An Examination of the Effects of Google’s New Close Variants
Published: March 25, 2019
Author: Will Cozart
- Exact match keywords now map to queries with the “same meaning” as the keyword
- Share of overall exact match clicks from close variant matches has increased to about 25%
- The average query captured by exact match now maps to more than 2 distinct keywords
- The average exact match keyword now maps to more than 21 unique queries (misspellings and “similar intent”)
- Exact match CPA increased to the end of the year at a higher rate than 2017 – in line with the close variants changes
- Advertisers will need to analyze the performance of close variants in their accounts, and make adjustments to mapping as necessary. Negative additions at the ad group level for accounts in a SKAG structure will be required as query performance dictates
Roughly 15% of daily queries on Google are brand new. To help advertisers capture these, Google announced on September 6, 2018, that it would once again expand its definition of close variants to include keyword intent.
The definition would change to begin “including close variations that share the same meaning as your keyword.” Google implemented the change in order to match keywords to searches that include “implied words, paraphrases, and other terms with the same meaning.”
Goal of Analysis
We wanted to see the impact on exact match traffic in the wake of the change and subsequent rollout to Google’s ad-serving algorithm. With 3Q’s Alpha/Beta account structure, any significant shift in traffic and impression volume on Alpha keywords from pure exact match to exact match (close variant) could be cause for concern and potentially necessitate increased negative scrub frequency and bid adjustments to compensate for performance shifts.
To determine the impact of Google’s close variant definition change on exact match traffic, we first looked at the close variant share of traffic on exact match keywords using the match type column in Google’s search term report. The subject of this analysis was a large subset of the 3Q Google MCC accounts (roughly 100 accounts), broken down by month over the second half of 2018.
Close Variant Share of Exact Match Keyword Traffic
There wasn’t much impact to close variants’ share of overall exact match clicks as Google rolled out the change in September. In October, however, the share of exact match clicks coming from close variants jumped up 13% MoM as the share of impressions increased 24% to 12% of total exact match, the highest in the observed period.
Moving into the end of the year, we saw an interesting trend of elevated click share from close variants even as impression share fell below the October peak. By December, close variants were making up nearly a quarter of total exact match clicks, despite impression share falling back below the level seen pre-announcement. December’s close variant click share was 23% higher in 2018 than 2017, evidence of Google’s inclusion of “similar intent”queries in addition to misspellings. In December, CTR from close variants jumped up well above its recent average, a trend not present in 2017.
Shifts in Query Mapping
Next, we looked at the number of exact match keywords to which a particular query maps, and how that has changed over time. Using the same date range as the above analysis, we pulled top-spending Non Brand search queries from a sub-set of accounts in the 3Q Google MCC. After removing redundant matches for different targeting types (RLSA-only campaigns, international targeting, Experiments, etc.), we determined to how many exact match keywords, on average, a specific query mapped. For example, does the query “red shirts” match to both [red shirts] and [shirts red], and how has that changed?
The trend is similar to above, as we saw an increasing number of keywords matching to a single query as time progressed. At the end of July, queries were mapping to an average of 1.6 EXM keywords, increasing each month to a peak of 2.1 keywords in December (a 33% increase).
The trend here is potentially troubling for accounts, as it shows that increasingly, building out an exact match keyword for a top-performing query won’t always capture all searches for that query or serve your purpose-built ad for that keyword without additional negatives.
Shifts in Keyword Matching
Lastly, we looked at how many unique queries matched to a single exact match keyword for the same sub-set of accounts used above. In July, the average exact match keyword for our test set matched to 12.9 unique queries. This number dipped a bit into Google’s September announcement, increasing from there with the average exact match keyword matching to a whopping 21.7 queries in December.
For account teams, the shift reinforces the need to monitor search query reports and scrub poor performers. Overall exact match CPA increased 28% from November to December, 2018, a larger change than we saw in 2017, where December’s CPA was 19% higher than November. So, it does appear that the increased number of queries matching to individual keywords is having a tangible negative impact on overall campaign performance.
It’s important to note that these close variants were mostly misspellings, but the number of “intent-based” variants increased in step with the overall number.
We continue to see Google automate various aspects of the search landscape. Chief among these changes is what the algorithm considers the same intent as a user’s query. Since the definition change in September 2018, we are seeing more queries match to exact match keywords, and more keywords matching to unique queries. Historically, exact match has been the easiest match type to control, but close variants muddy the water for account structures like Alpha/Beta, which were built on 1:1 mapping. We believe the close variants change is potentially the next step in Google’s push toward a keyword-less search account of the future. Google already recommends an account structure consisting only of BMM and DSA campaigns to take full advantage of automated bidding and Machine Learning.
With the increasing number and variety of queries matching to individual keywords, it’s more important than ever to remain vigilant in doing negative scrubs and ensuring keywords are not matching to poor-performing queries – “same intent” or otherwise. Account teams may need to begin adding exact match negatives into Alpha ad groups to control query mapping. It will be important for individual accounts to perform analysis on query performance when matching to the “pure” exact match keyword compared to a close variant match. If we see little to no performance variance between the two, we should explore further consolidation to reduce segmentation and improve data density.
Because of the variety of accounts, and individual team management styles, it must be noted that by no means can we assume that any shifts in traffic can absolutely be attributed to Google’s change in the close variant definition. Changes to accounts as part of normal management will affect the above results. Also, note there is a high level of variance among accounts in the number of queries matching to keywords by account.