Blog

Managing PLA Campaigns with Segmentation & Proper Bidding

In October 2012, Google Shopping became Product Listing Ads (PLAs), a paid advertising format. Since this transition, PLAs have become increasingly more important in driving revenue for B2C advertisers.

According to Kenshoo’s 2012  Global Online Retail Report, the visually appealing PLAs draw about one and a half times the clickthrough rate of regular text ads, and convert 23% better, resulting in a 31% higher return on advertising spending (ROAS).  Currently, CPCs are less competitive in comparison to search CPCs for most industries.

Taking advantage of PLAs is a no brainer but managing them can be tricky. We often take on accounts which contain one ad group targeting all items in the data feed with no negatives.  That’s like bidding on one keyword in broad match and just letting the account run.  Not a good idea! This set up allows for little control and limited optimization opportunities.

So how do we take control?  The first step involves analyzing the data feed to identify groupings of related products which will serve as ad groups within the PLA campaign. A few definitions for clarification:

1. Google Merchant Center Data Feed: A file consisting of a list of categorized products.

2. Product Targets: Instead of keywords, PLAs use product targets to determine which items from you data feed appears on the Google search results page. They allow you to bid differently on different sets of products.

3. Attributes: Product Targets are defined by attributes.  Attributes such as your items’ brand or category, define your item in a unique way.

Are you a reseller with inventory from numerous brands? We can use the brand attribute to target products by brand.  Do you sell clothing online? We can use the product category attribute to target by category of clothes (jeans, tops, skirts, etc.). There are infinite options and each targeting strategy will depend on our client’s type of business and data feed setup. The end result is multiple ad groups which each contain a specific product target or a select grouping of product targets.

Why segment? Our main tool to drive results is adjusting bids on the product targets. Just like in search where keywords perform differently, product targets will perform differently.  When we target all product targets with one bid, we can’t increase or decrease bids depending on individual performance of products. By segmenting by product targets, we can use the ROI per ad group to decide whether to increase or decrease bids.

Let’s look at a simple example: a reseller sells 2 brands, A and B.  Brand A drives twice the amount of revenue than Brand B. If we use the default all product targets strategy, we have to set the same bid for both brands. By segmenting the brands into two ad groups, we can place a higher bid on brand A and a lower bid on brand B, increasing overall ROI and driving more revenue.

We also utilize more advanced features using AdWords labels, but I’ll save those tips for another blog post!

PPC Predictive Bidding – Aggregate Data is Your Friend

In most accounts, we see big discrepancies in the performance of these modifiers across keyword categories.

How do we do this evaluation? I prefer to take a look at the search terms report for the entire account under the dimensions tab in AdWords. Make sure that you’re looking at a nice solid chunk of data and have included all relevant columns.

Once you’ve exported the data, apply filters and exclude all rows that contain branded terms. We want to evaluate non-brand queries only. Next, apply subtotal figures for clicks, impressions, cost and conversions. Apply formulas in this top row for CTR, average CPC, CPA & conversion rate. This will allow you to subtotal the data for different modifier filters.

After that, you’re set to go! Start evaluating the data by selecting the search term filter in column A and selecting “contains” then type in the different modifiers that you’d like to test.

For one B2B client, we saw that the modifier “software” had a CPA that was 2.5x the CPA of the modifier “platform”. On one eCommerce account, we saw a 150% delta in the ROI of “cheap” vs. “discount”.

So what do we do with all of this data? Whether you’re using manual bidding, bid algorithms or rules-based bidding, you’re likely concentrating your bid adjustments on keywords that are converting effectively (bid up) and keywords that are high-spend and non-converting or have high CPAs/low ROIs (bid down). Since only a fraction of your keyword inventory convert in a given period, this bidding methodology usually only affects a fraction of your overall keyword inventory.

Modifier analysis helps you implement more intelligent predictive bidding. Predictive bidding allows you to put yourself in a position to get more conversions at lower CPAs by making calculated bets on keywords that haven’t converted yet. Predictive bidding allows you to bid up the aggregate of long-tail keywords likely to convert, rather than just your root terms.

