New AdWords Estimated Cross-Device Conversions

A couple of months after the paid search world had to transition to AdWords Enhanced Campaigns in the name of simplified and more relevant cross-device ad management, we are finally getting some initial food for thought with regard to cross-device performance. I was calling it the “next frontier for online marketers” in a previous post, and we are now officially getting there!

A New “Estimated Total Conversion” Column Is Rolling Out

In AdWords, a new column called “Est. total conv.” was made available for some beta accounts recently. However, this column doesn’t seem to be available at the keyword level yet — only at the campaign and ad group levels.

The official field description is as follows: “This is our best estimate of the total number of conversions that AdWords drives for your business. What it means: Est. total conv. = Conv. column + Est. cross-device conv. column”

The description doesn’t say whether the “Conv. Column” field is referring to “Conv. (1-per-click)” or “Conv. (many-per-click).” However, I’d tend to think Google is using the “Conv. (many-per-click)” column for reference.

Also, Google plans to factor in other conversion types like phone calls and in-store visits in the future — not right now, though.

Leveraging The New “Estimated Total Conversion” Column

There are lots of different ways to look into this new column — let me just suggest two of them:

1. Since the “Est. cross-device conv.” column is not provided, one can easily calculate it such as Est. cross-device conv. = Est. total conv. – Conv. (many-per-click). This is assuming Google is using the Conv. (many-per-click) column rather than Conv. (1-per-click).
2. Analyzing a new ratio which could be called something like “Cross-device assist %” such as: Cross-device assist % = Est. total conv. / Conv. (many-per-click) – 1. The higher the ratio, the more cross-device conversions occurred as a percentage. A ratio of 0% means there were no additional cross-device conversions, while anything greater than 0% indicates some cross-device assists.

In the below paid search report by ad group and device, those rows with the highest cross-device assist % correspond to assists from mobile devices. That was expected; however, now we have some hard numbers.

Another way to look at it is strictly by device, such as in the below pivot table. I noticed a couple of odd things:

• The numbers in a report differ from the numbers from the AdWords interface by roughly 5-10%.
• The total number of estimated total conversion is lower than the total number of last-click conversions.
• Those additional conversions attributed to mobile devices are somehow subtracted from other devices. As a result, this advertiser’s true mobile CPA is 22% lower than it seems, looking at just the last-click, while the true desktop CPA is 5% higher and the true tablet CPA is 15% higher. You’ll definitely want to take this into consideration when adjusting your desktop/tablet bids, as well as your mobile bid modifiers.

Conclusion

With the new “Estimated total conversion” column, Google is attempting to tackle the increasingly complex issue of cross-device revenue attribution in paid search — and it is definitely a leap forward.
However, in order to make it fully trustworthy and actionable for online marketers, it’d be great to get more transparency from Google regarding the way the numbers are put together (with some kind of confidence interval?), then have the Est. total conv. column available at the keyword level for more granular insights — and ultimately across multiple channels for more sophisticated revenue attribution modeling.

Estimated Total Conversions: New insights for the multi-screen world

People are constantly connected, using multiple devices throughout the day to shop, communicate and stay entertained.  A September 2013 study of multi-device consumers found that over 90% move sequentially between several screens for everyday activities like booking a hotel or shopping for electronics.

As consumers are increasingly on the go and switching between devices, marketers are telling us they want to see a more complete and accurate picture of how their online advertising drives conversions.  Conversions can come in many forms: visits to stores, phone calls, app downloads, website sales or purchases made after consulting various devices.  Getting better insight into these complex purchase paths can help you optimize your online advertising and allocate budget more effectively.

Introducing Estimated Total Conversions

Today, we are introducing Estimated Total Conversions for search ads on Google.com.  This is an exciting first step to give marketers more insight into how AdWords drives conversions for your business by showing you both the conversions you see today, like online sales, as well as an estimate of conversions that take multiple devices to complete.  Over time, we’ll be adding other conversion types like phone calls and store visits as well as conversions from ads on our search and display network.

Estimated Total Conversions will provide you with a holistic view of all of the conversions driven by your Google search advertising that can be used to make important decisions like how much to bid and how to assign budget across your various marketing channels.  For the last few years, many sophisticated advertisers have been using their own analysis to get to these insights.  Today, we are beginning to bring this level of insight and measurement to all advertisers.

Estimated cross-device conversions

Estimated cross-device conversions is the first new conversion type to launch as part of Estimated Total Conversions. Cross-device conversions start as a click on a search ad on Google.com on one device and end as a conversion on another device (or in a different web browser on the same device).

