Back in 2009, a long-time engineer at Google was looking at their search results page and thought to herself “That shade of blue we use for links.. It seems a little… I don't know… off?”

And so she tested a couple of different shades. And a couple more. Eventually, Google tested 41 different blues to determine which one people were most likely to click. Turns out, people will click links more if the blue leans over a bit into purple, compared to blues that have a tinge of green.

You might laugh at the seemingly tiny change, but that small tweak had enormous consequences. Indeed, people started clicking on more links — and that included links on ads. Google says switching to a slightly different shade of blue earned the company an extra $200m a year in revenue.

That engineer, by the way, was Marissa Meyer who went on to be Google's VP of search, and later, CEO of Yahoo.

The Role of A/B Testing

A/B testing, of course, is a staple in a marketer's toolkit — especially those who sell things. For good marketers, increasing the conversion rate — even by fractions of a percentage point — is an obsession that can pay off in spades.

But how do you do it? What mistakes should you avoid? And where is the middle-ground between providing your customers with a rich brand experience and pushing them to buy.

Oddit Interview — CRO Audit

For answers, I recently spoke with Shaun Brandt from Oddit, which provides advice to e-commerce merchants on their product pages and web sites. I asked him if he could only change three things on an ecommerce website's homepage, what they would be.

SHAUN: The first thing is when you land on the page, probably the most common thing that we're constantly pushing to customers is when the customer scrolls past that first section: do they understand what you do? And you'd be surprised how many submissions and customers we get where you go to that first snippet of the homepage, where you're introducing yourself to the customer. And they don't tell them, they use marketing jargon, or they're very vague.

It sounds silly that it should be in that top part because there's obviously a whole homepage that you can do that in. But so many customers make that first impression and either leave or scroll.

So it really is critical to be telling them and in a really simple way, exactly what you do right off the hop.

I think the second thing is… when you come to a site, and you see a brand with no reviews, or no one else talking about it, it becomes really hard to build instant trust for a brand and build that kind of rapport with a consumer.

A lot of what we're working with brands on is how do we build that even if you don't have reviews?

So maybe it's as simple as saying, Hey, okay, you don't have reviews, let's reach out to some of your Instagram followers and get some opinions from them. And let's start building our own reviews or pulling in, you know, reaching out to customers more personally and saying, hey, we'd love some feedback. Really helping them figure out unique ways to build social proof into their homepage, even if they are a new brand that doesn't necessarily have a ton of that.

TOD: It's interesting, because I discount the reviews on a on a product or a homepage, if it's clear that they've been cherry-picked. I don't mind seeing things that are automatically inserted from TripAdvisor or something, because I know that there's a sense of editorial control.

Whereas the ones where reviews say “I love this product!” and then the review name is just “Dave.” You can't tell whether it's real, or whether they've just made that up.

How do you combat that if you were using your approach of reaching out to Instagram commenters and things like that?

SHAUN: [That's] where the design and UX layer comes in. I think it's a constant design problem of trying to make sure that they're actually real reviews. We would never recommend making up reviews.

I think Amazon is like the baseline of this, of trying to make sure that they're proven. I think that comes down to publishing what app that they were collected through, allowing users to access the database of them, rather than just saying, “Hey, here's a static image of it.” Anyone can photoshop a review in… So there's different layers and levels to making sure that they feel more trustworthy.

More About CRO and Oddit

Our full conversation was about a half-hour long and covered topics like:

• What the big broader shifts have been in how direct-to-consumer brands sell products
• The most common conversion rate optimization mistakes that brands make
• How to do ANY CRO when you're using something like Shopify where you don't have a lot of granular control over the design
• The surprising effect on sales caused by Apple shifting the search bar on mobile Safari to the bottom
• And much more

The full interview is on our Premium Podcast Feed.