For instance, if I have an account average CPA of $80 and I see the following modifier CPAs:

- sevices – $118
- solution – $112
- system – $104
- SaaS – $54
- Web-based – $52

I’m going to bid up lower volume, low position, non-converting keyword terms with “SaaS” and “Web-Based” in them, while bidding down non-converting services/solutions/systems terms.

Note, Google recently allowed you to filter by search terms in the search query report:

Enhanced Campaigns on Google – Quick Summary of Major Changes

Google recently announced that they will be making changes to how their campaigns work.  These new AdWords campaigns are being called Enhanced Campaigns.  The main change is that AdWords is no longer allowing advertisers to easily split out campaigns by device.  Overall, most of the changes are positive, but Google has taken away some flexibility and control.

Here is the AdWords blog post announcing Enhanced Campaigns.

 
Below are the biggest pros and cons of the new campaigns.

Positive changes
1. Bids based on location (can set different bids for different states in 1 campaign)
2. Ad group Sitelinks
3. Mobile-specific ads in same campaign as desktop ads
4. Sitelink scheduling, reporting per Sitelink, and other minor changes

Negative changes
1. Tablet and desktops are grouped together and must have the same bids
2. Budgeting for mobile (can’t set separate budgets)
3. Mobile keyword bidding
-Bidding is at the campaign level for devices.  You can bid mobile at 50% of desktops, but you might find a selection of keywords that perform better or worse on mobile and not want to have them bid at that default level.

 
Here at Metric Theory we are staying on top of the changes and developing transition plans for each individual advertiser.  Since Google has stated that current Legacy campaigns won’t be shifted over to Enhanced Campaigns until June, we’re advocating that most clients hold off on transitioning.  This provides the benefit of collecting some data and best practices from several test accounts in the meantime.  It’s also possible that Google may slightly alter some features of Enhanced Campaigns due to advertiser feedback.

Overall, most advertisers shouldn’t see a major impact from switching over to Enhanced Campaigns, but there are some details to pay attention to (i.e. – setting mobile phone bids or completely turning off mobile phones with a -100% bid).  As data geeks, we’re looking forward to analyzing and leveraging geographic and other data to take full advantage of the new options and features!  Please don’t hesitate to reach out to us with questions on Enhanced Campaigns.

If you would like to see more details, here are two links from Google which further explain the changes:
About enhanced campaigns
Upgraded ad extensions in enhanced campaigns

There is also a good round-up of the changes at SearchEngineWatch.

Lastly, Google is hosting a series of Enhanced Campaigns webinars.

Who is Metric Theory & Why Are We Calling?

My name is Jenny Gale Howe, and I’m a Sales Director with Metric Theory. If you’re reading this blog entry, you’ve probably received a call or an email from someone on the Metric Theory Team. And if that is the case, you’re probably wondering … “Who is Metric Theory, and why are they calling me?”

To be honest, if we’ve reached out to you at all, it means we’ve found something worth discussing in your Google AdWords performance. With that said, I’m guessing you get many paid search related phone calls a week, so let me tell you what sets this one apart from the others.

The business development team at Metric Theory also happens to be an extremely talented group of SEM professionals who audit both internal and external performance for paid search accounts. So if we’re calling you directly, you can expect that we’ve found enough evidence to suggest that we can likely help with the performance of your account.

Let’s put it this way:
If I’m shopping on Google for a new windshield for my 1965 Ford Fairlane (and yes, I do drive this car) And your company shows me a paid search advertisement for “Lawn Mower Batteries,” … True Story … Then I’m going to give you a call.
Plain and simple. I’m shopping on Google, and your company is displaying a less than desirable “user experience.” And you likely don’t even know it!

Our team proactively reaches out when we’ve found something that isn’t functioning properly in your live online advertising campaigns. We pride ourselves on providing solutions so you can walk away with actionable insights to improve your account’s performance.

And to be honest, we won’t call you with just one example.
That would be a waste of our time, and yours. We only call organizations where we see multiple issues indicative of larger scale problems throughout the account.

Bottom line – we don’t pick up the phone without a reason.
So what happens if you answer – or heaven forbid – call us back?
Well, to begin with, we will offer some free advice.

