Frederick Vallaeys Interviewed by Eric Enge
Published: April 28, 2010
Frederick Vallaeys is a Product Evangelist for Google AdWords. In this role, he helps advertisers learn about which Google products can best solve their marketing needs. He also represents the needs of advertisers with the engineering and product management teams. His main product focus is on ads quality and bulk tools like the AdWords Editor and the AdWords API.
Prior to Google, Frederick was an engineer at Sapient and a part-time wedding photographer who found new customers through AdWords. He joined Google in 2002 to help bring AdWords to the Dutch and Belgian markets. He earned his B.S. degree in electrical engineering from Stanford University in 2000.
Eric Enge: Can you start us off with an overview of the new types of ad formats you've added?
Frederick Vallaeys: Because we think ads should provide users with great information, just like organic results, we decided that it is sometimes the right thing to do to include an image, a map or a video in an ad, because sometimes that is a better piece of information than a simple text ad.
If you do a search for a movie, instead of just getting a text ad, now you actually get a little video clip, and you can play that video right there inside of the ad unit.
Say that you are going out and shopping for Mother's Day, looking for a handbag for your mom. Wouldn't it be great if you could see product images and product prices right there, so that before even choosing which merchant's ad you are going to click on, you could see the range of products that they have, the brands that they have, and their pricing? That is the type of information that we call ad extensions, which are essentially pieces of information that we tag on to the ad text because it makes it more useful. (For an example of this, try searching on Toy Story action figures).
Another new ad format is Ads Sitelinks. This is very similar to sitelinks on the organic results. Typically, a top organic result doesn't just have that one link to the homepage of that website, but it has four or five options, allowing the user to choose where to go, like directly to the gift registry, or customer support pages.
What we've done is taken that concept and made it available to advertisers through AdWords. The advertiser gives us ten links that they think most of their customers would consider going to, and Google then automatically figures out which four links to show for each specific query where that ad format is supported.
You typically see that in an ad that's on top of a page like on a query for a company’s name. If you do a search for eBay, instead of just seeing that little text ad that says, please come visit eBay, you also have four other links that will take you directly to where you want to be on eBay’s website. (for another example of this, search on Trip advisor.
The other approaches we've taken is to create completely new ad formats with new pricing models besides cost-per-click, and more interactive ad formats that help consumers who are looking for more complex products.
A good example of this is Comparison Ads, which we use specifically for mortgages and refinancing. In the past, if people did a query for refinance or mortgage, there were a lot of open questions, as Google didn’t know what that person’s credit score was or how expensive their house was. The ads that we would have shown people in the past would have been useful in terms of showing them mortgage lenders, but they weren’t that useful because those people would then have to call each lender and tell them how much they were looking to borrow.
Users can now tell us what their credit score is and how much their house costs, and we’ll return mortgage rates immediately. We get this data back very quickly, and we give people a whole bunch of lenders that they can then compare against each other.
The thing that is really compelling from an advertiser’s perspective is that these leads have been prequalified because the consumer has already provided quite a bit of information about their needs. By the time that they come to these advertisers, they are most likely looking exactly for the product that that advertiser or company offers. This creates more valuable leads for these advertisers.
This new system is pretty appealing from the consumer perspective as well, because in the past a lot of their information got resold. If people went to a mortgage lender that didn't have exactly what they were looking for, that lender might have sold their information to a bunch of other lenders, and then these users could be getting calls from different lenders every night.
With Comparison Ads, Google actually anonymizes phone numbers and contact information. Once the customer gets his or her mortgage, they stop getting calls about it, and their information hasn't been sold to a whole bunch of other lenders. We believe this is pretty compelling from both the consumer and advertiser perspective.
We also have a new ad format called Product Listing Ads using data that comes from the Google Affiliate Network. What we do for every query is automatically determine if there is some product inside the Google Affiliate Network that appears to be a good match. If there is, then we go and show an ad for that. What’s nice for the advertiser is that they only pay us on a cost-per-acquisition basis. Instead of paying per click regardless if that click actually converts, they only pay us when somebody goes and purchases the specific product. It’s pretty much as close to a no-risk advertising system as you can get from an advertiser perspective.
Eric Enge: To back up a little bit, you have essentially set up something to communicate directly with mortgage vendors and generate leads for them?
