered from a search query is the clearest window into what someone wants and when, because it is both explicit and immediate.
Behavioral data, especially that used for remarketing, is also very influential, because of the ability to target people based on recency of behavior (varying the cookie length) and value of the action (by creating small segments using multiple tags).
The greater of a predictor of someone’s likelihood to click, the greater the weight that signal carries. The less explicit, recent and specific the signal, the less influence it has. I hypothesize the priority looks like this:
Google prioritizes what drives their bottom line (CPCs, clicks). We need to understand what they care about and manage accordingly, for example improving CTR to boost quality score.
Ultimately, however, we need to prioritize what drives ours (profit, leads). The more accurately we can value a click, the better we can extract profit from it.
If there’s one thing that should be clear from this article, it’s that data drives results. Newer data is better than older data. More data is better than less data.
Google clearly understands that whoever has the data wins.
This is doubly true for advertisers, because we know what happens after the click. We can optimize landing pages, offers, checkout, and other conversion variables. We can also determine which kinds of customers are most profitable in the short term and over time (LTV).
Like Google, we must compete with data to maximize our yield. As auctions become more competitive and new ad formats remove some of the typical controls, such as is the case with product listing ads, which don’t require keywords, the ability to successfully target, value, and convert niche audiences will separate the most profitable companies from the least.
Put more simply, we need data to move budget from early funnel to late funnel targeting.
Operating at scale can’t be done manually. Google doesn’t have a sales team negotiating each search result, even for the highest volume, most valuable terms. Automated systems are required to handle data at scale
All advertisers of significant size need their own suite of paid search tools to automate the use of signal data to the degree possible. They reduce human error, operate at a higher speed and breadth, and free up talent to focus on strategy instead of tactics.
This list is not exhaustive. It couldn’t possibly be. Like organic search rankings, there are surely more signals that Google uses to manage the auctions that are not published. Not every tool Google uses to maximize yield is available to advertisers. Likewise, not every signal we use to maximize our yield factors into the auction.
It’s also important to remember that, like SEO, not every signal carries equal weight. If you’re dayparting campaigns with big, unrelated ad groups, poorly written text ads and sub-optimal landing pages, then you’re just rearranging flatware on the Titanic.
My hope is that this article will establish signals as part of the PPC managers’ vocabulary, so that we can continue to document and prioritize them for everyone’s benefit and to inspire Google to offer more transparency. If you can think of any signals I missed, or want to take a stab at prioritizing them, please leave a comment.
Editor’s note: This column originally was published on June 7, 2011, and comes in at No. 7 on our countdown of the 10 most popular Search Engine Watch columns of 2011. Over the final two weeks of 2011, we’re celebrating the Best of 2011 by revisiting our most popular columns, as determined by our readers. Enjoy and keep checking back!