When you realize you need to start AB testing, your mind runs a million miles an hour while your hands are stuck with the speed of your internet connection.
(Which even if you’ve got the highest speed in the world, still isn’t fast enough.)
You get struck with the brilliance of 6,000,000 great ideas of things you want to test, but since AB split testing only allows for one element at a time per an optimized landing page, you’ve got to scale things back a bit.
You can only do one thing at a time, so you have to try to choose the best things first.
You’ve got to figure out a way to prioritize.
While the initial idea of figuring out exactly what to start with and what to test next (or worse, which ideas to put on the back burner) can be intimidating and soul-crushing.
No one likes to see a good idea getting shoved to the side.
But with AB testing, it’s the nature of the beast.
Shoving things to the side is simply necessary for success.
And if you don’t do it, you’ll only create a bigger mess for yourself because haphazardly AB testing doesn’t do anyone any good.
So, how exactly do you choose what ideas come first?
Here’s our best tips:
1) Start Closest to the Money
By starting closest to the money, you start with changes at the end of your marketing funnel, closest to the purchase point that have the highest potential to make the biggest impact on your bottom line now.
By starting here, you can optimize from the inside out, knowing that each incremental change that comes about from AB testing will be as optimized as possible, so you won’t have to keep jumping around your funnel based on speculation and best guesses.
This checkout for a trial of a blogging course is a little intimidating.
It takes you from the website you were on to a totally different website (Clickbank) to complete the payment so you can buy the product.
I may have had some really excited feelings about the possibilities of this course before clicking through, but now that I’m on the page to actually hand over my money, the info-demanding and emotionless new website turns me off a little bit.
ModCloth, on the other hand, has a much more cohesive feel from their shopping pages to their checkout.
(Which helps reduce the checkout turn-off feelings.)
It also looks like they’ve already experimented a little bit with their three prominent, pink checkout buttons.
This would help to move more people from the idea of buying into the actual action of it…to the actual money-collecting page itself.
For example, as a new user to the site, I find the splash of five fields asking for my information a little intimidating.
For people like me who might be tempted to click off at this point, they might want to try a more interactive screen where I first only have to select whether or not I’m a new or returning member.
And then only show me the 2-3 fields of information they need from me, rather than slamming me with all five at once.
2) Define: Site Goals & Numbers
First, define your overall website goals.
For many businesses, it’s to make sales.
But for some business models that might exclusively sell via email lists, it would be email signups instead.
Then, decide what numbers best represent those goals.
For example, total traffic probably counts less than actual sales made.
(You might have heard of these numbers referred to before as KPIs. But I think that’s boring industry jargon that makes me want to stick sharp pencils in my ears.)
If those numbers (KPIs for your boring marketing folks) aren’t where you want them to be, decide where you want them to be. (i.e. Set your goals.)
Try your best to find out why those numbers aren’t where they should be, and prioritize your AB testing ideas based on your hypothesis of which one will have the most impact.
For example, if you think adding user-friendly descriptions to your checkout process has the potential to increase your sales by 5% and changing the color of the buy button might have a difference of 0.2% to 0.3%, for goodness sakes people, go after that whopping 5% first.
Save the measly color changes for later.
And be realistic with your hypotheses here.
This is no place for infinity and beyond Buzz Lightyear-style counting.
Hint: one of the best ways to find out what your problems are is to talk to your visitors to find out.
Surveys work well to help you see where your funnel is lacking education to help move prospects from point A to point B, for example.
3) Make Sure Your Most Common Exit Pages Are Actually Meant to be Exit Pages
No, your visitors are not going to stay on your website forever and ever until the end of time.
At some point, they will bounce off.
The bouncing off isn’t necessarily bad, as long as it’s happening on pages that you’ve specifically designed to be exit pages on your site.
You can easily figure out which pages are most often used as exits in your Google Analytics, and see where you’ve got a problem.
Just go to Reporting -> Audience -> Users Flow and hover your mouse over the red colors to see drop-off rates.
So, what’d you find?
Do people lose their cool after reading your obnoxious history on your about page?
Or is there a page where the CTA is nearly invisible so people bounce because they don’t know what else to do?
