The digital world runs on testing. Online retail managers revamp the look and feel of their website to encourage more sales. Online travel companies test the CTA buttons and their flight listings to guide their customers through the booking process. Financial services companies optimize their landing pages in order to increase the likelihood that a visitor will fill out their form and start the process of applying for a loan. Whatever the desired outcome–companies are testing each and every element and feature on their websites to optimize for the highest conversions.
Even the tiniest change to a website could have a huge impact on conversions and therefore online revenue. For instance, shifting the font, copy, and color scheme of the cart checkout page or product page of an ecommerce site might enhance the customer experience and therefore result in increased purchases and revenue. Rather than let politics or human bias reign in decision making, website managers and digital marketers are letting data lead the way via website testing to guide decisions about their websites. And everyone seems to be doing it—from the smaller online clothing brands such as Cosabella, to the leading tech giants such as Microsoft, Amazon, Booking.com, Facebook, and Google who each conduct more than 10,000 online controlled experiments each year. Controlled experiments can transform decision making into a scientific, data-driven process rather than an intuitive reaction.
Types of testing
For those who are still unfamiliar with testing, there are two main types of testing methods that are at the tip of everyone’s tongue–A/B testing and multivariate testing. The difference lies in the number of elements you can test–A/B testing can only test 2 variations per test while multivariate can test several elements simultaneously. A/B testing is ideal for testing large changes to the website or for situations with low traffic. Multivariate testing is good for testing smaller changes and testing many ideas at the same time.
Introduce AI to the multivariate method and you get a turbo-charged, super efficient optimization platform for website testing such as with Sentient Ascend, which we have dubbed a “massively” multivariate testing system. Sentient Ascend will take all of the different ideas on how to tweak and change the website and then, using evolutionary algorithms, “evolve” the different website versions to find the winner. After a multi-generational experimentation process, the platform finds the winning website version that possesses all the best feature elements that encourage users to convert.
Just look at how Clicksco won
Clicksco for instance, a global marketing technology business, used Ascend to help increase the conversion of its more than 50 different web properties from a number of different clients. Clicksco was able to test 1027 candidates and saw an average lift of 1.75% with some sites seeing as high as a 17.9% conversion lift. According to Paddy Johnson, Clicksco’s Director of Group Creative and Optimization: “Having a tool that lets us test all our properties in a single place, and test more of our ideas, faster means that our bottom line grows more quickly.”
More experiments are better
As the old adage goes, practice makes perfect. The same concept applies to testing–the more you test, the better your website will be. According to Harvard Business Review, the majority of tests fail. Even at large tech giants like Google and Bing, only about 10% to 20% of experiments generate positive results. A/B tests especially, have a high failure rate with only 1 out of every 7 tests producing positive results. So in terms of numbers, testing more is necessary just to arrive at a positive result. Furthermore, conducting more tests, faster allows companies to find out what works and what doesn’t work faster, allowing them to fail faster, pivot quicker, and respond to the shifts in their industry to stay ahead of competition.
Testing is Time Consuming
So why not just run more tests? Well, that’s easier said than done. Simply stated–testing is expensive and costly in time and human resources. In a recent webinar, a leading A/B testing solution advised web managers to hire more developers, contractors, and marketing professionals to be able to run more tests, stating that “tests with 5 or more variations were 75% more effective in terms of finding a winning result.” Hence, they would need to hire more people and run more tests to be able to test more than 5 variations.
Website experimentation requires a number of people involved in the set up and analysis of tests–marketing professionals, UX Designers, web developers and data scientists for instance–and if a company is using an A/B testing tool to test 5 or more variations on their site, they would need to manually set up and run each test for each change. This would be costly in time and money since the company is paying each person a salary for the hours they are spending testing. Not only this, but each test requires a number of weeks to find a statistically relevant winner.
Traffic is Expensive
Another constraint is traffic. Website experiments require a certain amount of traffic to be able to run the tests in real time. Traffic is like the fuel to a website testing’s engine, without it, it couldn’t run. Hence, this same A/B Testing platform advised website testers to buy or acquire more traffic–most likely from paid sources such as Facebook ads and Google AdWords–to be able to run more experiments and test more variations.
Seems simple enough. However, this approach of buying more traffic and hiring more people doesn’t take into account the limitations a company might have in terms of resources. For the small to medium sized companies who don’t have the resources to hire a whole team or spend hundreds of thousands of dollars on traffic like Google and Facebook, running more A/B tests may not be financially feasible. Luckily for marketers, AI-powered optimization tools like Sentient Ascend solves these issues by allowing testers to test multiple designs all at once and throughout an entire website funnel, while making the best use of that company’s human and traffic resources.
Scaling efforts with AI
In the race to capture online conversions, testing 5 or more variations on your site is key to be able to find significant lift. The more marketers test, the better their sites will be, and the more lucrative their online operations will become. Luckily, running more online tests doesn’t have to be as time costly or expensive as you think now that AI-optimization tools like Sentient Ascend can solve the issues of time and money and help you scale your testing efforts. With Sentient Ascend, you can test more variations, find out what works and what doesn’t faster, and pivot your experiment to better suit the needs of your customers. AI-optimization is the way of the future and the key to efficient conversion success.