Omn 1903 goog_shopng

1
The Path to Effective Keyword
Targeting in Google Shopping
optimizeShoppingcampaignswithn-grams
Howto AccountforDiscrepanciesinPerformance AcrossQueries
As search marketers, we know the value of search queries varies
drastically from query to query. Metrics such as conversion rate or
revenue per click (RPC) help uncover valuable search terms. However,
even with this information it is difficult to optimize Google Shopping
campaigns down to the query level without a strong, standardized
process for evaluating performance.
In this white paper, we will show you an easy yet extremely effective
method for query segmentation across product listing ad channels.
First, we will look at the customer decision process and the
importance of segmenting queries by query type. Then, we will
explore how to identify valuable queries in a performance-based
way using an n-gram analysis. Finally, we will learn how to
segment campaigns to bid according to the value of queries
using negative keywords.
Often, Google Shopping search queries align
neatly with the consumer purchase decision
process. Buyers who are casually researching
products tend to search using generic terms and
plurals. On the other hand, more qualified shoppers
search with refined queries, often including words or
phrases that indicate some additional level of intent.
For example, a customer looking to purchase a new lacrosse
head begins by searching terms like “Lacrosse Heads.” As they
move closer to purchase, they gradually refine queries down to
terms like “Nike Lacrosse Heads” or “Nike CEO Lacrosse Head.”
Since the customer is much more likely to purchase after searching for
more specific terms, those terms are correspondingly much more
valuable to the business.
Once we’ve identified high and low value queries, we can assign bids to
Google Shopping ad groups based on the value of the query. On an ad hoc
basis this sounds easy, but how do we apply performance data to identify
valuable (or non-valuable) queries across an entire account?
N-GRAMS:

2
Performing anN-gram Analysis
Now, let’s roll up performance across all search
terms containing the word “lacrosse.” In the
table shown above, you can see the word
“lacrosse” existed in 46,579 unique search terms
within our data set, resulting in a total spend of
$152,575 at a 5.53x ROAS. Looking solely at
these metrics, the term “lacrosse” appears to be
very valuable to the business; however, its
performance is actually roughly in line with the
overall average of the account.
Since “lacrosse” appears in the majority of search
terms, the word itself isn’t particularly useful for
determining which queries are valuable. In the
next step, we should find more actionable
insights in the campaign’s outliers.
An n-gram analysis evaluates the performance of a word or series of words across the entire
account, and determines which terms are valuable and which are not. In order to effectively
determine the value of queries, we must look at the performance of one, two, and three-word
terms (unigrams, bigrams, and trigrams) which each hold significantly different implications
for tracking performance.
In this example, we will look at
unigram, bigram, and trigram
queries generated by Google
Shopping ads for a lacrosse
equipment retailer.
By looking at the performance of
a word or phrase and how that
particular term performs across
all queries it exists within, we can
decipher their value with regards
to the campaign.
In the table to the right and in
the following example, we will
explore the performance of
search queries containing the
word “lacrosse”.
NOTE ON UNIGRAMS
Unigrams help identify single words that are
strong indicators of performance. For example,
a unigram word might be a popular brand name
or a discovery word (ex. what, how, which).
STEp1: Unigram
Search Queries
Unigram

3
Once we’ve pulled all the insights from the
bigrams, we can evaluate at a deeper level by
looking at three-word phrases: trigrams. Here, you
can see queries containing “womens lacrosse” vary
widely in performance depending on the qualifying
phrases that exist alongside the base term.
Queries containing the phrase “womens lacrosse
shaft” show strong above-average performance,
and therefore may merit more aggressive bids.
Conversely, search terms containing “womens
lacrosse stick” do not perform as well and may
merit a reduced bid.
Sure enough, two-word phrases containing
the word “lacrosse” vary significantly more in
performance than the single word alone. For
instance, queries containing the phrase “cheap
lacrosse” or “for lacrosse” perform well below
average, while “youth lacrosse” is above average.
Knowing this, we can implement a phrase-match
negative keyword for “cheap lacrosse” and “for
lacrosse” to reduce spend to those terms. That
spend then becomes available for other more
valuable terms such as “youth lacrosse.”
NOTE ON BIGRAMS
Many bigrams help
identify where a customer
is in the purchase decision
process. In the above
example, “for lacrosse”
indicates the shopper is
still in the discovery phase.
We would, therefore, bid
lower since the shopper
is not ready to purchase.
NOTE ON TRIGRAMS
Trigrams often indicate the
most qualified words and
phrases. For instance, many
trigrams include the brand
name and model number.
You’ll also notice there’s a
lot more diversity in this list,
compared to the unigram
and bigrams.
STEp2: bigram
STEp 3: Trigram
Bigram
Trigram

