Before I dive into this article, I need to throw in one caveat: experiments will likely fail many times before you find something that works. This failure may not impress your boss, so be smart about your experimenting.
The unfortunate reality of experimentation is that you get better by finding various ways that don’t work before we find one that does.
Sometimes we’ll strike gold right away, but this is seldom the case.
So if you want to make your processes better, be prepared to buckle up and see where the results take you.
Setting A Control: The Foundation Of A Good Experiment
If there’s one thing that can screw with a good experiment, its manipulating your control variable. A control variable is an aspect of the analysis you keep consistent throughout your various tests. The control variable could be anything you are looking to measure.
Recently my partner and I have been experimenting with various linking activities for some of our SEO clients. Our primary concern has been achieving a high return on our time investment. So for our experiments, our control factor is the duration of the test. How much time does each activity take to achieve our desired outcome? If we didn’t cap our experiment time at two hours, we wouldn’t have that answer.
The control variable is what makes your tests an experiment. Without it, you would have nothing to compare your results to or be able to note whether it was a worthwhile activity or not. When deciding on a control variable, consider what you are trying to measure. What would be of value in your findings? The chosen control will be your primary tool for comparing results across your various tests.
Documentation And Attribution: Impressing Your Boss With Concrete Findings
Once you have conducted your experiment, you need to document the results in a way that accurately communicate its findings. If you don’t record the results of an experiment, you have no way to express its potential value to your boss.
Here’s the thing – you have to be crystal clear on the findings of the experiment. Don’t sugarcoat the results, or pretend like it wasn’t a wash if that’s what it was. There is immense value in finding something that doesn’t work.
Just this week, my partner and I were experimenting with content outreach as a linking activity. We found we could earn valuable links, but it was far too inconsistent and took way too long to be worth our while. We spent two hours finding an inefficient way to get links for our clients.
So what did we do? We documented our findings, slipped it in the “doesn’t work file” and moved on. Next week we’ll try something different, and the week after something different still.
We love finding link building activities that don’t work because each time we get closer to finding one that does.
For an entire quarter of experimenting two hours a week, we have found three specific ways we can level up our linking activities. Actual concrete processes that are getting serious results for our clients.
For us, setting time apart each week for experimentation has been a game-changer, and I hope your experiments prove just as valuable.
If you’ve been leveling up your processes by conducting experiments, what have you found to be a valuable model for it? Anything you would recommend avoiding in an experiment?