Query Refinements: How Google Helps Users Find Products Faster

Shopping SERPs person been looking much for illustration a feed than classed results for a while now. In December, I wrote astir nan integration of Google’s shopping tab into nan main results for shopping queries in e-commerce shifts. The consequence is simply a marketplace that looks much for illustration Amazon than web hunt results.

SERP features and query refinements play a large domiciled successful this transition. They nonstop users from unrefined searches to uncovering products arsenic quickly arsenic possible, having an outsized effect connected clicks and revenue.

In this heavy dive, I analyzed +28,000 shopping SERPs to understand really query refinements activity and really e-commerce sites tin usage them.

It’s a spot early for shopping season, but I’m penning a batch of e-commerce because astir shops request immoderate clip to make changes connected their tract (especially nan large ones). So, if you want a fruitful 2024, now is nan clip to get nan activity connected nan roadmap.

This portion builds connected 2 analyses I’ve published previously:

Boost your skills pinch Growth Memo’s play master insights. Subscribe for free!

What Are Query Refinements

Query Refinements are “pills” astatine nan apical of nan hunt results that thief users refine their search. In essence, query refinements are merchandise filters.

First announced at Google Search On 2022, Google explained that refinements (and filters connected desktop) travel real-time hunt trends (bolding mine):

Whole page shopping filters connected Search are now move and adapt based connected real-time Search trends. So if you’re shopping for jeans, you mightiness spot filters for “wide leg” and “bootcut” because those are nan celebrated denim styles correct now — but those whitethorn alteration complete time, depending connected what’s trending.1

Examples of Query Refinements. (Image Credit: Kevin Indig)

The intent of query refinements is to bring users from nan “messy middle” to conversion arsenic quickly arsenic imaginable arsenic they literally refine the query. When you click connected a pill, Google sends users to different SERP, conscionable for illustration erstwhile users click nan merchandise filters connected nan left.

To understand really query refinements activity and really e-commerce sites tin usage them to their advantage, I dug up immoderate data.

How Query Refinements Work

I analyzed 28,286 shopping keywords (US, desktop) with seoClarity and recovered that query refinements:

• follow chopped patterns sites tin usage for keyword targeting.
• lead to hunt queries without hunt volume.
• trigger caller AI Overviews connected mobile.

Common Refinements

I analyzed which refinements travel up astir often successful “position” one, 2 and three. Think astir position 1 successful this discourse arsenic nan first refinement from nan left, which is nan astir visible.

Most refinements specify gender. The word “women” comes up astir often successful nan apical 3, but “men” comes up astir often successful nan first refinement. 45% of query refinements mention 1 gender astatine slightest once, 61.4% if you see kids. It makes sense: earlier diving into merchandise attributes for illustration colour aliases size, you want to make judge a merchandise is “for you.”

The astir communal query refinements. (Image Credit: Kevin Indig)

The 2nd astir communal group of refinements is location. Ten percent of nan apical 3 refinements see “nearby,” which is overmuch much visible connected mobile. Google shows maps by default connected mobile devices, arsenic mobile instrumentality users are much apt to beryllium connected nan go.

The 3rd group is attributes astir queries that see “for” aliases “with”, wherever users effort to specify usage cases (9.8%), and nan 4th is value (9% of refinements see nan word “sale”).

Query refinements person a precocious overlap pinch merchandise filters connected desktops and often characteristic nan first fewer filters arsenic refinements. Product filters don’t beryllium connected mobile, apt because users mightiness expect nan select sidebar connected desktop, but it doesn’t make consciousness connected mobile.

Product filters (desktop) and query refinements thin to person precocious overlap. (Image Credit: Kevin Indig)

The sorting and visibility of refinements are different connected mobile and desktop. Due to nan quality successful bigger surface size, mobile hunt results show ~4 refinements connected load, while desktop tin show complete 10.

Since query refinements are based connected realtime searches, they besides overlap heavy pinch autosuggest.

(Image Credit: Kevin Indig)

Interesting Findings

Three insights from nan information amazed me:

First, Google keeps refinements strictly focused connected merchandise attributes but not personification intents. I expected searchers to beryllium willing successful opinions and reviews connected Reddit, but neither “Reddit” nor “reviews” came up arsenic a refinement a azygous time.

Two, query refinements precisely lucifer nan query, meaning you won’t find synonyms aliases intimately related position successful them. As a result, brands don’t look successful refinements, either.

Three, astir query refinements don’t person hunt measurement aliases a CPC. Only 10,696 / 27,262 keywords successful nan first refinement person hunt measurement (median = 70), and only 6,514 / 27,262 keywords person a CPC. Since query refinements are based connected hunt behavior, we tin reason that hunt measurement and different keyword planner information are very constricted metrics.

AI Overview Refinements

Of course, I came across AI Overviews (AIOs) successful my research. For nan queries I analyzed, mobile results returned AIOs but desktop doesn’t. An illustration is brown mascara.

Brown mascara connected desktop, nary showing an AIO. (Image Credit: Kevin Indig)

Brown mascara connected mobile, triggering an AIO. (Image Credit: Kevin Indig)

You apt spotted nan AIO tabs connected nan screenshot above, which look independently of refinements and explicate communal merchandise attributes.

(Image Credit: Kevin Indig)

Notice really nan AIO provides further guidance and accusation successful tabs (see screenshot below).

(Image Credit: Kevin Indig)

At this point, it’s unclear whether citations successful AIO tabs are bully because they thrust postulation to reappraisal articles aliases bad because they springiness nan reply away.

(Image Credit: Kevin Indig)

For different queries, for illustration “air compressor”, I was capable to spot refinements in the AI Overview alternatively of supra it. Clicking an AIO refinement leads to different hunt pinch nan refinement successful nan query. For example, connected nan SERP for “air compressor”, one refinement is “for coating cars”. Clicking it leads to different SERP for nan query “air compressor for coating cars” (with different AIO and tabs but nary refinement). Notice that I was logged into nan SGE beta, which intends those features mightiness not yet beryllium unrecorded for each user.

(Image Credit: Kevin Indig)

5 Lessons

5 cardinal lessons surfaced from my study of complete 28,000 shopping queries:

• You should create circumstantial merchandise and class pages for men/women/kids erstwhile it matters for products (e.g. fashion).
• Mine query refinements and autosuggest to find applicable query variations for your keyword investigation (for example, seoClarity tin do this).
• Monitor ranks by query refinement to thrust your decisions astir facetted indexing (like Nike or Target). Refinements showing different URLs are an parameter of building circumstantial facets.
• You request to place searcher liking beyond hunt volume. The truth that much than half of queries don’t person hunt volume, but query refinements are optimized for hunt behaviour shows that you mightiness miss a batch of opportunities by limiting yourself to queries that person hunt volume. Instead, leverage onsite hunt data, surveys and qualitative investigation to heighten keyword targeting.
• Monitor and comparison clicks from desktop and mobile results to understand nan effect of merchandise filters (desktop) and AIOs (mobile).