Back in September 2014, Google made close variant matching (CVM) a mandatory, default setting for phrase and exact matches, pushing them into auctions for close variants like misspellings or plurals.
However, critical voices have raised that the new close variant match type is not performing as good as the true match type, speaking of more than 50% drop in conversion rates. As a consequence, many advertisers fear that there will be less control over bidding due to an increased volume of waste-traffic that they can’t control. This prompts them to use scripts, which categorise all misspelled variants as negative keywords.
We wanted to see ourselves if the effect is actually that bad or if CVM could maybe have an opposite effect. But instead of lumping together all keyword types, we decided to analyse search terms according to semantic cluster like “designer” queries or “generics”. Let’s take a closer look!
Semantic Clusters Make Closer Look Possible
Before we dig into the results, some words about semantic clusters. Instead of taking all keywords into consideration together, we wanted to know how conversion rates for exact variants behave in different semantic clusters. What do I mean by semantic clusters? Well, semantic clusters are groups of semantically similar keywords that allow us to take into consideration that “adidas predator” (designer + specific) is something totally different than “soccer shoes” (generic), thus optimising campaigns in more detail. With this in mind, let’s see the results.
Conversion Rate of CVM Is Lower – Difference Between Semantic Clusters Obvious
In order to find an answer to the questions we investigated search terms from a large, international fashion retailer over a period of 90 days in spring /summer 2015. The data basis was a total of 22,462 search terms, from which 11,998 were exact matches and 10,464 close variants.
We analysed these search terms according to the following semantic clusters:
- Designer + Specific
Firstly, we analysed the conversion rates of CVM in relation to the true exact conversion rate, excluding Brands.
The results within the three semantic clusters were as follows:
For Designer + Specific, the CVM conversion rate comes out to 7.42% lower than true exact match while for Designer, the CVM conversion rate is 12.98% lower compared to that of true exact match. At 47.69%, the Generics’ conversion rates show the highest discrepancy.
Secondly, we analysed the conversion rates of the three semantic clusters compared to the average conversion rate of the overall account.
For Designer and Designer + Specific, the results are very similar to the first graph, coming out with 9.38% respectively 18.53% compared to the overall exact account’s conversion rate. Generics though only have a 29.64% lower conversion rate.
The Bottom Line
As expected, the CVM conversion rate is actually lower compared to the true exact match. This is the same for all semantic clusters analysed, although there are clear differences between the clusters as well as compared to the overall account performance. But in the end, only the Generic cluster shows a similar strong discrepancy to the 52% mentioned in the beginning.
For both Designer and Designer + Specific keywords, the absolute deviation from the average conversion rate turns out to be rather small. “Close variants” can even improve performance by bringing up the account average. In individual cases, it may be necessary to check whether Designer keywords actually perform above average. With generics, working with negatives – whether with scripts or not – is still a good strategy.