Frederick Vallaeys: Yes, exactly. The consumer has two options for communicating with lenders; they can just pick up the phone and call the number that's right there, or they could use the form, fill it out, and send it to the lending company. In the case of the form, we then send that on to that merchant. When we give them that information, however, the phone number that they get is not the customer’s phone number, it's an intermediate phone number so once the customer finds a mortgage lender, they will no longer get calls.
Eric Enge: Do the mortgage lenders pay on a pay–per-lead basis?
Frederick Vallaeys: Yes, exactly. They pay either when they get a phone call or when somebody fills out a contact form.
Eric Enge: So it's a CPA model deal, it just happens to be mapping into a comparison tool? It’s essentially a Google Affiliate Network?
Frederick Vallaeys: Yes, it's similar in that advertisers are charged every time a lead is generated. We work with a limited set of mortgage brokers right now, and they use the AdWords system to place their ads. Comparison Ads and Product Listing Ads are both in limited beta in the United States. These are much more experimental than location extensions and Ads Sitelinks, where we already know how we want to proceed forward.
Eric Enge: If we dig into the affiliate network a bit more, what kind of businesses would participate in that?
Frederick Vallaeys: Product Listing Ads are mostly used by retailers. If you have any sort of product you are trying to sell to consumers, this is a great ad format to help drive sales.
Eric Enge: So people who are remarketing products? They might be selling golf clubs, for example. Not the manufacturers, but the retailers?
Frederick Vallaeys: Yes, exactly. Usually not the manufacturer but a reseller, the type of company you’d expect to see in an affiliate network. It could be eBay, Amazon, any other online retailer, or it could also be a brick and mortar store that has an online shopping system.
Eric Enge: Can businesses looking for leads participate in this model (i.e. that derive revenue on the lead as opposed to the sale of a product)?
Frederick Vallaeys: Not at this point in time. What we are trying to do here is really show products, and that goes back to the whole notion of wanting to think about how to deliver the best piece of information to each user that comes to Google. When they search for a product, it's very clear to us what they are looking for.
When it comes to leads, I think it's a bit more complicated, and that's where you are seeing something like what we do with Comparison Ads, where we actually have to ask users for a bit more information upfront before we can go and show the best results. In that way, those two things will be treated somewhat differently, just by nature of how well we can know ahead of time what it is users are looking for.
Eric Enge: The Ads Sitelinks only relate to a brand name of a business, and that matches up with a website for that business, correct?
Frederick Vallaeys: Yes. The example that I gave was very specific to brand names, but it can actually be any keyword for an advertiser. Because this is only shown in the top ad unit, it usually is a brand, as that's typically where an advertiser has the top position.
Eric Enge: So a top position ad has an opportunity to get a set of sitelinks?
Frederick Vallaeys: Exactly.
Eric Enge: Are there any other limitations or criteria?
Frederick Vallaeys: They have to be the top ad and meet certain quality criteria. The way we think about this is much like the way we think about sitelinks for organic results. There has to be a strong indication to us that you've already made up your mind as to what site that you want to go to. Within that site, we then give you a couple of options.
That’s why a brand is the most likely to show these sitelinks . If someone types in eBay, they are most likely trying to get to the eBay website, but if it is a more dubious query like ski jackets, it's not as clear to us as to where they actually want to go. You can buy ski jackets from North Face, but they can also be found on Amazon and other websites like that.
Eric Enge: Right, it mostly relates to a level of certainty about the whole business?
Frederick Vallaeys: Exactly, and the searcher's intent.
Eric Enge: I think you also mentioned something about local extensions.
Frederick Vallaeys: Yes. That's a great one if you operate a franchise business with multiple locations across the country. You can put in all of your locations, and then if a user comes to us and does a local search for Pizza Hut, for example, we would show them all the Pizza Hut locations around them, based on where they are located. They could also drag the map, and see the different locations. Say someone traveled to a city they haven't been to before but wanted to eat somewhere they were familiar with, we can show them any of those locations, regardless of where they are.
Eric Enge: That’s similar to what web search has been doing for a while.
Frederick Vallaeys: Yes, exactly, you hit the nail right on the head. We see a lot of these behaviors on the organic side of the business, and some of them make a lot of sense to do on the advertising side as well. We do an experiment, and if it turns out to be the case that it's a good user experience, then we go and roll that out.
Eric Enge: There is a lot of opportunity to do the experiment there, that's for sure.