You already work with another agency?
Fine – take these problems back to them and tell them to do the dirty work.
They should be doing that already, yes?

You do this in house? – Great!
We might be able to teach you a thing or two.

Regardless, we can promise you an interesting, and free paid search discussion. If that “Wows” you, we can offer you a free in-depth Audit of your account. Again – no strings attached. And all of this is done to help us prove our salt, and gain your trust.

So with that said, I strongly encourage you to take a minute or two to hear us out. The worst thing you will get is some expert PPC advice. Or, on the flip side, we may uncover a very expensive mistake that will make a world of difference in terms of your return on investment! And in a world where so little is truly free, I’d say that is a really unique, and exciting, opportunity.
We look forward to speaking with you!

Using Google Trends for Analyzing Search Query Spikes with Hurricane Sandy

Here is one chart for the entire US:

Here is another chart that shows the spike was more dramatic for New York:

Even though there was an obvious increase for New York and other affected states, Hurricane Sandy also drove increased awareness around emergency kits in states as far away as California:

There are a number of events that may impact one area of the country or the world, but have reverberations elsewhere. Whether you are running efforts for disaster supply companies, relief organizations, or other groups, Google Trends is a useful tool to examine and analyze spikes and swings in search query volume.

PPC Ad Copy – Testing is Not Enough

There’s a decent chance that I’ve lost your attention before I’ve even begun.  That’s because PPC blog posts espousing the importance of ad copy testing are more common than candy on Halloween.  If you’re still with me though, I’m not here to tell you that you should test ad copy.  You already know that.  The biggest issue is that most advertisers stop at testing.

In almost every new account that we bring on board, we see advertisers testing 2-3 ads in each ad group.  The problem is that as an advertiser you don’t realize improvements in account performance just from running a test.  Let me repeat to be very clear: you do not improve your PPC account performance just by testing ads.  You improve performance by acting on the information gained through ad copy testing.

If you go to Foot Locker to buy a new pair of running shoes, you’re likely going to try on a couple different pairs (ad copy testing your running shoes).  One of those pairs is going to be more comfortable, lighter and fit your foot better than the other.  The reason that’s important is that you want to optimize your performance.  You want a shoe that allows you to comfortably run further at a faster pace.  Once you figure out which shoe you prefer, would you buy both pairs to rotate them on runs?  Would you wear one of each shoe on each foot?  Of course not!  So why, when advertisers know that a certain selling proposition in ad copy is working better than another, do they continue to run both ads!

Whether you’re running a multivariate test or a split test, you need to gather enough data to make an informed decision on ad copy.  That said, at that point the impact on performance can be significant:

If you pause ad #2 and anticipate the same performance out of ad #1, you can then apply ad #1’s rates to the 17,991 impressions given to ad #2.

17,991 impressions * 2.43% CTR = 437 clicks (compared to 367)
437 clicks at $1.56 CPC = $682.00
437 clicks * 4.62% conversion rate = 20 conversions

Now, instead of this ad group producing 44 conversions at $38.73 each, it will produce 52 conversions at $33.77 each.  That’s an 18% increase in conversions at a lower CPA.  What did you have to do to achieve it?  Take one second to set an ad to pause.

Your work isn’t done though!  Maybe you were testing a unique selling proposition in ad #1 that concerned price and selection vs. a USP focused around quality in ad #2.

    VS.     

Thus, you have data that shows the USP from ad #1 outperforming the USP from ad #2. Now you should test that winning ad copy around selection and pricing in other ad groups. This allows you to more broadly leverage the winner of your ad copy test and realize better performance account-wide rather than just telling your VP of Marketing, “We’re pretty sure pricing and selection works better in ad copy.”

In this situation, you would want to implement that same line 1: 300+ Blue Widgets from $9.99, in other applicable ad groups throughout the account.  You can either run this head to head with the winning ad copy in each ad group, or if you have conclusive data, replace existing ads.  What you are effectively doing is shifting a higher percentage of your overall account spend from lower performing ad copy to higher performing ad copy.  This will positively influence the overall account performance.

