The Elephant in the Google Offices
According to a recent article by Eric Jackson on Forbes.com, Google has an odd habit of focusing most of its prepared remarks on things outside its core business (AdSense/AdWords). In its Q4 analyst call, Google executives spent the lion’s share of their prepared remarks on Android, mobile search, mobile display, Google Enterprise, and Google +, and only focused on its core business in a granular way when analysts asked. They did this despite the fact that the ad business comprises 96 percent of Google’s revenue.
The obvious reason Google wants to focus outside of its core business is because growth to the core business is slowing. Google’s numbers in its core business were down year over year for the first time in a while, despite that fact that query volumes are way up. The core business is not hurting. The business still put $10 billion in cash into Google’s coffers in 2011. But analysts expected more because Google promised more. And Google promised more because its own query volume data predicted it.
So why did paid search decline while search volume increased? Google executives explained to the analysts that Google made many improvements to the product in Q3 that all matured in Q4. I’m speaking of Panda and the related ad quality improvements. Somehow, improvements in paid listing quality (and organic quality) led to a decrease in click-through rates (CTR) and cost per click (CPC).
Normally a company sees at least incremental improvements in revenue when it makes massive improvements in its product. But in Google’s case, this isn’t necessarily true. The greater the relevance for organic listings, the lower the CTR. Also, the greater the relevance of ads, the lower the CTR. Ads are run like a commodity exchange on an auction basis. CPC is driven entirely by the laws of supply and demand. If demand goes down, CPC goes down. Though product quality is necessary for Google to beat its competition, it does not necessarily improve its bottom line. I will explain why this happens in this article.
Analog: IBM.com Masthead
I mentioned in a previous article, Two Common Myths of Web Design and Information Architecture, that I helped lead the effort to replace the ibm.com masthead or site search function. This was the kind of massive overhaul to our site search function that we’ve seen in Google’s search algorithm since it released Panda. Two things happened almost overnight:
- Query volumes started increasing at a surprising rate as more customers gained confidence in the site search function. Over the course of one year, query volumes nearly doubled.
- Revenue for our own internal version of paid listings in our site search function decreased drastically—about half as much in the first month and a steady decrease since.
The Forbes article makes it clear that Google’s problem is analogous to what happened on ibm.com in the sense that improving the product reduced ad clicks. In ibm.com, that is a good thing. For Google, not so much.
We know from reams of research that users prefer organic listings to paid listings, all things considered. If organic listings give them what they need, they will click them rather than the paid listings. The only reason our paid listings got a lot of clicks before we updated the search engine was our organic listings were horrible. As soon as we returned good organic listings, users stopped clicking paid listings. That might be what is happening at Google.
Lower Bounce Rates
Consider the following scenario. A user clicks a paid listing on a search engine results page (SERP) and finds herself on a page that is not relevant to her query. Perhaps it is nominally relevant in the sense that the page she lands on is related to the query. But it is not what she is looking for. So she bounces off the page by hitting the Back button. Perhaps she then clicks another paid listing, and finds the page she lands on irrelevant. This happens several times before she finally finds a page that is relevant to her query.
This scenario happens frequently. Only a few years ago, in my conversations with agency folks who run our paid campaigns, it was common to have 80 percent bounce rates for search ads. I even had one agency person explain to me that 80 percent is the benchmark. Only when bounce rates are higher than 80 percent did they even bother to change the ad in some way, either in the “creative” (the link and the snippet of text) or in the landing page quality.
Google gets money for each one of these clicks, despite the fact that the search experience is not particularly good or that the company paying for the click is wasting their money. On the other hand, if a user finds what she is looking for on the first click, obviously Google only gets money for one click.
All things considered, Google gets more revenue the higher the bounce rates for search ads. Any improvements in ad quality reduce Google’s revenue. Despite this fact, Google is always improving ad quality. Ads with higher CTR and lower bounce rates get better Google quality scores, which tends to elevate them in search position. The higher the ad position, the better the CTR. Users then click on higher quality ads and have lower bounce rates as a result.