Frederick Vallaeys: Exactly. And our experiments are not limited to just big advertisers. A mom and pop shop, could participate equally well, even with just one location. Another way that we think about many of these product launches is that we don't want to put anything out there that's restricted to a limited list of advertisers. We want to show the ads that provide the most relevant information, regardless of the size of the advertiser. We want to surface the best information to the user, because that leads to the best experience.
Eric Enge: Let’s talk about some tricks and tips.
Frederick Vallaeys: A lot of advertisers that I speak to say that it's fantastic how many product launches AdWords does, but it can be a little overwhelming. Sometimes it's difficult to know what and where the biggest opportunities are for an advertiser, so I picked three product areas or features that I think most advertisers could get some significant benefit out of. These are the Google Content Network, the Conversion Optimizer, and the Search-Based Keyword Tool.
Eric Enge: Let’s start with Content Network.
Frederick Vallaeys: The Content Network has a bit of a bad perception to some people, and that's not completely unjustified because of the way that the product was seven years ago when it was first introduced. We certainly have come a long way with it, and many concerns that advertisers had along the way have been put to rest. We actually did a study last year and looked at how advertisers using the Google Content Network are faring, and we found that those using the network would typically get about 20% of all of their leads and conversions from the content network.
We also found that they typically did that at a cost that was slightly lower than what they were paying on their search ads. There is actually a lot of traffic and there are a lot of conversions they are getting from the content network, and they are doing it at a cost that's very similar to what they’re used to paying for search ads.
Eric Enge: My own opinion is that the big reason why the content network got a bad reputation was because the default way of turning it on was by using the same keyword set as you use for web search.
Frederick Vallaeys: Yes, exactly.
Eric Enge: The way keywords are used is quite different. In the content network, what happens is that the keywords are basically used as hints of what to look for on a webpage. You are scanning those pages to see how that matches up, as opposed to the way it works in regular keyword search.
Frederick Vallaeys: Exactly. There is a difference in user intent when people are doing a search, and very specifically saying what they need right now, versus when they are just pursuing other interests like doing research for an upcoming purchase. With these two examples, the way advertisers chose keywords is very different, and also the way that you write ad text is very different as well.
You always want to have a must-click ad, but a must-click ad when somebody is searching is all about being relevant. An advertiser may have exactly what someone is looking for because they know what the user just typed into the search box, but on the content network, advertisers don't know exactly what it is users are trying to do. The ads can be a little more captivating in a way. They are not exactly the same type of ad texts that work well on search and on content.
Eric Enge: Right, and that's because the user who is visiting the particular webpage has a different motivation then someone who has entered a specific search term related to your product.
Frederick Vallaeys: Yes, exactly.
Eric Enge: So you have to try to pull them there and attract their attention, whereas in search it's a different messaging approach.
Frederick Vallaeys: Exactly but aside from having a different campaign for search and content which is an option advertisers always had, there are other things we’ve done to improve performance. Recently we've introduced features like view-through conversions which is particularly useful for brand advertisers. You want to run some display ads and you want to see if those display ads actually help lift the conversions that you are getting on your search campaign, so now we can actually measure and report back on that. The notion is very simple; you show a banner ad through the Google Content Network, somebody might see that ad a few times, and they are becoming aware of your brand, even though they are not clicking on it.
Now they go and do a search about that particular product category, they see your search ad, and they want to check it out because they are familiar with the brand. They actually go and do the conversion, so now we can report on which of those display ads actually participated in that whole process. That's pretty powerful from a display advertiser perspective.
Eric Enge: Attributing some credit to the ad in that situation is just something you need to do because otherwise you will undervalue the impact of the campaign.
Frederick Vallaeys: Exactly.
Eric Enge: Can we talk a little bit about using Conversion Optimizer to manage your bids to a target CPA?
Frederick Vallaeys: As far as the Conversion Optimizer goes, what's really powerful about this for advertisers is the fact that it's free to use, and it targets bids to a maximum cost-per-acquisition. It essentially removes the hassle of doing bid management from the advertiser’s perspective. Now they can focus on different aspects of the business that might be done better by humans as opposed to computers.
The other thing that makes the Conversion Optimizer really powerful is that Google has access to a ton of data about each specific situation when a keyword triggers an ad. As opposed to a more conventional bid management system that has to make generalizations and place one bid for a keyword, the Google Conversion Optimizer actually adjusts your bid for every single auction.
