Retail Performance Marketing - The Mercent Blog

Google Product Listing Ads: A 2011 Review and Outlook for 2012

About a year ago Google released Product Listing Ads (PLA) from limited beta, allowing all retailers to participate.   Ever since, we’ve emphasized to our clients the importance of this online retail for growth. 

Here I offer a quick overview of our experience last year and some thoughts on this channel for 2012.  In short, it’s growing and every retailer needs to be engaged on PLA.

We analyzed a cohort of retailer clients consistently active on paid search (i.e., text ads, referred to as Adwords below), Google Product Search (GPS), and PLA throughout 2011.  As a group and for each individual merchant, client sales and traffic from PLA grew significantly, with growth particularly accelerating in Q4.  As a percentage of client paid search text ads sales, their PLA sales accounted for ~19% of attributed sales in December for these clients.  By the end of the year, their PLA sales accounted for about 13% of their total Google attributed revenue (GPS + PLA + Adwords).  These results are typical of our client base that activated PLA later in the year.  The figure below shows the acceleration.  

The growing importance of Google PLA:
Analysis of sales from a cohort of Mercent clients

In short, Google’s PLA is growing and every retailer needs to be engaged and selling through this channel.

Additional observations:

Retailers can grow both GPS and PLA simultaneously.  Use of best practices saw a fair number of Mercent clients drive achieve growth in both of these programs.  Growing GPS was no small challenge in 2011, but if merchants continuously monitor performance and competitive positioning, optimize their Google Merchant Center (GMC) feed, and maintain high data quality they performed well on both programs.

PLA does not appear to cannibalize paid search text ads.  We cannot scientifically assert this one way or another with certainty yet (that’s a future project).  But, we did not see any evidence of cannibalization of or negative correlations with paid search text ads.  If anything, they seem to grow together.  This is obviously a key question for marketers so we plan to pay close attention to this.

PLA performs well.  Typically we see clients generate ad spend to sales ratios for PLA that are similar to their best performing CSEs.  As such, PLA is now effectively perceived as a major CSE within their CPC and PPC program.  The continued attraction of PLA, though, is that the volume opportunity seems substantially greater than that of most other CSEs, at least in the near term.

PLA optimization is a lot of work.  If you think clicking the button to make your feed items available for PLA and adding a bid is enough to do well, you’re missing optimization opportunities.  Not only must the text in your feed be optimized, but promotional copy needs to be added.  The biggest chunk of work is properly structuring your product targets in your feed, taking into account past performance for each SKU as well as logical product category structure, so that you can implement effective and successful bidding strategies.  And your feed quality needs to be as good as possible.

Quick thoughts on PLA for 2012:

• Expect more growth and distribution of PLA.  Google has every reason to grow the program and we expect them to do so.  In fact, in November they announced PLA launching in Europe which is a good indicator of their intentions (http://adwords.blogspot.com/2011/11/product-listing-ads-rolling-out-to-uk.html).
• Free up more budget specifically for PLA.  I would look at PLA not as a testing channel but as a growth channel in 2012.  I don’t mean to be a cheer leader for Google, but I think retailers need to be here.
• Competition will increase (doesn’t it always) making A/S ratios harder to attain.  But, I believe (expect?) Google will open up the supply side of the equation enough so that good results can be widely attained.
• Do not abandon GPS or sacrifice any efforts to optimize your GMC feed.  If anything, do more.  The best feed wins.  It’s still true.  PLA success depends primarily on the quality and structure of your product feed.  Effective bidding is dependent on the targeting enabled by the feed.  
• Measure Google traffic and performance holistically.  Look at your total return from Google (sales from GPS, PLA, Adwords, and organic) and divide it by your total spend on Google.  I’ve previously called this the Return on Google.  For practical reasons you will still need to manage channels somewhat individually, but the interconnections between programs at Google and the extent to which their algorithms do or do not use signals across programs are complex and probably unknown, or at best unclear.  Therefore, I encourage you to consider Google as an aggregator of an audience - measure how much of that audience you get, what it costs you to get your share, and how much you get from it.  It’s easy to get super-granular but keep the big picture in mind, too.

