We’ve tested out some big ideas here at ShareProgress — just last week, we told you about the time we sent AFL-CIO members an email about Robot President Scott Walker, and we’ve written in the past about the Facebook tests we’ve run on petition images. But we’re also constantly tweaking and testing our own website to see which things work well, and which could be improved a little less dramatic–but can add up to some pretty impressive gains over time.

We recently added a new Data Scientist to the ShareProgress team, Sandhya Anantharaman, who’s been running tests on our share pages. We were chatting recently about how the testing process has been going, and she told me that working on testing our pages changed her perspective on what can be tested, and what’s worth testing. I was curious about what that meant, so I asked her a few questions about her experiences.

What kind of tests did you initially want to try on the SP site?

“The first time I saw the ShareProgress share page was during my first week here. I took a summer internship with ShareProgress to learn about data science, but I had no idea what I’d be doing. Andy, our lead Data Scientist, sat me down, showed me the share page, and explained that I’d be spending a large part of the summer testing ways to increase the number of people that were sharing.

“The ideas that came out of my first brainstorm were all over the place. When I thought about testing, I figured I had to make large radical changes to the site to make a difference. I wanted to rearrange everything on the page, use big images, and add a paragraph to the page about why sharing makes a difference.”

How did you realize that those ideas wouldn’t work?

“First of all, there were some practical barriers to the changes I wanted to make. Basically, I didn’t have the CSS or Javascript skills of our excellent front-end developer Justine, so I couldn’t quickly make large changes to the website without making a lot of work for her. However, it turns out that this wasn’t a bad thing. When I brought my list of 100 big test ideas to Jim, I found out that most of them were actually likely to discourage people from sharing.

“I wasn’t exactly thinking about what kind of mindset people were coming to the share page with, or why they wanted to share. I was thinking about what I thought would make the site more attractive, which many times isn’t the best measure of a successful page.”

What kind of tests did you come up with after that?

“After I realized that I was going to have to write the JavaScript for the tests myself, I actually just started with changes that I could make on my own. The first tests involved mostly small, aesthetic changes–tweaking button language and color, changing border width, rounding corners, and changing the size of text. After I learned more, I experimented with getting things to change color when you hover over them and moving some small elements around on the page.”

What kind of results did you get?

“Honestly, the vast majority of tests didn’t make a difference at all. At one point I was a little depressed about the results, because I was worried that I was going to look bad because I couldn’t produce results. But as it turns out, I learned a lot about the way people use our site from all of that seemingly random testing. There were a lot of changes that I made (see 80px border) which I thought looked crazy and would make a huge difference, that literally had no effect. Then there were a few things, like changing the header text in the share boxes, that ended up producing really big gains.”

Two different versions of the ShareProgress share page, in a test to measure the effect of increasing the thickness of the surrounding border

Do you have advice about testing small changes?

“Test everything, even the changes that you think are really small and inconsequential. Most of them won’t make a difference. However, you will end up learning a lot about your users and what they notice on the site. Even with the changes that produced results, I often had to go through several iterations of those tests before I found variations that actually increased engagement. The best advice: never stop testing!”

Are you running small tests on your social sharing that are getting results? Let us know! We’re always eager to hear what works (and what doesn’t!), and what you’ve learned from it.