More to come in later posts on how to analyze ad copy tests to determine winners across multiple campaigns.  For now, don’t head out to the track with a different shoe on each foot.  Act on your ad copy tests!

PPC Keyword Research – PPC Keywords as Billboards

I hate to begin with a cheesy metaphor, but there is no better way to explain the role of keywords in PPC. Keywords are your billboards; not the advertisement that is pasted up there, but the physical apparatus.

Think about it…

If you are arduously grinding through San Francisco on the 101 (the stretch of highway separating the city proper from Silicon Valley), your eyes are drawn towards colorful, creative messaging for a variety of technology driven appliances and applications.  Conversely, if you find yourself on I-80, skirting through a string of rural communities, you are more likely to see an advertisement for Bass Pro Shop or the latest Indian Casino.  While there are some variances in size and surface area, the largest differentiator between a billboard in San Francisco and a billboard in rural California is the audience it serves.  I’m not saying that San Francisco lacks a population that would benefit from a Bass Pro Shop advertisement, but the heavy concentration of cutting-edge tech companies defines a demographic more receptive to a tech-oriented advertisement.

In principle, PPC keywords are no different.  Through keywords, you define your audience.  The model of delivery is more advanced and enables real-time data, but the core principle is consistent: if you don’t target the right audience, you will not be successful.

As a company, you define your audience based on a need/demand from potential customers.  Unfortunately, the way a potential customer expresses this need via a search engine isn’t always that intuitive.  In order to help advertisers ensure they are displaying their messages to the appropriate audience, I have included several tips for PPC keyword research.. 

Tip #1: Avoid Assumptions, Data Doesn’t Lie

PPC is a form of direct-response advertising; meaning, users tell you what they want to see.  However, assumptions that a certain search term is relevant can be detrimental to the performance of an account.  There is no guidebook to defining consumer sentiment (as it is expressed on search engines); fortunately, there is the ability to test.

Whether it is CPA, ROI, revenue or leads, a keyword’s value is determined by its ability to impact the overall performance of the account.  In an effort to avoid assumptions and eliminate guesswork, advertisers should use their accounts’ KPI as a baseline of evaluation.   For example, made-up company, Tennis Clothes Direct recently added the keyword “tennis shoes.”  You might assume that a keyword such as “tennis shoes” would target their desired market of tennis players; however, this assumption fails to acknowledge that a large portion of consumers still refer to day-to-day footwear as tennis shoes. 

Tip #2: General & Vague Keywords Drive a Generally Vague Audience

As a PPC keyword becomes more descriptive and qualified, the available audience diminishes (the volume of searches is smaller).  It’s an unfortunate reality that every advertiser faces, but if you wish to really dominate a market, you must consider the inclusion of general keywords to generate enough volume.

General keywords are high-volume head terms without the use of any qualifiers.  For example, the keyword, “desktop virtualization,” would be a general keyword for a company that offers Virtual Desktop Infrastructure (VDI).  The benefit of general keywords is the size of the audience.  The downside of general keywords is the unpredictable sentiment of that audience in terms of where they are in the purchase funnel.  To select the right general keywords without constraining traffic, advertisers need to ask themselves the following questions:

1. Does this keyword directly define my product and/or service?
2. Does this keyword define other products in the industry?
3. Is this something you would search if you were in the market to purchase your solution?

Using the VDI advertiser, let’s compare “desktop virtualization” and “virtualization technology” as they relate to the above questions.  This way we can see the differences between general and too general:

Tip #3: Consider Competitors for PPC Keyword Research

If someone is searching for a direct competitor, they are likely in your target market.  It’s fair to assume that they have done enough industry research to be familiar with a specific vendor.  In other words, they aren’t just doing introductory research about a product or technology, but rather they are actively referencing a specific vendor.  This searcher has indicated they are deeper in the purchasing funnel.

An efficient account places a heavy investment on traffic that is more qualified/more likely to purchase.  Since competitor traffic consists of “purchase ready” searchers, it would seem imperative to target this high-quality traffic.