If a specific keyword does really well in conjunction with some other words in the query, or if it does really well at a specific time of day or in a specific part of the country, we can then modify the bid up or down to make sure that you hit your target cost-per-acquisition. This is powerful because it’s a level of bid management that leverages all of the data that Google has at the time of the query.
Eric Enge: It sounds like what you are doing is leveraging the data across all the other advertisers in that category.
Frederic Vallaeys: Well, not necessarily. We don't necessarily look at other advertisers in that category, it’s more that we can look at every query on a case-by-case basis. So we could see an advertiser performing really well for a certain variation of their broad match keyword, but not so well on a different one. A good example here is if you are selling Alaskan Cruises, but you run on the broad match keyword "cruises". Say that 90% of people who search for cruises are actually looking to go to a warm place, like Hawaii or the Caribbean.
In the past, you probably would not have done very well on that broad match keyword because most users who clicked your ad didn’t want to cruise in Alaska. What the Conversion Optimizer can do is figure out your ad converts well on searches for Alaskan cruises and not for Hawaiian cruises. Therefore, it can bid up on the good queries and down on all of the others, and actually still give you a pretty good value from the limited set of queries that are most relevant to your product.
The average results that we see are a 21% increase in conversions, and about a 14% decrease in cost-per-acquisition. That’s the average, so others have seen better results and some have had worse results, but most advertisers do see a positive result.
It's really easy to turn it on and off so I recommend that advertisers play with this. The cost-per-acquisition is at the ad group level, so even if you have very different products that you are selling with different cost-per-acquisition goals, you can still do well because you can set a different bid for each ad group.
Eric Enge: Do you have the potential to allow people do this on an ROI level?
Frederic Vallaeys: We don't have the ROI level yet. If you want to bid a percent of the shopping basket value as a retailer, we don't have that capability. However if you have an ad group that's promoting roses and a different ad group that's promoting vases, you could take your average profit for every vase sold and use that for the max CPA bid for the vases ad group. For the ad group about roses, you could do the same thing, and then you could set drastically different maximum cost-per-acquisitions.
Eric Enge: So let's talk about the search-based keyword tool.
Frederick Vallaeys: This tool looks for keywords for which you have a result in the organic index and it tells you which keywords you don’t have in your AdWords account along with the landing pages from the organic index.
It essentially identifies missed opportunities, where if a company has something available that people are actually searching for with no ad result showing up, it will tell them. Because this tool gives you back the list of keywords along with the landing pages, it's very easy to group them into good ad groups. All the keywords with “roses” landing pages go into a “roses” ad group, and all of the “tulips” landing pages go into a “tulips” ad group. The reason this is so powerful is because it's really quick to run the tool. It spits out the data in a matter of seconds, you can filter through it, and then you accept some of those changes into new ad groups. We’ve seen some pretty dramatic increases in the number of clicks that customers get when using this.
It’s also cool because it includes the impression share metric, so it will even give you some insight into how many clicks you are already getting on that keyword from broad matches, along with how many new and fresh opportunities exist that aren’t currently taken advantage of.
Eric Enge: When you dive into the detailed reports, you'll see metrics that will give you feedback on impression share. The idea being that the keywords you bid along are relevant to 1,000 search queries, and you were shown on 721, which equals a 72.1% impression share rate. What are the things that affect impression share, because I think that plays into what you just eluded to with the Search-Based Keyword Tool?
Frederick Vallaeys: There are two things that affect your impression share, and we actually bring those out in reports. Impression shares are lost due to budgets, and impressions are lost due to rank. When it comes to budgets, it's straight up how much your daily budget has to be to capture all the users who are clicking on ads for these keywords. You can imagine on a keyword like flowers or cars that the daily budget recommendations are very high. If you are a smaller business or someone just getting into the space, maybe you would set your budget lower so that you can test the waters while limiting your risk. In that case, you are probably going to lose some impression share due to the budget of your campaign but that is okay. Once you are comfortable with the campaign, raise your daily budget to the recommended amount to increase your impression share.
When it comes to rank, if you are being outranked by other advertisers, the solution is either to increase your cost-per-click or to do an optimization of those keywords to improve your Quality Score. Maybe you’ll find some more specific keywords, so instead of having a broad keyword, get more specific. By breaking up keywords into different ad groups, you can have a more appealing ad text that more clearly states why your particular offering is really the best one for that specific set of keywords. That can also help lift up your rank.
