• Stop being a LURKER - join our dealer community and get involved. Sign up and start a conversation.

Used Vehicle SEM Comments and Best Practices

Feb 12, 2011
1
0
First Name
Scott
What are your experiences with programs like Haystack or competitors that automate PPC campaigns for used vehicles? We use Haystack and we are happy with it, but are curious about how to understand, use and manage it more effectively as we believe it is a great way to expand the digital exposure of our inventory.
 
Last edited:
What are your experiences with programs like Haystack or competitors that automate PPC campaigns for used vehicles?

Scott G,

If you have time, open an Adwords account and run a small campaign yourself. You can set the daily spending to any dollar amount you like, Limit it to $10 a day if you want. Its not hard at all and it's a great way to learn.

how to create a google adwords account - Google Search
 
Haystack seems to be a simple API with the engines. You feed Google your inventory and it creates ads accordingly. It's not new science, and I believe Dealer.com has a program similar.

It targets the long tail ("2005 GMC Sierra"). However, if you know how to buy Paid Search ads, there is a method to purchasing terms on a "broad match" or "phrase match". This means the engines pick up certain terms or a combination of terms within your ad, so you can show up for "2005 GMC Sierra" by just purchasing the term "sierra" or "gmc sierra", or any combination of the sort.

The other argument is that Google searchers are not looking for individual vehicles on Google (yet), so the long-tail targets are not going to yield you much traffic.
 
The primary rationale for us implementing the dynamic inventory for used vehicles is to manage a couple of goals.

One, we want to ensure that we have highly relevant ad copy. That means that the ad is not a generic used vehicle ad, but one that speaks to the searcher and reassures them that you have what they are looking for.

Second it's in effective use of ad dollars. It may or may not be everyone's strategy to ensure that any vehicles that you spend advertising dollars on are available for you to sell, but it certainly does help user experience to have a solid end to end experience.

When I say end to end experience, I'm seeing this chain of events ...
1) user searched on Google ->
2) relevant and compelling Google ad referencing searchers interest ->
3) click action to a page on site that is highly relevant to the user

When we built this, we wanted to bring the user directly to the exact vehicle in stock. If a searcher looks for a "Sierra", we can see two poor experience scenarios without dynamic inventory.

First scenario: you ignore the specific search and broad blanket ads with general used vehicle terms. That would either imply a really generic ad OR dynamic keyword insertion. With dynamic keyword insertion the user would feel tricked into your site because you may not have that vehicle. With a generic ad, you really can't deep link to that specific vehicle and the end user feels it's an overly generic ad when they read it.

Second scenario: You have the specific search always active, but when the vehicle is not in stock, you take them either to a "no vehicles in stock" page, or a generic page where their experience suffers due to lack of relevant content. They get a feeling of "You told me you had this on Google, and now you don't ... I won't trust any of your ads from here on out".

Both of those scenarios, when you are paying for that click, are bad experiences for the end user and a poor use of your money.

On a side note, Broad match terms are often hard to control. They enter in more bid auctions that you'd like if you look into the search phrases that can trigger them. "Sierra" is a great example. You'll be in "Sierra Mist", "Sierra Trading Post", "Sierra at Lake Tahoe", "Sierra Mountains" auctions, as well as a whole slew of session based matching that will frustrate you with poor metrics and relevancy, and ultimately effect your budget for poor quality scores.

At the end of the day, we just want to get the consumer to the most relevant thing. In this case it is the exact used vehicle details page, on your website, they're searching for.

Eric Mayhew
Technical Director
Dealer.com
 
When we built this, we wanted to bring the user directly to the exact vehicle in stock. If a searcher looks for a "Sierra", we can see two poor experience scenarios without dynamic inventory.

Eric, this is a great "by the book" PPC theory and one that stemmed from research for true eCommerce platforms. If I'm looking for "White Nike Air Force 1", I would expect to see an ad that contains my search terms, click on that ad and be taken to a page that contained the exact product. This is proven to increase conversions.

Your philosophy (which again, I am not arguing), works phenomenally for people searching Google for a "Used White 2008 Jeep Liberty"?

My first question to you is: what is the percentage of people searching Google that way vs. people searching for "used jeep liberty" or "used jeep"? My second question is: are you buying these terms on an Exact Match basis?
 
This conversation could blossom because now we're talking about the strategy that can make or break pay per click campaigns. Typically, I like to talk about the simple concepts and let the software handle the complexity, but since you asked ...

For each vehicle in stock, we actually aren't advertising to the level of specificity you mentioned typically. You could, but searching at that level is not typical search behavior. We are using keyword variations around the "used jeep liberty" and many variations of that theme. In another ad group, we may even get more generic (with ad copy to match) for the "used jeep" searcher.

The difference between the super generic term "sierra" and the "used jeep" terms are that in the "sierra" term, we're so generic that we can't see any searcher intent in our keywords. "sierra", both broad and phrase (since it's just one word) is much too generic. Once you've selected a vehicle that is either exclusively a vehicle term, like "RAV4" it's not as important, but if you hope to deep link, the "new" and "used" qualifiers do help with classification. Jeep is a very obvious vehicle term so that is a great first level filter, adding the phrase "new" and/or "used" almost guarantees vehicle interest (plus provides classification).

When it comes to broad, phrase, and exact matching ... (fortunately you don't have to choose just one, you can make smart decisions where it counts and take a mixed bag approach)

In any scenario where an advertisers budget is the limiting factor (ie. there are more clicks available than budgeted for), I would absolutely try to use the available funds on exact match terms first. They are the most qualified and highest converting clicks. Let me caution you, that means that you have a good list of keywords, not uber generic, nor weak. It requires that you periodically use Google Insights, mine organic search phrases for high volume strong converting phrases, and use other available tools to identify search trends. Those phrases need to exist in your campaign if you want to use the above strategy.

