The FindWAtt Blog

Google Shopping, Product Feed Management & Product Data Optimization


Share it!

5 Ways to Measure Search Term Specificity in Adwords Shopping

Tim Gilbert 2017-09-14

Why Search Term Specificity Matters for Revenue

This is in response to an idea for increasing Adwords Shopping Campaign revenue tested by Ed Leake and a possible refinement suggested by Dan Barbata.

How We Built & Optimized Google Shopping For 192% More Revenue by Ed Leake
Increasing Bids as Search Query Specificity Increases by Dan Barbata

To quickly summarize what they are talking about, you get more conversions if you capture clicks from people ready to buy today, and people who are ready to buy tend to use more specific search terms than those who are simply browsing.

It's easy for a human to look at a couple of search terms and decide whether they are more generic and specific. But your search term query report (SQR) probably contains tens of thousands or more. And to make it worse, the long-tail specific search terms are the lower frequency ones that tend to change over time. To fully optimize your PPC campaign, you need an automatic way to discover and prioritize those terms so you can focus your efforts where they will have the largest ROI.

How to Measure Search Term Specificity for Buyer-intent.

  • The text length in number of characters. This is by far the roughest approximation, but the easiest to calculate. One reason it isn't precise is that a short string like an model number (e.g. "Canon EOS 550D") is more specific than a more descriptive string using more generic terms (e.g. "entry level Canon DSLR camera").
    Note: I recommend removing stop words like conjunctions, pronouns, articles, and prepositions before calculating length because they don't have any significant impact on specificity.
     
  • Campaign structure. If you have structured your Google Shopping AdWords campaigns and product groups correctly into hierarchical groups, you may have a slightly better indication in your search query report. It is possible to look at the depth of the group that the search term was in (how many product type levels, brand, custom label are included). If the campaign group that included the search term click was a top node (e.g. "product_type_l1=lighting parts & accessories"), it's part of a more generic topic than a campaign group include another node layer and brand (e.g. "product_type_l1=lamps and product_type_l2=lamp shades and brand=design classics lighting")
    Note: In the SQR fields, product group is called "Keyword".
     
  • Word frequency in the English language. Some words are very common in English, while brands, trademarks, and more specific product types and attributes will tend to be less common words (or even not in the dictionary). The words in "river hip wader" are less common than the words in "black walking shoe", which means the first search term is more specific.
     
  • Product specificity. How many products in your shopping feed contain all the words in the search query. A highly specific search term will only match few or maybe even just one particular product. A less specific term may match many items in the same category, while a highly generic query will match products across various categories. If you're carrying apparel products, then a search for "blue jeans" will presumably show up in the product text or attributes for hundreds of products, while "tan capris" will match fewer.
     
  • Attribute recognition. As suggested by Dan, some attributes indicate a much higher buyer-intent than others. Someone who searches for shoe style + size  (e.g. "Jazz shoes 7 1/2") is probably more ready to buy than someone searching for shoe style + color  (e.g. "Jazz shoes brown"). Someone who searches for a particular manufacturer part number is even more likely to buy. If we combine the count of the attributes in the search query along with the specificity of each attribute, we get a very reasonable approximation of search specificity. This is made possible by our technology of context-specific attribute recognition in unstructured text.

My recommendation:

The best approximation of search term specificity is probably a combination of the separate measurements above, especially product specificity and attribute recognition. If we use both search term specificity, expected search term performance, and calculate per-search term revenue estimates, we should know exactly where to focus our attention to increase our PPC shopping campaign revenue the most.

If you want to use search term specificity to improve your non-shopping PPC performance, you can still use text length combined with word frequency.