 

Is Google Monetizing Google Product Search through Paid Search Ads?

“What’s going on with Google Product Search (GPS)?” is a question I’ve received a lot lately.  GPS was a smoking hot channel for most of the last 18 months.  But traffic and sales comps since about March 2011 have been down, in some cases substantially.  Well, several things are going on, but I think a major one is that Google is starting to more aggressively monetize GPS by shifting traffic away from Google Shopping results and toward its revenue generating programs - Adwords (paid search ads) and Product Listing Ads (PLA).  This is a hypothesis of mine, and I have no inside or other information from Google confirming or denying this.  Below I share why I think this is happening and some of the implications for retailers.

First, some very brief history on GPS performance:

• 2010 January-September:  GPS performed awesomely.  We were realizing YOY comparable store sales well over 100%.  In addition to great work by us and our clients, this was due to easy comps against the 2009 recession and some tactical missteps Google made on GPS in 2009.
• 2010 October-December:  Q4 2010 YOY comps got harder because (among our client base) the consumer spending recovery started in Q4 2009.  Also, Google changed some aspects of the user experience that negatively impacted GPS traffic.  In particular, in mid-October they made changes that emphasized local results and, we believe, larger stores.  We saw a dramatic downtick in traffic and sales, but Google acknowledged the mistake and corrected it before the brunt of the buying season.  Short story is that GPS still put up very impressive numbers in Q4, but not like earlier in 2010, meaning comparable store sales were in the 20-50% range.  (In hindsight, their experiments in October and November 2010 look like initial forays into monetizing GPS that had too much of a negative impact on GPS traffic.)
• 2011 January – June:  The comps started declining further.  Declines accelerated in April, May, and June.  By late spring we were seeing comparable same store sales for GPS in the single digits and lower.  Median account yoy sales growth for GPS only, went negative by late spring.  Essentially, Q1 2011 was still pretty good for GPS in terms of YOY growth, but Q2 was poor to very poor.

Several things are probably affecting GPS results, but I believe one of the biggest is that Google is monetizing GPS by shifting traffic away from GPS and toward its revenue producing channels.  The industry has speculated about whether and how Google would monetize GPS.  Some have speculated that Google would eventually make GPS a CPC program like other comparison shopping engines.  This is a legitimate option, but I think what we’re starting to see is that Google will stick to their tried and true revenue generator, paid search ads, as the vehicle to monetize traffic that comes to Google for shopping.  PLA, in this perspective, is a mere experimental riff on paid search ads.

Below is aggregate data from the set of Mercent clients consistently active on all Google programs since January 1st, 2011.  The subjects are all retailers – some large, some smaller.  It was not possible to make this a scientific test versus control study.  The data presents correlations, not causal evidence.  Nevertheless, it’s real historical data - let’s see what it tells us.
First, sales:  As mentioned above, GPS has seen tough times recently, especially since March (coincidentally around time that Panda started).  GPS comps were good early in the year but then, generally speaking, they declined.   The sales numbers below show that in absolute terms GPS sales have declined over the course of the year.  At the same time, however, paid search sales (which are labeled Adwords below) have increased.  The increase in paid search sales is more than the drop off in GPS, though the percentage change is roughly comparable.  What is also important, though, is that *total* sales have increased.

Daily Sales from Google Channels, 2011 YTD
$ per day

 
Now let’s consider what drives sales (since Google can’t itself directly affect retailer sales), traffic and conversion rate:  Traffic is something Google can directly affect and it shows a similar pattern to sales, with GPS traffic dropping off in March while paid search traffic increased at the same time and by roughly the same amount.  It’s worth noting that retailers are limited in their ability to affect traffic to GPS, especially in the short term.  Optimizing feed quality and content is, of course, critically important, but there is no lever like bidding that could be used to affect traffic.
 