Tip #4: Stay Away From Secondary Messaging

Anyone willing to stand behind their product will rattle off a seemingly endless list of benefits when asked, “how does your product benefit me?”  The problem is that a large majority of those benefits are secondary to the core offering of their product.  Moreover, most markets are so saturated that there are specialty products for every potential benefit.  For example, there isn’t a brand in the shoe industry that offers a multi-purpose shoe made for work, the field, the trail, and the beach.  Instead there are running shoes, comfort shoes, dress shoes, etc.

Of course, Nike running shoes are comfortable, may be appropriate in the workplace, and could be used on the trail, but given the implicit market segmentation it would be fruitless to advertise for Nike running shoes with keywords such as “comfortable shoes.”  While certain benefits may be valuable for ad copy, they might not work as well for keyword selection.

Follow the steps above when conducting PPC keyword research, and you can ensure that you are advertising on the right billboards and managing your spend wisely.

PPC Ad Copy – Showing vs. Telling

Little did I know in 1999 that my high school English teacher was going to teach me a lesson that would be extremely valuable for future PPC ad copy writing.

Somewhere between analyzing the corrupting power of unchecked ambition in Macbeth and teaching us to use three strong supporting points to advance a hypothesis, Mr. Maretzki emphasized the importance of showing vs. telling.  He instructed us to paint a picture with our words.  He told us not to get lazy with our adjectives.  Most importantly, he insisted that we distinguish our thoughts on a topic in a way that made those thoughts more valuable than other writing on the topic.

These lessons are surprisingly applicable to writing ad copy for PPC.  Your ad is meant to be a portal between a consumer’s search and an advertiser’s solution in the form of a landing page.  Your job in writing ad copy is to convince the consumer that you are providing the best portal to accomplish that.  There are often ten other ads on the page for a given query.   How do you distinguish yourself?  How do you stand out?

Every time I see an advertiser talking about their great selection, great prices, or easy-to-use product I cringe.  Did you hear me when I said you were competing against ten other advertisers?  What makes you think that sort of generic language will induce a click on your ad?

When I speak to some of these very same advertisers about what makes their product or service better than those ten other competitors, I’m typically met with an impassioned soliloquy about the fantastic and distinguishing features of their business.  Those features need to be reflected in PPC ad copy.

The best trick for showing not telling in ad copy is to ask yourself why.  Why is our selection great?  Why are our prices good?  Why is our software easy to use?

Some examples of generic ad copy that tells vs. ad copy that shows below:

As Mr. Maretzki also taught me, nobody likes to critique his own work.   How do you evaluate whether you’ve written PPC ad copy that actually stands out?  Be honest with yourself: if you searched a given query, would you be most likely to go find what you were looking for by clicking on your ad or a competitor’s ad?  If it’s the latter, that probably means it’s time to go back to the drawing board.

Finally, it should go without saying that these new ads should still be tested against the prevailing winning copy.  Ad copy that shows, like the examples above, typically beats out the generic copy, but making that assumption can get you in trouble.  Have the humility to test your ad and let the data do some show and tell of its own to confirm you’re on the right track.

Beyond the Paid: Jeremy Brown Interview

Melissa Mackey sits down with our VP of Online Marketing, Jeremy Brown, to discuss AdWords’ new sitelinks policy.  What are the benefits of sitelinks?  What are the potential pitfalls of the new policy?

“Sitelinks are absolutely great for taking up more space on the page.  That is their number #1 utility… I’m definitely not a fan of Google enforcing a unique link for every Sitelink, and I’ve provided feedback to Google through numerous channels that this will result in a worse user experience.”

Find out more here:

http://www.beyondthepaid.com/adwords-sitelinks-policy-interview-jeremy-brown/

The Future of Remarketing: RLSA & Google Analytics Remarketing

Across the past several weeks, Google has significantly beefed up its remarketing offering.  In the process, they have eliminated the need for clunky pixels and added depth to the level of segmentation.  Both Remarketing Lists for Search Ads (RLSA) and Google Analytics remarketing offer new ways to advertise to individuals based on their interaction with your site

Considering the limited exposure for each of these, I have yet to confirm my excitement and expectations through data.  Eventually, I plan to follow up with a more thorough evaluation, but for now, here is a simple introduction.

