Business Essentials

A/B testing for beginners - Brought to you by NatWest

A/B testing for beginners - Brought to you by NatWest

Thursday, 21 May 2020

If you’re not trialling different versions of your website or experimenting with your digital marketing campaigns, you’re almost certainly missing a trick.


A/B testing in its most basic form is not a new concept, but where once this might have meant experimenting with new signage or different window-dressing options, in today’s fast-moving world of digital marketing, A/B testing can – and probably should – be applied to almost every public-facing aspect of a business.


Neil Wright, founder of travel insurer CoverForYou, explains just one example of A/B testing in action: “Let’s say we’re sending out an email to loyalty club members. We’ll make two different versions, and send ‘A’ to 5% of the people on our list, and version ‘B’ to a different 5%. We’ll track how people open or react to them, and then we’ll send out either A or B to the remaining 90%, depending on which performed better. We use A/B testing for everything because it’s a great way to help us improve how we’re communicating with the customer.”


As well as helping business owners work out which version of their mailout is most effective – everything from length of newsletter to subject line to time of delivery can make a difference to engagement rates – A/B testing can show SMEs how they can make improvements to their website, Facebook ads and more.


Understanding what ‘better’ means

In many cases, you will begin with a ‘control’ – your website as it stands today, for example. The variations you then test will let you see if these help your site to perform better than before.


“Before you begin, though,” says Andrew McCaul, co-founder and creative director of creative digital agency The Bigger Boat, “you need to know what ‘better’ means to you – for example, if it’s a B2B business, it’s most likely to be an increase in lead generation. You may want to ask: ‘Does changing the colour of this button mean that more people have inquired?’ That’s a great measurement for a test, but there are all sorts of things you can measure.”


Where do you start?

The simplest way of A/B testing a website is to look at your stats for the past month, make some changes, then see what differences these make the next month. This is entry-level stuff, however, and is a rather imperfect science because trends, competition, fluctuations in organic search and a multitude of other factors mean no two months are the same.

“When you launch a site and never change anything, you’ll never know if there might be a better way”

Andrew McCaul, co-founder, The Bigger Boat

A better way to do it is to make use of a free Google tool called Optimize, although McCaul cautions that this does take a little technical know-how and may be beyond the capabilities of the average time-pressed SME owner. “It allows you to make slight variations, such as showing different images or using different colours and headlines, and these versions all run concurrently,” he explains.


Once Optimize has gathered sufficient data, it will show you which of your changes are having the most impact – vital information in the neverending quest to make a website a little bit better than it was yesterday.


Testing becomes the norm

When it comes to digital ads, Farhad Divecha, founder and MD of multilingual digital marketing agency AccuraCast, says that A/B testing is within everyone’s grasp.


“In fact, with Google’s new ad format, you almost can’t avoid A/B testing,” he says. “They ask you to write three to five headlines and two or three description versions, which they automatically rotate and then let their machine learning system decide which ones get shown more often – based on click-through rates, conversion rates, and so on.”


Facebook is simple as well, he says, and allows business owners to upload multiple images for ads, before giving prominence to those that perform best.


Small changes to digital ads can make a big difference. “Last autumn we were A/B testing ads for a client and saw that those that mentioned the word ‘Christmas’ saw a heck of a lot more engagement,” says McCaul. Things like this might sound obvious, but they’re easy for busy SME owners to overlook. Another client of The Bigger Boat once had a contact form that, upon closer inspection, had far too many fields to fill in. “We tested halving the number of fields, and conversions through that form tripled,” says McCaul.

Hiring digital marketing experts who are familiar with A/B testing will almost certainly result in improvements to your website and marketing efforts, though Divecha says it’s OK to start small and try the DIY route, too.


“I recommend SMEs begin with whatever tools and options their existing providers have,” says Divecha, who adds that the aforementioned Google Optimize is one of his favourites. “It does require a little bit of code integration,” he says, “but it can be done quite simply and allows you to test multiple elements all at once.”


Ultimately, A/B testing will nudge you towards incremental improvements that will be beneficial to your business and, say the experts - you’d be mad not to try it. “If you don't test anything, you don't learn anything,” says McCaul. “When you launch a site and never change anything, you’ll never know if there might be a better way.”


Five tips for better A/B testing

  1. Don’t over-react: Make sure the data you glean from A/B testing is meaningful, cautions Divecha. “Don’t make decisions based on just a few clicks.”
  2. Log your findings: “People run all sorts of tests, but if they don’t log the results they’re in danger of not remembering what worked,” says McCaul.
  3. Try something radical: Tiny changes may yield tiny – though still important – benefits, but with A/B testing you can also try dramatically different things, too. “When you want to make a big jump, you have to make big changes as well,” says Divecha.
  4. Be the customer: Says Wright: “Try and remember that visitors may not be as familiar with the processes on your website as you are, and use testing to make their journey easier.”
  5. Be careful with your prime ‘real estate’: When testing tweaks to a website, Divecha says you should avoid tinkering with your high-volume pages as this may adversely affect organic search. “Test somewhere else and then roll out the final version to your high-ranking pages,” he suggests.


Our Partners:

Sponsored by Specsavers Sponsored by NatWest