Still, higher organic and ad quality do not explain the decrease in CPC. It only explains a decrease in revenue from lower CTR. In the Forbes article, the Google executives cited CPC as the cause for the dip in revenue. Let’s explore why.
So why would CPC go down overall even as result quality improved? One explanation is better intelligence. Savvy search marketers know they won’t need to bid as high on words that have lower demand but nonetheless produce good yields. The science of bidding on words that will tend to have relatively high demand and relatively low competition is emerging. As it improves, CPC should naturally come down.
So, for example, at IBM, we now focus much of our paid budget on branded words. The reason is simple: A higher percentage of users who type branded words into Google are looking to buy products. If we build pages for the simple purpose to help them buy those products, we maximize our yields.
Now, we ought to own organic listings for our branded words as well. But our organic pages for our branded words will tend to be higher-level portals with a high percentage of learning, solving and comparing content on them. Purchase experiences are there as well, but they’re not front and center. Panda punishes pushy ad content. That’s why we also need to buy our branded words and point our paid listings at experiences specifically designed for the purchase phase.
Because branded words have less competition, we don’t have to pay as much per click for those words. This drives CPC down.
Our old search strategy was called Always On. According to this strategy, for all of our high-value non-branded words, we would need to be on page 1 in Google. We didn’t care so much if that meant paid or organic, as long as we showed up for those words. Typically, we would buy the word while we optimized the organic experience. Once we had first-page position for the word in the organic listing, we stopped buying the word. The trouble with Always On was two-fold:
- Non-branded words are relatively expensive and provide less value to the business in paid campaigns. Because the words have higher competition, it costs a lot more to bid against our competitors for these words.
- In our studies, up to 25 times more people convert from non-branded organic search referrals rather than paid. Because they’re not looking to buy yet when they type non-branded words, giving them purchase experiences too early in the buying cycle is just not a popular option for users. You still pay for the click, but you get a much lower yield for relatively high cost. Remember those high bounce rates? Those were primarily the non-branded words.
All things considered, it is better to focus our efforts on building better organic experiences in the first phases of the buying cycle—i.e., with prospects who are not as familiar with our offerings and need to do their due diligence before entering the purchase phase. It might cost a bit more initially, but it results in much higher engagement rates for our first–page listings. The new strategy is to lead with organic, and only use paid to boost the branding of organic listings. The only exception is in the purchase phase, where paid is the primary way we engage with customers.
As I explained in a blog post at my blog writingfordigital.com, Google’s own studies show that B2B tech users tend to click organic listings at a higher rate when they see a paid listing pointing to the same URL. We don’t do this for all our non-branded words, but we do buy words when we already have an organic listing for that page. Because users tend to click the organic listing rather than the paid in this case, we still reduce CPC for those words despite paying a high price per click.
Abandoning Always On led to a much lower CPC for IBM. This kind of analysis is available to every company through improved analytics. And Google is one of the leading providers of search analytics solutions. Improving those products drastically in the last few years has actually hurt their own core business by helping their customers reduce waste and increase yield. When companies reduce click waste, as IBM did with abandoning Always On, it drove down CPC.
Why does Google focus so much of its attention on things outside of its core business in its external communications? Because it is difficult to grow the core business while continuously improving the product. Google needs to improve the product to stay on top of competitors like Bing, so it can’t fail to continuously improve. But continuous improvement actually decreases CTR and CPC, the leading indicators of its revenues and profits, respectively. The main way it will grow, then, is by growing its emerging portfolio. That’s why we see so much about Android adoption and Google + sign ups on analyst reviews.
James Mathewson is IBM’s global search strategy lead and author of Audience, Relevance and Search, Targeting Web Audiences with Relevant Content. The opinions expressed here are his own and not IBM’s.