Google are continuing to push the boundaries of how they match queries to the keywords we include in Google Ads campaigns. There have been a few updates to the scope of “close variant” search queries in recent years, whilst Google announced BERT last month which is sure to impact on matching moving forward.

What is a close variant match? Google explain it best here, although in short it’s Google’s method of matching what it sees as relevant terms that wouldn’t have otherwise matched to your keywords. Whilst this has obvious upsides (more matching = more volume, without the need to add in thousands of keywords), it’s really important to know the extent of the close variant matching in any accounts you manage whilst also reviewing the relevance of the close variants you are being matched to.

To get some context on the scale of the close variants matching, we analysed ~50 accounts that have had presence in the last year and a half and came up with some pretty interesting discoveries:

The % of close variants appears to be increasing

The graph above shows the close variant matches as a % of total search query impressions across the accounts.

The ratio started at just over 10% in mid-2018, to consistently above 20% in the past few months in 2019

Close variant queries have higher CPCs

CPCs are consistently higher, at an average of 21% higher CPCs than regular matched queries. 68% of the accounts we analysed had close variant CPCs higher than the regular matched queries.

In our reviews, we isolated Brand & non-Brand queries and CPCs were up on both in the accounts we tested. For Brand, we found that there were instances whereby slight misspells had up to 10x the avg CPC of the actual exact match brand query.

The question we’d have here is; why are CPCs higher on the close variant queries?

We’re thinking Quality Score must be a factor here, as the queries are less relevant to the ads than the intended Exact/Phrase keywords. Quality Score has historically been calculated at keyword level, so if specific queries are being “punished” with higher CPCs then this calculation must have been tweaked by Google. Would love to hear any other thoughts on this!

Are there implications on wider performance?

The most important question here; are the close variant’s relevant and is performance comparable to the regular keyword match type? This is something that needs to be answered on an account by account basis, although I’ve given some examples from a couple of verticals below.

In an e-commerce Google Ads account we run; non-Brand Close Variant comes in with higher CPC’s ($0.48 vs. $0.40), lower CTR’s (11% vs 14%) and lower ROI (9:1 vs 11:1). Whilst there is definitely merit in having a huge chunk of the close variant traffic live as it is still providing Revenue, it’s clear performance isn’t as strong as the core matching queries (where we have more control on messaging/bidding etc).

In our research we found that the matching for close variants is generally pretty strong, with most the high volume queries tending to be plurals/misspells. However, we have noticed a few instances whereby the intent of the query has changed following the close variant match. For example, a query “university of birmingham” can match from a “university in Birmingham” keyword. The intent of the search completely changes; a user looking for a specific university rather than an option of all universities in their target location. If we’re paying the same CPC’s (or higher for close variant) then that’s not ideal, as we’d expect conversion rate to be significantly lower on the “competitor” type search.

Some next steps!

• Audit your account to see what extent your queries are matching to close variants. Establish the performance of the close variants, as this will determine the scale/scope of the changes you’ll need to make.
• Next, review the match quality of the close variants in your search query reports. For all close variants with strong volume, you will decide whether to leave matching to your keyword as it is, add in as a keyword in it’s own right or add in as a negative. For more advanced analysis, you can pull an nGram report for all close variants, so that you can diagnose theme performance at scale to help with any new adgroups or negative keyword lists.
• Finally, I’d recommend that you consistently monitor the ratio of close variants as a % of total, and make a rule to perform a search query deepdive if the ratio increases significantly on any given day/week.