Scientific Strategy in Your Business.
I’m a planner. I like to think through all the options, the variables at play, and calculate the best path forward. It’s made me a terrible skateboarder. Too many unknowns and physics that are too out of my control. My terrible skills at skateboarding was not my point here. I want to think about the failures, and how we can analyze them with science.
Anytime I think about my work in strategy, design, development, etc, I think about it as a science experiment. What is my hypothesis, how am I measuring, and what variables can I control. For those of us that have been out of school for a bit, a hypothesis is a theory, the idea, the assumption. It’s not your goal. Sure it may lead to your goal, but it’s not the goal itself.
Let’s work on an example. Say you are launching a new product to market. You have a site, you have the product, you have delivery. These are all controlled. You have a year-end goal of 10,000 sales and 12 months between now and then. So these are all our controlled aspects of the experiment.
What is the hypothesis? Well, we need to explore how do we get 10k sales in 12 months or roughly 833 a month, 192 a week, or about 28 a day. So we have some tools here. We have ad campaigns, email marketing, word of mouth. So we have a goal, and we need the hypothesis. What if our hypothesis is: With 500 clicks on an ad, we will convert 25% of these to sales, or 125, over a 2-day test. So if we break that down, we see a clear number, 500 clicks (another control), a target metric response (25% sales), and the time frame (2 days). It’s safe to assume or suggest, that you have multiple experiments running at once. So as part of your experiment, you need to make sure you have control or filters on the results.
With the right tools, you will see the top half of the experiment in the ad systems, It can be FB, Google, etc. but you will see the 500 clicks, and the time (since we are limiting it to 2 days). The conversion is where we need to be careful. We have multiple tests running, so we need sale attribution to the source. Luckily most eCommerce platforms have ways to manage this. I’ll avoid diving deep into ads and tracking as this post is about testing.
So every time we experiment, there is data collected. In the case of an ad campaign, we can track all kinds of things, like impressions, total cost, cost per click, time on site, time on site per converted lead, etc. So all of this data gives us valuable insight. So what should you be looking at? Some of it depends on what you need to make new experiments. If you had zero sales from the 500 clicks, it’s a deep dive into time on site, how far did they scroll down, where did they fall off, how long did the site take to load, etc. Look for patterns.
Say it did convert 25%. Sucess right!!!! Okay so now what? Well, that’s only 125 of the 10k we want. So what can we learn? How long did the converted sales users spend on the site, and is it more or less than the other 75%? What patterns lead to success vs the patterns that didn’t? Where can we create a new experiment changing a few variables? What if we change the text, the color, the images, the call to action type, etc. Each time you test, it should be fueling you for the next round of tests.
I love this stuff, the hard thing is it does take time and money to work through each. So what do scientists do? They read the research papers of their peers. They dive into what the hypothesis was, the variables, the controls, and the results, and they make up a new hypothesis for their test. In the world of business, every product, customer, and brand is different. So naturally, every test is unique, but the fundamentals are often well discussed and tested. You shouldn’t be starting at zero, but you may need to fail a few times on your own to help establish a baseline of your research.