Daily Traffic from Google Channels, 2011 YTD
Clicks per day 

 
 
As far as conversion rate is concerned, GPS stayed constant year to date, while paid search conversion rate increased substantially.  Google can affect conversion rate by preferentially ranking sites or content of high quality.  This has been a well publicized goal of its recent Panda changes, so the results here seem consistent with that objective.
 

Daily Conversion Rate from Google Channels, 2011 YTD
%, orders/clicks per day 

 
What does this all mean?  I believe Google is trying to increase revenues primarily through paid search ads and secondarily through PLA by changing the user experience so that some traffic to Google Shopping is effectively captured by those channels.  I think they have set a secondary goal of improving performance metrics of their paid search programs.   I also believe they’re constraining themselves by trying to achieve these goals by minimally impacting GPS traffic and sales.  I believe the sometimes wild fluctuations in GPS traffic over the last several months is Google trying to get the balance “right.”

This makes sense from a Google perspective:  the product Google sells is paid search ads and what they’re doing is trying to make that program look bigger and better to their paying customers.  They’re achieving this, in part, by leveraging the high quality, structured content from the “free” GPS program.

The data suggests they’re making it work:

• Total sales are up
• Paid search traffic and sales are up
• Paid search conversion rates are up
• And even though GPS sales have taken a hit, they’ve mitigated the impact to this channel by improving the quality of traffic and, hence, the conversion rate of GPS traffic

This is all fair enough – Google is allowed to make money.  But it creates several implications for retailers.  The first of which is that they should begin to assess Google holistically as a publisher rather than as a collection of individual channel silos.  Optimizing paid search, SEO, PLA, and GPS separately from each other only makes sense if that (i.e., separately) is also how Google runs them.  This does not appear to be a valid assumption.  Consider a metric as simple as Return on Google (ROG).  This is simply total sales from Google channels (in this case paid search, PLA and GPS) divided by total dollars spent on Google channels (in this case Adwords and PLA).  The ROG chart for the same client group is shown below.  (Note that ROG for GPS is undefined since spend = zero.)

Daily Return on Google (ROG), 2011 YTD
%, sales from all Google channels / spend on all Google channels 

 
This fascinates me.  Through all of this, Google has been able to marginally increase total retailer sales and make their paid search program look more efficient (at the expense of GPS) while keeping their total ROG essentially unchanged.  So, in net this looks good for retailers, but we shouldn’t presume this will always be the case. Hence, there are several key implications for retailers.  I’ll address these in future posts, but for now here they are briefly:

1. Retailers need to assess Google performance holistically, across all channels.  Retailers should use something like a ROG metric and evaluate all their Google programs; if Google pursues monetization of all programs through Adwords (and allows their algorithms to move traffic between programs) then looking at each Google channel individually may lead to business-shrinking results.
2. Retailers need to concentrate specifically and consistently on the quality of their product feed to Google.  Even if the numbers directly attributable to GPS go down (i.e., without using a ROG-like metric), the data above indicates that the data feed is significantly growing in importance to Google, especially as it relates to retail search.  In fact, I think the data feed may soon surpass other inputs in terms of importance to Google in determining relevance for retail related user searches.  There’s a great recent article on this topic (including pointing out that retail is leading the way on “paid search without keywords”) here: http://searchenginewatch.com/article/2083543/How-Inventory-Feeds-are-Changing-Paid-Search-Advertising. 
3. Retailers should hone their tracking so they can better understand which type of content on Google drove which clicks and which sales.  For example, was it the “Everything” search results page, a “Shopping” page search, an image, etc. from which the click originated.  Knowing which ad or ad group drove the results may not be adequate.

As stated in the beginning, this is only a hypothesis though one apparently supported by the data at hand.  I’m curious what others are seeing.

Frank Kochenash
Vice President Performance Services, Mercent