If you’ve got an exit page that isn’t supposed to be an exit page, try to use a heat map to figure out exactly at what point most of the bounces are happening, and test some key changes at those points on the page.
For example, right now we’re collecting signups for an AdWords training webinar we’re hosting with Kissmetrics.
This is an example of a page that we DO NOT want people to bounce off of.
Particularly if this were an on-going offer and we noticed a lot of bounces from this page, we’d want to do some AB tests to fix that.
This is an example of a page (after someone signs up for the webinar) that we’re totally okay with people bouncing from.
In fact, it’s designed to be an exit page.
So if Google Analytics shows us that this is one of our top exit pages, we know we’ve got very little to worry about when it comes to this particular corner of our website. (You know, beyond what we do post-conversion. But that’s another can of worms entirely.)
4) Create a Hypotheses for Your Fave 3 Ideas Per Page…
If you’re beyond optimizing things right next to the money and you’ve got your exit pages in order, it means you’ve got all your ducks in a row.
But, it doesn’t mean you’ve maxed out your conversion rate potential quite yet.
You might be trying to optimize your home page, for example, which probably has zero calls to action that are directly buy-based.
But just because your visitors aren’t making direct purchases doesn’t mean your homepage shouldn’t have an positive impact on your bottom line.
Even when you move away from testing right next to the money, keep the money in mind.
Make your hypotheses around how a certain change you’re thinking about will affect your bottom line (because ultimately an increase in revenue is more important than in increase in click through rates).
For example, if you’re looking at your home page, you’ve got loads of options.
Depending on your product, you might decide to show a video with a strong call to action for a free trial, or lead people through an interactive questionnaire like Oscar does to directly show visitors purchase options.
For a SaaS company, you might hypothesize that the free trials will have a better long-term impact on your bottom line than directly presenting people with plan-based pricing options, so you’d want to test the video first.
5) Address That Sudden Drop in Metrics Right Now
If you’ve all of a sudden noticed strange things going on with your heat maps, lower than normal conversions at a certain point in your funnel, or see anything else that’s a little funky and throws up a red flag, don’t ignore it.
You’ve got a problem that needs to be addressed ASAP.
So as much as you might be in love with the idea of the amazing test you’re going to do close to the money to increase your revenues, those tests aren’t going to be worth anything if you’re not funneling the traffic to the points in the funnel that are close to the money.
Hint: Sometimes, drops just happen.
If you’ve got an overnight drop that came out of nowhere, try to weather it out for a few days to see if things get back to normal.
You want to act quickly on red flags, but you don’t want a rash over-reaction that’ll take you even longer to recover from when nothing was really all that wrong in the first place.
When your numbers suddenly plummet, don’t kick them while they’re down.
Do something to get them back up again.
6) Let a Chart Do All the Thinking and Prioritization For You
I’m actually not kidding on this one.
Optimizely put together an absolutely incredible chart to help non-seasoned (and seasoned) AB testers cut through the crappy AB testing ideas to get to the real money shots.
And if you ask me, it’s too damn good not to share.
It’s a simple points system where each potential test either gets one point or zero points based on whether or not it fits certain criteria.
If you’re new to AB testing or would just rather not spend your brain energy picking through all of your teams crummy ideas that they think are amazing, run them through the mill of this chart to find the true winners.
Bonus: ConversionXL developed their own prioritization framework. Check it out here:
The PXL Framework benefits:
- Makes any “potential”or “impact” rating more objective
- Fosters data-informed culture
- Makes “ease of implementation” more objective
Cautionary Tale: AB Testing May or May Not Save You
While using an AB testing guide is an incredible way to improve your conversion rates, it’s important to remember that not every test will have a huge statistical significance.
Believe me, I’ve been there.
I once thought the end-all be-all solution to one particular funnel drop-off would be to add an eBook.
Even though it got lots of downloads, it hardly made a dent in our conversions and bottom line, so I had to go back to the drawing board.
So don’t get down on yourself if one of your “grand ideas” turns out to be a moot point.
It happens to the best of us.
Prioritizing is important, but prediction is impossible.
That’s why testing to actually find out what is important is so crucial.
P.S. Did you like what you read here? Gained some insight? Tweet and post this to your peeps to share the wealth.