Assessing the Impact of Image Extensions

By , Senior Account Manager, Community


Google first announced image extensions in beta in June 2013, providing an exciting opportunity for advertisers to include visual content directly in text ads. Playing off Product Listing Ads (PLAs), which have produced great click-thru and conversion rates for many advertisers, image extensions allow for three visuals to show at the top of text ads. These ads not only look more visually appealing, but also occupy additional search real estate. Here is an example of an image extension.

Example of Image Extension

I was initial excited about image extensions because of their ability to visualize concepts. With this extension, companies not using PLAs can still use visuals. The example above uses images of a hotel and of the Sydney cityscape. These images grab the eye, helping the ad stand out even more than what other ad extensions can achieve.

I first heard about the beta in April 2013, and was able to get one of my sporting goods clients in the door. This client was running PLAs, however, so we used the image extensions for concept related terms (ie: terms around training and specific skills).

Before going live, the advertisements must meet a few requirements. First, the images need to be associated with specific ad groups. Based upon five popular ad groups I asked my client to provide three images for each group, fifteen in total. Images must be representative of all keywords and ads in the particular ad group. In other words, you can’t use an image of someone drinking coffee for an ad promoting running shoes. Second, image extensions will only show if your ad is in position number one above the organic rankings.

Reviewing the Data

Fast forward one year and I’m ready to present the data, organized by click type, consisting of:

  • Headline Clicks
  • Image Ad Clicks (Image Extensions)
  • Mobile Clicks-To-Call
  • Product Plusbox Offer Clicks (now retired)
  • Sitelink Clicks

Note that these clicks are solely on Search Network text ads. Display and PLA campaign data are not included.

In reviewing this case study there are a few other considerations to keep in mind.

Small Sample Size – This extension only had the opportunity to show in five ad groups. Remember that this extension only shows when text ads are in the top position.

Images – The images used in this extension remained the same throughout the year. Though normal optimizations were made (keyword bids, negative keywords, etc), the images were not touched.

One Client – This data only represents one account in one vertical. Since I was only able to get one client on the beta I have a limited sample. The data may look different when analyzing multiple accounts.

The table below showcases click and impression data by click type.

Table of click and impression data

It’s no surprise that headline clicks are the most abundant. Even with the additional real estate that extensions occupy, the headline text is larger, which generally makes it more eye catching and easy to click. I am surprised that image ad clicks accounted for a lower CTR than headline clicks. Because they occupy roughly half the ad, I believed more searchers would click an image based upon the same principle as the headline.

When reviewing the average cost per clicks, image ads are the highest. Again, this doesn’t come as a surprise; image extensions need to be in position one to show. It does beg the question of how conversion and revenue metrics are impacted with such a high CPC. Speaking of which, here are the conversion and revenue metrics.

Table of Image Extensions conversion data

First off, I am disappointed to see that image extensions only accounted for 11 conversions during the year. That’s less than one conversion a month! Even though conversion rate and average order value are higher than headline and sitelink clicks, the sample size is relatively small. Additionally, cost per conversion is higher, as our CPCs are that much more expensive than other click types.

The final table looks at each click types’ percentage of overall metrics.

Table of percentage of Image Extensions data

Image ad clicks only accounted for 0.18% of total clicks and 0.20% of total impressions. This data highlights the scarcity of how often the image extensions were presented.

Final Thoughts

Image extensions are a great idea in theory, but put to the test they had minimal impact. They gained so little exposure that it was tough to get a statistically significant sample size. Even when they did show, the metrics were okay, but generally did not perform as well as other click types. I still like the idea and would run this feature for other clients, but I don’t expect the needle to significantly move.

What have your experiences been with image extensions? Please leave your comments below!

Twitter Facebook LinkedIn Google+ Email Print More

17 thoughts on “Assessing the Impact of Image Extensions

  1. Chris Oliver

    Its interesting to see the above stats. I’d generally agree with this. I too thought it would be great however the image ads show so infrequently my clients don’t think they’re even enabled !! I thought it was a crafty way for Google to ‘encourage’ higher cpc’s to achieve position one. I’ll be disabling them soon.