Eric Enge: So when I see an ad that has an average position of three and an impression share of 85%, clearly it isn't the bid amount. It must have to do with the fact that the impression share is calculated on things of varying relevance, correct?
Frederick Vallaeys: Right. Even if you have an average ad rank of 3, there could be certain queries where your keyword is a match but your ad is actually not that relevant so maybe you have an ad rank of 17 in that specific instance, so your ad didn't get shown because it was not on the first page. That would have been a lost impression.
Eric Enge: So that just confirms that the impression share calculation does in fact use the keywords, and all the data in which the keyword matches, even if the relevance on some of those is low.
Frederick Vallaeys: Right, and we also have a metric, for exact match impression share which does not include those variations.
Eric Enge: So if you are bidding on your own brand with no budget limitations, you put your money behind it, and you do an exact match impression share, you presumably be quite close to a hundred percent?
Frederick Vallaeys: Right.
Eric Enge: Are there ever times when it's useful for keywords to bid on it in multiple match types?
Frederick Vallaeys: Advertisers sometimes add the same keywords in multiple match types for the wrong reasons. They often assume that they'll get a better Quality Score by doing this, and that is not the case because the quality score that we assign to a keyword is for the exact match variation.
Eric Enge: So even for a broad match version of a keyword, the calculation of the quality score is based on that keyword as an exact match.
Frederick Vallaeys: Exactly. As an advertiser if you add all three match types you will see the same quality score number in the system because we just calculate it based on is the data for the exact match.
A valid reason for adding multiple match types, like you eluded to, is when you know that you have different conversion rates on the different match types. If that is the case you could set different bids, or you could even have the match types broken out into different ad groups and have an ad text that's a little bit different.
Again, it goes back to the notion of what user intent is. If they typed in exactly the keywords that you have and your keyword is an exact match, then you probably want to tell them about exactly that. But if they came in on a broad match, if they were searching for digital cameras but were matched on a more specific keyword like Canon EOS Rebel, maybe you want to show them an ad text that says, ‘we have many different types of cameras from all different brands,’ because the query wasn't as specific to what you had as your keyword.
I think varying ad text and bids are two good reasons to use multiple match types. With that said, if you used the conversion optimizer then that whole bid aspect would go away and we would handle it for you.
Eric Enge: So for those with a large campaign, they might only do this for larger volume keywords. It's not worth the trouble with the occasional click?
Frederick Vallaeys: Yes, people should pick their battles, and that’s sort of why I am trying to lay out the tools that I think are most useful and quickest hits for making improvements. If you can focus on the things that humans do really well, like marketing messages and making sure ads are relevant and include information that users are searching for, then that's a great way to spend your time, whereas bid management may be better left to machines. The whole question of how deep into the long tail to go, and how many keywords to manage, that's a decision that needs to be made individually by every company. There certainly are limitations to the long tail where keywords can become too expensive to manage. It might not be worth it anymore, and you are better off using broad match keywords and some of the other tools that Google has without wasting a lot of your resources managing it.
Eric Enge: Thanks Frederick!
Frederick Vallaeys: Yes, thanks a lot Eric!
Have comments or want to discuss? You can comment on the Frederick Vallaeys interview here.
Other Recent Interviews
- InfoGroup's Pankaj Mathur, April 5, 2010
- Matt Cutts, March 14, 2010
- Compendium Blogware's Chris Baggott, February 7, 2010
- AudetteMedia's Adam Audette, January 30, 2010
- Tribune's Brent Payne, December 13, 2009
- Google's Josh Cohen, November 15, 2009
- KeyRelevance's Chris Silver Smith, Sep 17, 2009
- Oyster's Eytan Seidman, August 6, 2009
- Google's Peter Linsley, July 12, 2009
- Yahoo's Dennis Mortensen, June 15, 2009
- Enquisite's Richard Zwicky, June 8, 2009
- Microsoft's Scott Prevost, May 24, 2009
- YouTube's Tracy Chan, Matthew Liu, May 18, 2009
- Google's John Mueller, May 14, 2009
About the Author
Eric Enge is the Founder and President of Stone Temple Consulting (STC). STC offers Internet marketing optimization services, including SEO, Social Media and PPC optimization, and its web site can be found at: http://www.stonetemple.com.