"If it ain't broke, don't fix it"
- Thomas Bertram Lance
Humans have a distinct ability to solve problems. Where this becomes detrimental in practice is when we try to solve problems that do not exist. This is especially prevalent in the digital marketing world. Most notably from my experience, this surfaces most often when it comes to split-testing practices.
In this article, I'm going to lay out the top 3 ways sales and marketing funnels go sideways through making a few small but catastrophically crippling mistakes can be hard to reverse.
Let's jump right in.
Split-Testing Sin #1: Testing The Wrong Things In The Wrong Order
First things first... do NOT, under ANY circumstances... launch a split-test on something that is working (unless you hate your funnel). I have unfortunately done this so many times myself out of greed for higher conversion rates, and it has ended badly every single time.
I have also been part of a project where something was 3x'ing on the front end (I.e. without any backend or high ticket products/services) and we split-tested it for 8 months into oblivion, only to revert the funnel back to the original control.
Save yourself the time, and resist the urge to run a test for the sake of running a test... or trying to go from a 40% optin rate to a 45% while your checkout page converts at 20%. Greediness for an extra percentage of conversion will almost always lead to lost conversions more often than gains. Which brings me to my next point...
Always optimize the point of most friction first. You should analyze the individual conversion of every step of your marketing or sales funnel and have clear benchmarks to what those are. Whatever metric is the lowest... let's say it's your checkout page has a KPI benchmark of 50% and your checkout page is converting at 20%. That metric should be like a red spinning siren going off indicating THAT STEP is the step you should be split-testing to improve.
Pro-Tip: If you try beating a control multiple times and it wins every time… just stop trying to beat it. Instead, test a completely different page in isolation while the control brings home the bacon.
Split-Testing Sin #2: Deploying Variants Too Early
However long you think you should run a split-test... you should probably run it longer (assuming that your split-test variants are not dropping overall performance of the funnel). Here's why...
Most marketers ignore the velocity of their incoming data (I.e. how many visitors they're getting on a daily basis to their funnel/site) which dictates how short or long a split-test needs to run, and at what percentage of traffic.
But without getting too complicated, all you really need to know is that even the most promising split-tests can be proven to be statistical anomalies with enough runway. You absolutely must let each test run long enough than you to sufficiently validate the conversion economics and sequential impact of each step of your marketing or sales funnel.
Note: Sequential impact is the resulting changes in conversion rate to the following steps of your funnel of the test you're running (e.g. split-testing your optin page can change the purchase conversion rate and/or call booking rates).
Split-Testing Sin #3: Not Leaving Breadcrumbs
The final and most debilitating mistake marketers make with split-testing is not keeping a detailed log of their tests. If don't leave a trail of crumbs, you'll never find your way back to the cookie (or in this case your control).
You can absolutely optimize your way OUT of performance gradually over time if you’re not careful. This happens way more often than most realize, and if you've ever split-tested a webpage or funnel page more than twice, I can almost guarantee you've lost a variant (and the data of that variant) to prevent yourself from repeatedly testing the same concepts over and over again.
And the worst part about failing to keep a detailed log of split-tests, it makes it extremely difficult to retrace your steps back to a place where you were profitably converting visitors into leads and customers.
Here's the top 3 things you should be keeping a log of:
A full-page snapshot of the variant in colour (I.e. version history)
A description of what the variant is testing and a hypothesis of the test
The outcome of the test
Remember to always have your original control somewhere safe that you have easy access to at any time to roll back your changes.
You are now armed with the knowledge of the top 3 split-testing sins that can save you years of headaches and thousands of dollars of wasted ad spend.
A Note On "Conversion Contrast":
One important thing to remember (and sort of a bonus #4) that trips up most well-to-do marketers is that they fail to provide enough of what I call "Conversion Contrast" in a test.
If you're conversion rate is significantly below your target... no button colour change is going to save your funnel. For big changes you need big differences (I.e. contrast) between the control and the variant.
One of my favourite ways to apply this is by having drastically contrasting the big idea on the sales page or even testing different promises and angles.
But, Conversion Contrast is an article for another time.