Alistair Dent, 13 January 2010
Google introduced the Plusbox in the UK in beta a few months ago. The Plusbox is an extension to ads that will allow them to display related product images underneath the main ad text. We’re going to take a brief look at how it has been performing and what effect it is having on Adwords ads.
What is a Plusbox?
A plusbox is an extension to regular Google ads that shows a new line underneath allowing the searcher to view more information.
The below ad shows on the keyword “mountain bikes”:
It is a standard PPC ad, but with the additional line underneath indicating “Show products from X for Y”
When a searchers clicks on the plus symbol or the link, the ad expands out:
Google then display some products (usually 4-6) from the site in question, giving the user some choice (or inspiration) about where to go next.
The principle behind this move is the make ads offer more to users. The more products on display for a search, the higher the probability that something might be shown to a user that will be close to the reason for their search.
Background
The Plusbox feature works by linking an Adwords account with a Google Base feed (the Merchant Centre product feed that is used for Google Shopping). A Google Base feed is a list of products with stock status, price, images and descriptions. This makes the Plusbox extension suitable only for advertisers running an e-commerce platform, since it inherently is linking up to products, rather than areas of your site.
When a keyword in a campaign is searched on, Google takes that keyword and matches it as well as it can against the product names and descriptions in the Google Base feed. It will pull out a selection of products. How it decides is unknown, but given Google’s constant push for relevancy it seems reasonable to suspect that it is based on popularity, perhaps with some kind of categorisation. E.g. The search above for “mountain bikes” will probably match almost 50% of the inventory for a company like Wiggle. So Google will choose some highly searched categories, then from each it will pull out some of the most popular products. You can see from the screenshot above several different types of product at several different price points that all have relevance to the original search. The feature is still in beta and is changing all the time, but it seems like the algorithm in use is a sophisticated form of this.
The products selected will show their name, image and price. The name and image become hyperlinks to the Google Base landing page. Not the ad’s landing page. So a user will go to the specific product page they have clicked on rather than what you decided for the ad/keyword.
Data
What we haven’t performed here is a controlled, blind experiment. We have observational data, and we haven’t manipulated the data significantly to control for other factors. For instance, do Google show our plusboxes more often if we’re in higher positions? That would lead to a bias that would show us data like we’re seeing below. Anecdotally we have seen our plusboxes equally in a number of positions, but with the data Google provide it is very difficult for us to control for this (positions are only reported as averages over a particular timeline. If on 50% of searches we were in high positions, and often showed the plusbox and 50% in low positions and rarely showed it; our data would look much like the data below but we would never see those figures in the average-out stats from Google).
Experimental data is much stronger than observational data even for a lower number of data points. This is because there are many many ways to accidentally bias observational data, and unless the appropriate control group is set up correctly in advance, every statistical attempt to control for external factors is likely to influence the data in an unknown way. So this cannot be considered a rigorous statistical approach. What we have here, is raw untouched, unmanipulated (even to improve the quality) data, straight from Google about our Plusbox ads.
We ran with Plusbox ads for about two months, and took data from the second month (to allow the campaigns to settle a little after the Plusbox introduction) about click-through rates in the campaigns. Specifically about how often the Plusboxes appeared, how often people opened them, and click-through rates in the various states.
Advertiser A is an electronics retailer. They are selling highly commoditised, branded products (e.g. laptops, printers etc). They are using predominantly keywords that are very specific to the products, with almost no general terms. Traffic on these is very competitive, with everybody using similar ads that are almost all highly targeted.
Advertiser B sells retail display solutions. Products are not usually commonly known about by searchers beforehand, so most traffic is on more generic terms. Competition on these keywords is much broader, with each advertiser choosing what it thinks is the most attractive ad.
Overall impressions across these two campaigns over the data period was in the 7 figure range.
Analysis
The electronics retailer with very specific, targeted ads in a commoditised market saw a very low instance of people expanding the Plusboxes to view the products. Only 19% of keywords showed a Plusbox at all, and those were only opened 0.16% of the time. However click-through rates on ads with a Plusbox was over double the campaign average.
The display products retailer saw a slightly different pattern. Their ads included a Plusbox much more often, and were expanded more often. The click-through rate on ads was affected less by the existence of a Plusbox (but it was much higher already) and the products themselves attracted a lower click-through rate than advertiser A.
We believe that it looks like a pattern is emerging from this that explains a lot about the way these Plusbox ads are shown. The differences between these two campaigns (one very targeted and competitive, the other more general and trying to differentiate itself) are leading to dramatic differences in the Plusbox activity.
On a very specific term, chances are that there are less products in your Google Base feed that match. For instance, on a search for “black samsung n130″ there might be only one or two matching products in the feed. Google may well choose not to show Plusboxes on those keywords. On a generic term (such as those in advertiser B’s campaign) there are more likely to be many matches from which to choose the best products in the Google Base feed. E.g. “literature holder” may well match every product in an entire category, so a Plusbox will show every time. This seems to be a reasonable explanation for the dramatic difference in Plusbox appearance rates between the two campaigns, but there are other explanations. It could be that click-through rate is a factor in deciding whether to show a Plusbox, in which case you would expect a higher proportion of advertiser B’s keywords to show Plusboxes.
The numbers appear to be different the other way around when looking at the click-through rates on products shown. Again, a reasonable explanation would be the difference in keyword specificity. Using the same searches from above, “black samsung n130″ would be likely to show you product images that are exactly what you expect to see. But “literature holders” would probably show you a wide variety of product types, possibly none of which match what you had in your head. The difference between these two figures is not as significant so it could be chance that this explanation matches, so I’d take it with a pinch of salt. But it’s a good starting point to develop a hypothesis.
Conclusion
First off, it’s clear that the Plusbox extensions are benefiting both campaigns. The click-through rates for ads with a Plusbox are double the campaign averages, even if the Plusbox is not expanded. The ads stand out on pages of ads where differentiation is key. Whether this effect will continue when Plusboxes become more prevalent is unknown. Are these clicks coming from people who are clicking our ads instead of other ads? Or are they coming from people who otherwise wouldn’t have clicked an ad at all? It’s likely that the answer is both, and some of this effect will remain. Only Google can know at this stage whether Plusboxes are improving click-through rates for ads on any given search results page.
Secondly, it needs to be re-iterated that this data is raw, and uncontrolled. So this is not data on which to make a campaign decision. If we want to design a test however, then this gives us a great starting place. If we can control for certain factors (like making sure we’re not only showing ads when in higher positions, or on ads that already have better click-through rates, etc) then we can start to learn exactly how these Plusboxes are affecting our campaigns with enough confidence to start making changes as a result.
Tags: plusbox