Testing Theory: Academic Studies You’ll Actually Use is a series that provides practical testing ideas based on the study of how people make decisions (formally known as behavioral economics).
Photos have the power to greatly influence conversions and engagement, but testing them is tricky. There are so many potential images you could be using, how are you supposed to know which ones are likely to have a big impact? For example, how did the Obama and Romney 2012 campaigns decide to use the images below on their donate pages?
One concept that can guide you to find the ideal image is priming.
A simple definition of priming states that exposing a person to a stimulus (image, word, idea, etc.) will trigger related thoughts (in an automatic, subconscious way). This has the power to affect people’s behavior and decisions. For example, a recent study (pdf version) divided fund-raising call center employees into three groups. Each group was given a fact sheet to use while they made phone calls. In the corner of the sheet there was either a picture of a woman winning a race, a generic photo of employees at a call center, or no photo. The groups primed with the call center photo raised the most money (and the woman winning a race came in second). Just showing images of people acting the way we want them to act is enough to influence their behavior, without needing to explicitly tell them how they should act.
This same concept can be applied to your web testing strategy by choosing photos that subconsciously influence people to act in a way that’s aligned with your organizational goals. A great real-world example of this is the websites of Obama and Romney. Both had the same goals – to raise donations and ultimately win the election – but had different content strategies that lead to different outcomes. We don’t know how much money was raised through each candidate’s website, but a close approximation is private donations under $2500. According to the New York Times campaign finance infographic, Obama raised $637.3 million in this category, whereas Romney only raised $388.1 million in the same category. Digging deeper, 56% of Obama’s sub-$2500 donations were under $200 (as opposed to just 24% of Romney’s). There are many factors at play here, but a big one is the images and headlines used throughout each candidate’s sites. Obama’s analytics team ended up presenting his campaign as a grassroots movement that the public and Obama were equal members of. Photos of Obama amongst the people, addressing the public, and phrases like, “Help build this movement”, are common throughout his site. Analyzing this through the lens of priming, he made his campaign feel larger than just himself. By making a donation you’re part of a greater effort to help the U.S.
In contrast, when we analyze Romney’s site with priming in mind, the imagery and tone is detached from the public, and so is the individual user browsing the site. Phrases like, “help Mitt Romney turn around America,” disconnects the user from Romney’s campaign goals. Photos with a tight crop of Romney against an American flag further reinforces this feeling. Consequently, the feeling of only helping Romney, rather than the U.S. as a whole, may have contributed to reduced donations.
The lesson here is your tone and imagery affects users. Think about whether the tone on your site is positively or negatively influencing your audience. This can be hard to know upfront, but testing alternate phrases and images will provide insight. A great place to start is conversion funnels. First, go through your funnel and take note of the images and tone of the copy. Are there images throughout the funnel? If not, add some that keep your site’s goals and message at the forefront of user’s minds. If you do have images throughout, do they prime users in the right way? Are they consistent? On e-commerce sites, it’s common for final checkout steps to be mostly devoid of imagery and copy. Increase engagement by adding imagery and text that represents your organizational beliefs. For example,Mailchimp does a great job of this. This is also a great use of Optimizely’s multi-page testing feature.
Knowing what responses people will have to different images and words can be difficult to predict. This is a good list of different priming effects that have been studied to give you ideas. One last example: this fascinating study showed that a proposition to increase school funding received significantly more votes when the polling station was at a school than when it was at other locations. Imagery of a school triggered school-related norms, such as supporting children’s futures, which subconsciously influenced people to support the proposition.
Additionally, here are some specific recommendations for tests that can be run on individual pages using any decent A/B testing tool:
- Switch out stock photography for actual photos of your product, service, users, etc. – whatever’s appropriate. People know when images are stock photography.
- A photo is worth a thousand words – try adding photos to pages that lack them, or increasing the prominence of existing imagery.
- Use photos with people in them. These are well known to increase conversions. 37signals has a good write-up of results from testing people on the homepage of their Highrise marketing site.
- Make sure your imagery and copy use a consistent tone throughout your site. For example, use a consistently positive or negative tone throughout.
Priming is a subtle effect, but one that can make a big impact on your visitors’ behavior. By priming people in the right way you can increase engagement and conversions throughout your site.
For more on priming, read this Usabilla post: http://blog.usabilla.com/6-primes-that-increase-your-conversion/
I’m making a board on AB testing best strategies and techniques/ uses. I like this post and I’m going to add it to my board. Here are some other resources that I came across at a meeting at the Hive in Microsoft’s silicon valley HQ. http://www.verious.com/board/AKumar/the-hive-a-b-testing-information-session/
Hi Jeff,
Great post as usual. I would second that in my testing experience the use of “real images” in place of stock photography does have a significant impact for the positive. I think seeing photos of users who aren’t quite so “done up” or a photo that doesn’t have the perfect depth of field makes us feel more connected to the image because it feels more real.
Jeff – thanks for the quick and thorough response. I think i understand the primary goal of the post. What I don’t fully understand are the examples.
I think my main issue was to point out that you were comparing Optimizely testing vs. Optimizely testing and that was never pointed out in the blog post. The blog read as testing vs. no testing.
While all tests are obviously different, you stated that your “imperfect” data led to the “best approach” for the Obama team because they tested and then an assumption was alluded to that we did not test. Bottom line was both campaigns were using Optimizely and making decisions based off the results they were receiving from that technology.
Thanks again,
Ryan
Good point – the fact that both sites got to their final versions through testing is actually incidental to the post’s content. And you’re right that I specifically said Obama used testing, whereas I made no mention of it for Romney. I amended the post to correct this. I don’t know what the data told either campaign, or if it led them to their differing tones of voice at all! But it was still interesting to analyze the differences.
Jeff – To clarify, the point that you are making here is that although at the Romney Campaign your Optimizely tools told us what images, copy and layout worked best on that donation page, we should have done something else instead?
Also – the Obama team must currently be seeing different results based on their most recent donate page.
donate.barackobama.com
Hi Ryan, good question. To clarify, I used the Romney and Obama campaign websites as comparisons of two sites with the same broad goals, but different imagery and overall tone. There’s no “best” approach that’s guaranteed to work for everyone. Based on my (imperfect) data, Obama’s approach seems to have resonated more with site visitors. But there’s a lot of channels outside of the website that affect people’s donations (such as ads, speeches, emails, etc.), so it’s hard to isolate the effect of priming. That being said, there’s no guarantee that this tone of voice would have worked on Romney’s website. But it’s worth considering how photos and word choice subconsciously influence users, and testing alternate approaches. I don’t have access to Romney’s testing data, so I don’t know what the data told them. If you have any knowledge of this I would love to hear it.
And thanks for posting Obama’s newest donate page. I find it interesting that they changed it so drastically since the previous one was so heavily tested. In my opinion, this new one looks worse and I doubt it has a higher conversion rate than the previous one.