  2. Matthew Umbro

    Hi Chris,

    Thanks for your feedback! I love the idea of image extensions but in practice they see very little traffic.

  3. Pete Leatherland

    We’ve used them but they rarely show up, normally just on a brand search. In theory it lets us show certain products (or categories of product) our customer or someone searching for us may not be aware we sell so in that way it is a bonus to introduce a range and to help with our branding by associating us with what is in the image before they get tot he site. It also makes the site look more established by having more area, especially if we have G+ posts showing and all the other extensions.

  4. Tom Holder

    It’d be interesting to see the stats from the ad group as a whole, rather than segmented by click type? Although the click volume and CTR appear low for image extensions, there’s nothing in your stats to show that the image extensions might capture more attention, but the clicks end up going to the headline? My experience of image extensions is just that – there was clear improvement in topline ad group click through rate and volume after introducing image extensions.

  5. Matthew Umbro

    Hi Tom,

    Thanks for your comment! Certainly it is a possibility that the presence of the images resulted in higher headline CTR and/or more clicks (very similar to sitelinks). Unfortunately, Google does not allow the use of the “This Extension vs. Other” segmentation like it does for Sitelinks. Even if we did have this metric, however, I don’t know how drastically different the results would be. Nonetheless, I would still use this extension but not have any expectations of great improvements.

  6. Matthew Umbro

    Hi Pete,

    Thanks for reading! I do like the fact that the image extensions make my brand look more credible. Just like any other extension, I don’t expect significant clicks on the actual extension, but I do like that I’m able to take up more search real estate.

  7. Andrew Bethel

    I tested the image extensions last fall and experienced very similar results, Matt. The unique ability to showcase a high quality and relevant image to a user sounds in theory like a great idea.

    However, I had to increase bids to the point of unprofitably and when you tie that together with the increase in CTR (because it happens to look like a Google image result, maybe) it just didn’t make sense to continue with the image extension.

  8. Matthew Umbro

    Hi Andrew,

    Thanks for your input! I hadn’t thought about the idea of the extension looking like a Google image result. Could definitely be a reason why CPA was so high for the extension.

  9. Emma

    We had a terrible time with image extensions! I work for a shoe company and we’re very creative so we were super excited by the idea of image extensions. However they got disapproved several times for reasons that didn’t really make sense – ‘too much blank space’ was a favourite. Once we did get a set approved they barely showed up – largely I guess to the fact that all our other extensions were deemed more relevant.

    Great idea in theory, not so much in practice sadly!

      1. Emma

        Hi Andrew – we did it for our own brand term where we have no competition and are constantly in the number one position. I think when push came to shove our sitelinks, location extensions, etc. were all deemed more relevant – or at least that was the impression I was given by our rep.

        1. Andrew Bethel

          We were running this on non brand but the image extension was very max bid sensitive. Small threshold between not showing and showing.

          We began by testing pretty high bids to get a read on how often the image extension would show ($2.00-$5.00 which is high for this traffic) but bringing it down, even incrementally, caused the image extension stop showing entirely. I except the increase in SERP real estate was coming at a major cost.

          This was a small test but we didn’t end up pushing it any further. Bigger fish to fry so they say.

    1. Sam OwenSam Owen

      Hi Emma,

      I had a similar experience with disapprovals. Acting as a middleman between clients’ creative teams and Google meant getting them set up could be a slow process.

      It’s the reason I don’t think Google will bring these out of beta. The resources needed to ensure these images are of high enough quality for the user experience not to suffer is way too high. I wouldn’t be surprised however, if they did do something along the same lines as Bing where they introduced a rich-ads-in-search style set up for branded ads only.

      1. Emma

        Hi Sam,

        It was a nightmare – especially as we kept finding ads that had a lot more blank space than ours. It would be great if they could find some way to improve on them as they really were a great idea!

  10. Mark H

    May have to do with the fact that the images are high up that they could easily be mistaken for organic image results.
    In theory they sound good to almost everyone because you’d think people would click so maybe the Beta will tweak this on the SERPs to somehow distinguish paid images from organic images.

  11. Pingback: This Week in Ad Tech: Week of June 17, 2014

Leave a Reply

Your email address will not be published. Required fields are marked *