For example, say someone shops for “blue jeans” on her mobile phone while waiting for the morning train.  She clicks on a mobile ad for ABC Blue Jeans.  When she gets to her office, she goes directly to the ABC website to make a purchase.  This is an example of a cross-device conversion.  We calculate cross-device conversions using a sample of data from users who signed into multiple devices. 

Estimated cross-device conversions will begin rolling out globally to all AdWords advertisers starting today and continuing over the next few weeks. To see these new statistics, you’ll need AdWords conversion tracking and a sufficient volume of conversions on which to base a reliable estimate.

In the last few months, we’ve analyzed data across thousands of AdWords advertisers to learn more about cross-device conversion patterns.

When advertisers in the travel industry use AdWords estimated cross-device conversions, they are able to measure 8% more conversions, on average, than they did before.  In addition, they can now measure 33% more conversions that originated on a mobile phone and later converted on different device.  This helps them attribute all those sales -- from customers who searched for flights and hotels on their mobile phones and then made a purchase from another device -- to the right ad.

Similarly, other verticals, like entertainment and retail are also seeing positive results.  Businesses in these industries are now able to measure 12% and 7% more conversions, respectively, than they could before using Estimated Total Conversions.

Sean Singleton, Marketing Manager at American Apparel noted that, "We always knew our online ad investment was influencing conversions across devices, but we didn't know how to begin estimating these numbers. Once we saw that 5.3% more conversions could be attributed to cross-device conversions in AdWords, we knew we could more accurately calculate the value we were receiving from each ad click.  We also learned that mobile ads are driving 16% more conversions than we thought, so we are now investing more into this channel to gain more sales.”

More results from other verticals can be seen below.

Paving the way for marketers to measure the full value of their online advertising

We are committed to helping you gain insight into the new conversion types that are part of a constantly connected, multi-screen world so that you can make the best advertising decisions possible. In addition to cross-device conversions, both phone calls and store visits will be included as part of Estimated Total Conversions in the coming months. These are important conversions to consider — people make more than 40M calls to businesses each month directly from Google ads and are often looking for physical store locations when they search on Google, particularly on the go.

The PPC Experiment You Never Dare Run

A question that PPC account managers frequently have to deal with is, “Why are we paying for this traffic? Aren’t we going to get that traffic anyway?”
It’s a fair question, even if it is completely annoying to hear for the twentieth time by the twentieth new accounting manager you’ve had to break in. No matter what data you present, no matter how perfectly your charts demonstrate perfect, positive correlation between ad spend, revenues and profits, they never seem satisfied with your answer.
“Fine,” you say. “Let’s try an experiment. We call it the PPC nuclear option.”
“The nuclear option?” the accountant gasps. ”What’s that?”
“Well,” you continue obligingly, “We take all our PPC ad campaigns offline for a few months, and see what the true impact on our bottom line looks like. I just need you to sign off on it here….”
That usually ends the discussion — at least until next month’s AdWords and Bing Ads bills come due.

The PPC Nuclear Option — For Real!

I won’t bore you with the juicy details of how it came about (unless you promise to buy me a drink next time we meet), but the long and short of it is that I am now in the middle of a real-life PPC nuclear option experiment.
Campaigns that had been running for a few years were taken offline abruptly three months ago, and we are just about to put them back online. The website is now relying completely on referral and organic search traffic. The website itself and all downstream processes have remained virtually unchanged.
I thought it would be interesting to make a few simple observations at this point in the experiment. I’ll save deeper analysis for a later date, after we’ve put the campaigns back online.
Observation #1: Greater Than Expected Decline In Keyword Conversion Performance

Keyword conversion performance  before and after the “PPC Nuclear Option” was detonated.