4
Implementing Results
Once you’ve identified your valuable and non-valuable phrases, you need to ensure the campaigns are
structured in such a way that you can actively segment by query type.
However, there’s one big problem with this strategy: Google Shopping ads do not allow advertisers
to directly segment ad spend by query intent. Ad placement is determined by probability of a match
between what the user is searching for and your product feed attributes. Worse still, the PLA structure
requires advertisers to determine spend on the product level. This means all queries related to that product
are lumped together, no matter how relevant. So how do we ensure we are serving the highest-value ads
to the highest-value customers, and limiting spend on customers who don’t intend to purchase?
The answer is a unique campaign structure—and robust negative keyword lists. Every product should
live in at least three separate places within the account: in a high-value campaign, a mid-value
campaign, and a low-value campaign.
Create a series of negative keyword lists based on your n-gram results—one for low-value terms, one
for mid-value terms, and one for high-value terms. At this point, you definitely want to double-check to
ensure each term is in the correct list, because these negative keyword lists help determine how much
you’ll spend to acquire traffic at each level of the funnel.
w
w

5
Summaryofn-gram Analysis andnegativekeywords
N-gram analyses offer a performance-based approach to segmenting by query in your account and
provide key insights into understanding the value of those queries as they relate to your business.
In this white paper, we have shown you an extremely effective method for query segmentation across
product listing ad channels using n-grams and negative keywords. We looked at the importance of
segmenting queries by query type according to likelihood of purchase, and how to identify valuable
queries in a performance-based way using an n-gram analysis. Using unigram, bigram, and trigram
words and phrases we gained valuable insights into the performance of those queries. Finally, we
learned how to segment campaigns to bid according to the value of queries using negative keywords.
If the strategy above is implemented correctly and kept up-to-date, it allows for not only immediate
improvements both in total revenue and profitability, but continued benefits in the future as well.
We can also help—if you’d like to gain a better understanding of performance by
n-gram, contact us at sales@omnitail.net for a free analysis!
www.omnitail.net | (617) 307-4969
Unigram
Bigram
Trigram
N-gram
Analysis
Query
Segmentation
Profit
Once your lists are in place, bids for each product group can be determined based upon the
combined performance of each product and value segment, rather than a blended average of
each level of query intent. This strategy isn’t maintenance-free; you’ll still need to regularly review
and update the lists and monitor campaign performance to ensure your campaigns are on track.
The best way to do this is to keep a cache of known high-value terms to reduce the time needed
to scrub data. Small businesses with less traffic and a fairly static product offering can probably
classify all search queries manually once the initial n-gram is run, but larger businesses will need to
run queries through the n-gram process regularly to ensure the lists remain accurate and up to date.
NOTE ON PRIORITY SETTINGS
Google Shopping offers three priority
settings for campaigns: low, medium, and
high. These setting are key to driving low
value traffic to low value campaigns and
high value traffic to high value campaigns,
so make sure you have them correctly set
correctly for each type of campaign.

Omn 1903 goog_shopng

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Omn 1903 goog_shopng

Similar to Omn 1903 goog_shopng (20)

More from EvoLife.bg

More from EvoLife.bg (10)

Recently uploaded

Recently uploaded (20)

Omn 1903 goog_shopng