If there aren't enough exact match searches available to efficiently consume an advertisers desired goals, introducing phrase match would be the next best option. it's actually a great match type to utilize as long as you don't use it on any single word keywords. It's even a little shady on two word phrases, but if they are good qualified words like "used jeep" then they are fine. They are less qualified than exact match but are easier to manage for the end user (but no harder for an automated system obviously). In our strategy, we typically bid a little less than the exact match variation for these keywords since they incur some unintended searchers and in the bid auction, would prefer our exact match to "win" whenever it can (since we can really deliver the best matched ad very reliably). That means that when exact match fails to match, we still have a phrase variation that may be considered for the auction.

Lastly, we would bring in broad match. Typically I don't recommend broad match for short phrases, but on long tail phrases, they work since Google has enough information to really understand the intent of your broad match. That said, even in that circumstance you get enough confusing search phrases showing your ad that you want to control, but can't without a constantly evolving set of negative keywords. Again, broad match gets a lower bid value than even phrase match in our strategy since it's the least qualified audience for our ads.

This staggered bidding lets the best matching keyword win. We've done this, so ... no one has to make this mistake themselves anymore ... but if you were to:

set up three ad groups, one for service, one for parts, and one for brand.
create specific ad text relevant to each of those themes.
create a keyword for "honda service" in the service, "honda parts" in the parts, and "honda" in the branding campaign (set all bidding equal, or better yet, say that you want to bid less on parts keywords).

Then have a searcher type in "honda parts". You'd think Google would pick your Honda parts ad group and show your Honda parts ad and let you deep link to your Honda parts landing page ... but they don't. They look at all keywords that are candidates for match (of which, "honda" is a candidate), look at your quality scores and bid and pick one. If all is equal, it ends up randomly selecting (I suspect it has to do with search volume, not really random). Therefore, you'll eventually end up with all your searches for "honda service" and "honda parts" being picked up by your generic branding term "honda". Thus, staggered bidding helps sift this out.

I'll stop here ... it's easy to turn this into a novel.
 
I think there's two conversations going on here: philosophy and functionality.

I'd like to talk about the functionality first, as Haystack's functionality is not tied into any specific search philosophy.

Speaking specifically for Haystack, it automates search campaigns based on inventory and user settings.

If a client wanted to enact a search campaign that included broad-match keywords that targeted used vehicle makes - the system could be set to generate it.

If another client wanted to enact a very targeted search campaign, that focused solely on exact match keywords that included information right down to the color, the system could be set to generate that campaign.

In addition to keywords, the system also generates appropriate ad copy and deep links based on client settings.

So philosophy aside, Haystack simply develops search campaigns based on a dealer's inventory and the search philosophy they find the most effective - whatever that philosophy might be. This increases speed, accuracy and user experience and reduces the amount of time necessary to manage a search campaign.

In regard to philosophy, I agree with Eric. Different dealers in different situations will utilize different philosophies - and the system can be leveraged to account for them.

Beyond functionality and philosophy, the other advantage to a program like Haystack (or Eric's TCD) is that there are some very highly trained people and mountains of sophisticated software behind them.

So you can see how the system performs, I've attached a search query report for a Haystack campaign. You can see how accurate the system is right down to the odd-ball ATV they happen to have in inventory. For those of you unfamiliar with a search query report, the first column is what a user actually typed into a search engine, the second is the match type and the third is the vehicle ad we delivered to them (that they clicked on).
 

Attachments

  • Sample Search Query Report.zip
    94.3 KB · Views: 17
Haystack seems to be a simple API with the engines. You feed Google your inventory and it creates ads accordingly. It's not new science, and I believe Dealer.com has a program similar.

It targets the long tail ("2005 GMC Sierra"). However, if you know how to buy Paid Search ads, there is a method to purchasing terms on a "broad match" or "phrase match". This means the engines pick up certain terms or a combination of terms within your ad, so you can show up for "2005 GMC Sierra" by just purchasing the term "sierra" or "gmc sierra", or any combination of the sort.

The other argument is that Google searchers are not looking for individual vehicles on Google (yet), so the long-tail targets are not going to yield you much traffic.

In response to this, it is a simple concept and not a new science. Retailers like Amazon and eBay have used technology like this for quite some time. However, there aren't many options for the automotive industry (Kenshoo offers software like this, however, you can spend $120k per year licensing it). In addition Haystack has additional automotive features that make it useful specifically for auto dealers. A dealer can utilize the tool for a couple hundred dollars per month.

As with my previous post, you can utilize the tool to make the campaigns as long-tail or short-tail as you like - that's a philosophical discussion.

You are right that people who know how to buy paid search have a lot of options at their disposal. However, most dealers don't know how to buy paid search nor is their time most effectively used doing so.

Even as knowledgeable search marketers, this tool saves our team and improves our campaigns.

For example, if you're an Acura dealer in Jacksonville, Florida you wouldn't typically buy the keyword 'sierra' or 'gmc sierra' or even 'used gmc sierra' since it would be unlikely you have that vehicle in stock at any given time. But, when it does come in stock, the system generates a specific keyword set (which we recommend to be tightly targeted), and an ad that lists the year, make, model, price and perhaps color and a deep link to that unit of inventory on your web site. When the vehicle is sold or otherwise out of inventory, it takes the ad down. Even with all the tools at our disposal, we couldn't recreate this process by hand.

Our typical dealer increases search marketing traffic between 15% and 25% with the tool. In addition, you can look at the search query report we attached on my last post. People are searching in that manner and we are getting good click volumes.

Hope this helps further the discussion.