Using the Scientific Method in Web Development

Reading Time: 3 minutes

I just finished up reading Make Time, a book about making time for the work that matters, and one of their tips was to use the scientific method for fine-tuning your days.

It broke down into observe, guess, experiment, measure the results, and repeat as needed.

I haven’t formally studied the scientific method since school, but it jumped out to me that a lot of what I do with development follows the method.


The first step in the scientific method is to observe a problem.

In the sense of web development this could be a bug that a user reports. Typically the first step is to see if we can recreate it and observe what is happening.

For the sake of this post let’s say we have a client that comes to us with a site that isn’t converting as well as they’d like.

We can observe their current conversion rates and review the site to see how it is currently setup.


Next we’ll make a guess, or more formally hypothesize, as to what is causing the observed issue. As you gain experience in the field you’ll get better at this step. The more experience you had to draw on the more focused your initial guesses will become. Until then it’s perfectly fine to resort to Google even experienced developers do a lot of Googling.

In certain situations we can make educated guesses with other tools. Let’s say as we review the client’s site the design and user experience looks good, but we notice the page performance leaves something to be desired. Pages are loading slowly and that can really hurt conversions.

So we can hypothesize that by decreasing the page load we can increase conversions on the site.

We can finetune that with existing knowledge and research done on page speed and conversions.

A 1 second delay in page response can result in a 7% reduction in conversions.

Neil Patel – How Loading Time Affects Your Bottom Line

We can also validate our hypothesis by using performance testing tools like GTMetrix, Google PageSpeed Insights, and Lighthouse.


Next we can build an experiment to test our hypothesis that improving page performance will increase conversions.

Essentially this comes down to building the feature and implementing it. In the case of page performance this could be something like optimizing images across the site. That could be done in various ways such as implementing lazy loading, optimizing images, or advanced techniques like using object-fit to serve responsive images in place of background images.

Each of these could be implemented and tested independently to determine how they effect the overall goal of increasing conversions by reducing load time.

At this point in the process we’re building out our experiment and how we’re going to measure the results.

Measure Results

Speaking of measuring results, that is the final step. Clients typically want the results a website gives them rather than just having a website. So being able to measure results can set you apart.

Luckily the web has many ways for measuring results. Tracking conversions is easy, you can use analytics to view the number of users visiting your site against those making a purchase and compare it to other points in time.

A/B testing is another great way to measure results. It can be a bit complex to setup but there are plenty of tools out there to let you manage this easily. This comes in handy if you’re experiment involves testing how two different page layouts convert users.

Another option is good old user testing. Getting feedback from users that are on your site can provide great insight into how people are experiencing your site. It can be hard to do on your own, but tools like UserTesting can make it a little easier to get real feedback.

Do It All Again

The web is constantly evolving. To stay at the top of the game you need to be consistently iterating on your projects. Observing issues, making guesses, building experiments, and measuring the results.

Pin It on Pinterest