Remarketing Lists for Search Ads Beta

RLSA is the largest step Google has taken towards Search Remarketing to date.  Before you jump for joy, understand that this is not true Search Remarketing based on impressions and queries; rather, it permits advertisers to differentiate search advertising based on whether an individual as visited their site or not.  Regardless, it’s still exciting stuff.

What Does It Do?

If you have ever employed remarketing through the GDN in the past, you’ve placed pixels across your site to cookie specific audiences.  RLSA allows you to leverage these audiences/remarketing lists in search.  In similar fashion to traditional GDN remarketing strategies, RLSA enables advertisers to augment bids and messaging on the search network based off a searcher’s previous interaction with your site.   Moreover, it also allows you to exclude audiences.

Why Is This  Awesome?

Someone who has visited your site before is more likely to convert and therefore holds more value than someone who hasn’t.  RSLA allows you to employ this principle on the search network.  Let’s take a look at a hypothetical use case to illustrate this more clearly.

Joe is in the market for a high-end fashion shirt.  He conducts a search on Google and clicks an organic listing for BestMensFashion.com and is immediately cookied by a remarketing pixel.  Joe shops around, but isn’t ready to purchase and exits.  A few days later, Joe resumes his search and types “affordable mens fashion.”  Usually, for that query, BestMensFashion.com would display an ad in position 7-8.  But through RLSA, BestMensFashion.com would be able to increase their bid and position due to the fact that Joe had been cookied.  In order to implement this, BestMensFashion.com would have to create a mirror campaign that contains this keyword, target it to the appropriate audience and adjust bids accordingly.  With that increased bid, BestMensFashion.com now displays an ad in position 2 for Joe.  We bid higher because we know Joe (and other members of this remarketing audience) are more likely to convert. The greater visibility and brand recognition leads Joe to click on it.  After a few minutes of browsing, Joe purchases a shirt and in the process unknowingly demonstrates the potential impact RLSA can have for any advertiser.

Google Analytics Remarketing

Another new offering from Google allows you to run remarketing using audiences from Google Analytics as opposed to just AdWords.  By incorporating a small DoubleClick parameter in your analytics code, you can leverage analytics audience segmentation for GDN remarketing:

What Does It Do?

Google Analytics allows you to segment data across a variety of metrics and dimensions.  These insights help advertisers optimize their websites and online marketing programs, but now they have an additional functionality.   Through Google Analytics (GA) remarketing, you can create remarketing lists based on audience segments (standard or customized) found only in GA.

Why Is This Awesome?

First and foremost, it negates the need to place a different Google AdWords remarketing pixel every time you want to narrow down on a specific segment.  Previously, if you wanted to target visitors who bounced with a full shopping cart, it took at least 2 unique pixels.  Now, all that is required is a slight addition to your existing Analytics code as demonstrated above.

Second, imagine the possibilities.  GA remarketing grants access to a devilishly appealing level of audience segmentation.  Previously, these data sets (new vs. returning visitors, direct vs. organic, etc) were exclusively used for evaluation.  Now, you can identify a particular segment of users that are quantifiably more valuable to your business and maintain a front-of-mind presence.  But take caution.  With such a wide plethora of opportunities, there may be a tendency to overcomplicate your remarketing strategy.  More than anything, this would result in over-segmentation and leave you with a convoluted list of audiences that make-up a minimal portion of site traffic.

You can use default or custom segments to define an audience.  Using Joe in another example, let’s take a look at how we could use the non-bounce default segment.  This time around, Joe is in the market for some organic underpants.  He conducts a search on Google and stumbles upon GreenUndies.com.  Once Joe visits this site, he is immediately cookied by GA.  During this visit, Joe engages with a number of pages, but does not purchase.  Through his active engagement, Joe defines himself as a non-bounce visitor.  Due to the fact that Joe did not bounce and was actively engaged, he is a valuable segment.  GreenUndies.com realizes this and creates an Ad Group that employs the non-bounce default segment as an audience to re-engage with Joe, and visitors like him, across the GDN.