Before we detonated the nuclear option, we were aware we’d lose substantial traffic to the site because about 70% of our traffic was coming from paid traffic sources. What we didn’t know was how our brand keywords or our highest ranking keywords (which were also part of the URL) would fare.
In the chart above, I’ve summarized the before and after conversion volumes for most important keywords that have driven conversions and was surprised by some of the results. We looked at our brand keywords, keywords that were prominent on the site and which appeared in our website URL, as well as our most productive non-brand keywords.
Branded Keywords - Our client’s brand is not a household name and there’s not a lot of search volume associated with it, so our brand keywords have never been our largest source of search conversions. They do, however, rank highly and so we were surprised to see conversion volume on them drop by 28%. We figured, based on earlier studies on brand keyword cannibalization, that we’d see only 10-15% drop in conversion volume for our high-ranking brand terms.
What would account for a greater than expected decline in brand term performance? I have a hunch that our Google GDN display ad campaigns may have been providing a lift while they were running, but since we are not a famous brand, that halo effect only lasted while our display ad campaigns were running. If my theory is correct, we should expect to observe an uplift once we start advertising again. Stay tuned.
URL Keywords – Our next best performers were keywords that literally described our products and which appeared in our website URL. These keywords have generally been at or near the top of the organic search listings, and they perform more like brand terms than generic keywords. We were most surprised that these would have dropped by a whopping 59%.
Non-Brand Keywords – We were also astounded at how deeply the PPC nuclear option demolished the performance of our best non-brand keywords, which dropped pretty near to zero.
Clearly, this client is too dependent on paid search on its most important terms.
Observation #2: Organic Traffic Needs To Contribute More Search Volume

After taking PPC campaigns offline, online organic and referral traffic remain.

This Google Analytics graph shows all search traffic from 2009 through 2013, and to my last point, demonstrates a highly unbalanced mix of online traffic sources, since 75% of their visitor traffic dropped when the nuclear option was detonated. I don’t know what the perfect ratios should be for organic to paid search traffic, but in general, I think most of us would agree that, over time, the mix should start to skew toward more unpaid sources of traffic and rely less on paid traffic.
Of course, figuring this out didn’t require taking the nuclear option because the data has been there the whole time. It does certainly lay bare the situation unambiguously, however. The accounting manager will love this data, and so will the SEO team (that doesn’t exist at the moment) because it shows how even modest investments in SEO can be justified financially.
Observation #3: Friendly “Ghost Clicks”

Google Analytics reports a small amount of PPC ad “ghost clicks” even after taking campaigns offline.

Even two months after taking the campaigns offline, we are still getting visitors from “ghost clicks.” Most ghost clicks happen when a visitor types your URL into their browser after first clicking on your ad. When they start typing the URL, their browser cache types ahead for them, showing them your URL along with your original URL tracking variables. We observed about 4% of our visitors coming in this way, but they represent only about 1% of the original ad click volume.
I don’t know what a good “ghost click” ratio might be, but I’d like it to be higher because that means that our site was so memorable that people typed it into their browser rather than doing a new search.
Observation #4: Unfriendly “Ghost Clicks”
I’ve observed that not all PPC Ad ghost clicks are friendly. In some cases, they may be competitors doing deep dives on your site for whatever nefarious purposes they have in mind. Here’s my evidence of unfriendly ghost clicks:

Hmmm… Lots of visits all of a sudden from PPC ad ghost clicks. Does not look like a friendly ghost to me.

Why would we, all of a sudden, get hundreds of PPC ad ghost clicks when our campaigns are clearly turned off? My first guess was that someone accidentally turned them on; but no, that wasn’t it. I dug one level deeper in Google Analytics and found that all the clicks were coming, on different days, from different locations.

Who clicks from Beverly Hills, Little Ferry and Trinidad? Jet setter, perhaps?

Using my best TV detective deduction skills to develop a profile of the perp, I see that the clicks come from Beverly Hills, CA and Little Ferry, NJ. Both are pretty affluent suburbs. I also see clicks from Trinidad. Could this be a rich jet-setting venture capitalist trying to discover our secret sauce? If so, I’ll bet they are also reading my column; so to you I say, “I know who you are, and I saw what you did.”
Observation #5: Google Analytics’ Accounting For Paid Search Clicks Is Excellent
I had a weird observation when looking at my paid search clicks versus my analytics data. Google Analytics was reporting 7% more clicks than we actually paid for from all our paid traffic sources. I know that Analytics and AdWords account for clicks differently and that Google Analytics reports clicks (suspicious or otherwise), but I never took notice of how great the difference was.
I ran an AdWords invalid clicks report and subtracted those from my Google Analytics total, and found that the number of clicks from our paid search campaigns matched up almost perfectly (99.1% match) with the clicks Google Analytics reported receiving.
An added surprise was my discovery that Google AdWords continues to credit invalid clicks to the account even after turning off all spend. It wasn’t a large amount, but it was a very pleasant surprise nonetheless to see additions to our account balances.

Questions About The PPC Nuclear Option?

I’ve touched on just a few of my own casual observations from this unplanned PPC nuclear option experiment, but there are so many other observations to be made as we bring this campaign back to life. If you are curious about any part of our experiment, and have questions of your own, please leave a comment below, and I’ll see if we can get an answer for you.