How ThirdLove used AI for better personalisation
- 02 December, 2019 07:07
For most women, a bra is a very personal and rarely spoken about thing. So to use artificial intelligence (AI) for not only determining a better bra fit for women, but for personalising the marketing journey, could easily have flipped over the line from helpful to creepy.
And yet, US underwear brand, ThirdLove, has struck that careful balance, growing revenue in the process.
Since 2013, ThirdLove has been on a quest to connect women with comfortable bras that actually fit. After taking the measurements of millions of women, the company introduced the industry’s first half-cup sizing, and developed an algorithm using the data to match each woman with the right size for them.
While the innovation was astonishing, ThirdLove needed to convince shoppers to take a chance on a new and relatively unknown brand with one of their most intimate requirements.
While it had been doing the usual marketing, including personalisation and split, A/B testing, it knew it needed to do better. After a research phase, ThirdLove engaged Dynamic Yield for its AI-enabled personalisation platform so it could take a far more tailored approach to its customer journey from the top of the funnel.
Personalising to build trust
ThirdLove required a solution that would help them to tailor messaging and creative to different audience groups, continuously optimise landing pages for each visitor, and surface the best product recommendations for each visitor. Its director of product growth, Desirae Oppong, said while ThirdLove’s products speak volumes about the company’s understanding of customers’ needs, for many women who have been disappointed buying bras in the past, a higher degree of personalisation was needed to gain their trust.
Since implementation, ThirdLove has offered initial landing page experience for different audiences, adjusted homepage messaging and creative based on browsing history, and has been collating data from users via a quiz to better improve its offerings.
Initial results are encouraging, including a 23 per cent increase in average revenue per user, a 3 per cent uplift in homepage hero banner click through rate, and a 75 per cent improvement in completion rate for its Fit Finder Quiz for first time visitors.
Oppong told CMO this is the tip of the iceberg in terms of what the business is hoping to achieve as it navigates through the new AI platform.
“We inherently knew there was a lot of opportunity around personalisation for us,” she told CMO. “And looking at examples of personalisation across other brands, it has been proven to drive considerable value by really honing in on what a certain user needs. So that was one of our big reasons for looking at Dynamic Yield.
“We also needed an updated way of doing A/B testing, as the prior tool we had was not nearly as robust.
“The change has allowed us to really hone in on what is the right experience for different users on our site, and how we customise and tailor that in a way that's truly effective."
Oppong cited growing diversity of the user base over time. "When you start off as a very young startup, almost all of your site visitors are brand new. And it's a little bit easier to think about them as one collective group," she commented. "But as you get larger, that diversity increases quite a bit, and suddenly you find yourself with a healthier balance of new visitors versus repeat customers.
“The content each of those groups is looking for is completely different. And so we've found it's incredibly important to personalise and make that journey a little bit tailored to each of those types of users. And you're missing the opportunity to really showcase something that they're likely to be interested in if you don't.”
To this end, instead of dropping first-time visitors on collection pages with entire product catalogues, ThirdLove directs a portion of traffic to specific pages more tailored to the segment. Additionally, visitors who land on the Fit Finder Quiz landing page from Facebook campaigns but have already completed the quiz in the past are automatically redirected to a landing page highlighting more relevant content. Simply doing this resulted in a dramatic increase in purchases and improvement in overall campaign performance.
ThirdLove also leveraged its hero banner on the homepage to drive engagement by tailoring it according to the individual’s browsing history, and for new versus returning visitors across desktop, tablet, and mobile.
Those who had not previously browsed the site received more evergreen content and messaging around core products. Returning visitors were shown newly launched products. By optimising the homepage experience, ThirdLove saw a 6 per cent increase in hero banner CTR and a 3 per cent lift in conversion.
“What we do is, first of all, is we look to understand what challenge the customer is coming to us with. So in that way we're able to tailor our recommendations on what might be the right products to recommend," Oppong explained.
“Then we do filtering based on the size that we recommended or sizes that you've bought in the past because one of the interesting things about this space is they might not have a size that fits them properly, and so being able to showcase the products that we have that truly fit you is one of the important experiences for our brand and for our site.
“If you're doing personalisation in a way where you're using information people are readily sharing with you, then it ends up being helpful instead of something else.”
Up next: How ThirdLove is convincing women through data
ThirdLove's Fit Finder Quiz furthers ensure a good fit in what is a difficult category already without adding all the buying online impediments. In fact, 80 per cent of the women who share their information are recommended a different size than what they reported wearing, making it all the more difficult to persuade first-time shoppers to take a chance on new sizes and products.
But after changing the first-touch experience for new visitors to provide more foundational and introductory content for the start of her journey, ThirdLove saw a significant impact at the top of the funnel, increasing its Fit Finder quiz completion rate by 75 per cent.
“We encourage people to take it before they buy, because one of its primary uses is to really help find the right product more likely to fit. We've actually done a lot of testing to make it a more prominent part of the journey, especially for new users, because we think it is fundamentally important to getting that person in the right fit,” Oppong said.
“We were also doing A/B testing before, it has always been a fundamentally important thing. But the velocity wasn't nearly as high as it is now. Previously, we were running maybe one or two tests a month. Now we run more than a handful every single month and we are aiming to get even more than this. So that testing velocity becomes super critical for being able to learn quickly and understand what works and what doesn't work and confirm whether some of your hunches are truly right or not."
For Oppong, one things that can be helpful for thinking about a new user is what campaign are they coming in from and what have they seen before they get to your site. "We have done some testing where we try to send certain types of our traffic to specific landing pages to understand what is most helpful for a person at this stage in their understanding and in their journey with us," she said.
“The landing page testing was really helpful because we had a hunch some audiences were ready to shop and we should just help them buy that product as quickly as possible. But what we learned and realised through the process is that journey wasn't nearly enough and they really needed more of an introduction to the company and to what our products offer them and what problem they solve. This was truly beneficial.
“So we have test running on all of the major pages of our site. And some pages, like the homepage in particular, we actually have several tests running because they're the most important pages.
“You do actually have to be quite intentional about thinking through the right content, and how to personalise it."
ThirdLove also continues to learn there are a lot of customers just not quite ready to make a purchase yet and who don't come to the site to buy.
"It’s really is more of an experience they're looking for. This was one of the things we had a hunch about, but it's still surprising the degree to which it shows up in the data," Oppong said. “We now realise we have to take a step back and think through the lens of somebody who has no idea about this brand and is coming to you for the first time.
“Even looking at things like what traffic goes to what page have been strong drivers in increasing revenue per user, which is super exciting to see. I would say most often our biggest successes come from really understanding what the user is looking for and honing in on that.”
ThirdLove is hoping to drive the platform to even great heights, including taking on churn and retargeting.
“We've barely touched the tip of the iceberg so far. There is a much deeper journey when you think about personalising, especially in the content experienced by user," Oppong said. "A lot of that just takes time to build. The fundamental journey is just really different based on where you come from, and we have certainly some way to go before we truly get to one point.
“Of course, marketing channels and bringing people into your site in the first place continue to be fundamentally important. But this, plus personalisation, really go hand-in-hand with each other. The more you improve the experience, your revenue, per user conversion, all of these metrics, it also just helps your acquisition marketing perform much better as well. So every dollar you're spending, you're getting much more out of it at the end of the day.”
Follow CMO on Twitter: @CMOAustralia, take part in the CMO conversation on LinkedIn: CMO ANZ, follow our regular updates via CMO Australia's Linkedin company page, or join us on Facebook: https://www.facebook.com/CMOAustralia.