Booktopia uses big data to test assumptions before implementing changes
Booktopia has always known that data is a valuable business asset. But in the early days the company was using very little, if any, of the data at its disposal.
Over the last decade, with more resources at hand, the Australian online bookseller has increasingly become a data-informed business, using tools like Emarsys, HotJar, Google Analytics, and Optimizely to make discoveries, decisions, and money.
A data-informed approach has contributed to Booktopia's growth from a AU$10-a-day side project in 2004 to a business generating more than AU$100 million in annual revenue in the last financial year.
The 13-year-old company -- which attracts 24 million website visitors and sells 4 million books, DVDs, magazines, and stationery every year -- is currently using data to personalise search results, optimise the checkout experience, and validate new features and functionality before broadly implementing them.
Wayne Baskin, CTO and deputy CEO at Booktopia, joined the company in 2007 and built some of its systems from the ground up, including its warehouse management system and content management system. He stressed the importance of testing assumptions about customers before making any UX changes, and told ZDNet that data should be used to discover customer truths, not validate biases.
Seeing red
Wayne Baskin, CTO and Deputy CEO, Booktopia.
Booktopia's marketing team had a theory that the small red flag used to signal an item is low in stock was subliminally discouraging the customer from making the purchase. The team's rationale was that the colour red represents an error and suggests that one should stop what they're doing.
The company decided to conduct an A/B test using Optimizely Web Experimentation, where 50 percent of its customers continued seeing a red flag on products that were limited in stock, while the other 50 percent saw an orange flag. The red-coloured flags were found to be more successful in converting customers, disproving the marketing team's theory.
"Both are very strong colours ... but if we had listened to marketing and just changed it from red to a colour they suggested, we would have seen a 20 percent decrease in conversion rate," said Baskin, who worked as a software engineer at GE Capital prior to beginning his tenure at Booktopia.
"We don't go into the reasoning behind it ... we just listen to the data, and make decisions based on it."
However, displaying stock availability -- whether 'in stock' or 'low in stock' -- increased the conversion rate by 16.9 percent and revenue by 6 percent.
"Our hypothesis was if we show a customer that only a few of this item are left in stock, that will encourage them to take action on their purchase ... We took it into testing and our hypothesis proved to be correct, so we implemented it to 100 percent of our users," Lara Atechian, UX lead at Booktopia, told ZDNet.
Shipping forecast
Booktopia has also pondered whether presenting the shipping cost one page earlier in the checkout process -- that is, on the shopping cart page -- would boost conversions. The idea was that it would increase the perception of transparency, even though the company clearly displays its flat AU$6.95 shipping rate on the header of its website.
"Pretty much every single ecommerce specialist out there would say 'be transparent with your customers and show them what the shipping [fee] will be from the beginning'," said Atechian, who joined Booktopia about three and a half years ago.
However, this theory was similarly disproved. Both the conversion rate and revenue per customer had decreased 7 percent and 7.9 percent, respectively, when the shipping cost was presented on the same page as the shopping cart.
"It's not a matter of hiding or showing [the shipping cost]. It's a matter of the location in which you [present the information] and the value it will add to the customer's experience," Atechian said.
"We didn't just follow what everyone else thinks is right; we tested it first and realised it wasn't right for us.
"Companies shouldn't try to shoehorn some other company's solutions without testing them with their users first. Even industry standards may not be best for your business."
Lara Atechian, UX Lead, Booktopia.
Iconic status
In another example, Baskin explained that a graphic designer who had joined Booktopia six years ago said at the time that the website's header was too cluttered, and suggested the icons that were used to showcase the company's multi-award winning status -- and build up social proof -- be removed. The designer said this would allow customers to find the search bar more easily.
However, after removing an award icon from the header, Booktopia saw a 2 percent decrease in conversion rate.
"If we were, say, AU$50 million in revenue at the time, that's AU$1 million less revenue just by removing those awards," Baskin said.
Baskin said these examples show the importance of letting go of personal biases and trusting the data, even if it doesn't seem logical.
"I think a lot of old-school marketing people or people from a store environment don't always trust data -- especially when you put two weird correlations together -- because they don't understand it. They strive for perfection and there will never be perfection in big data," Baskin said.
"You can take one data point and it could mean one thing, but if you actually look at the full picture -- if you combine different data points -- it could mean something totally different."
An unobvious correlation Baskin pointed out is when fathers purchase expensive AFL books, they're likely to pick up a few children's books to assuage their guilt for spending so much on themselves.
Atechian communicated a similar sentiment to Baskin, saying that it's important to stay neutral in the process of discovering customer truths.
"If you are trying to get items out of the data that you can action on, be neutral. Don't be attached to a specific idea in your mind, and then paint your data with that idea. You don't want to take it in the wrong direction," she said.
Atechian said over the last three years, as Booktopia became more "disciplined" and increasingly embraced experimentation, the company has seen a 40 percent hike in desktop conversion rate and a 50 percent hike in mobile conversion rate.
Angus & Robertson versus Booktopia
Booktopia's 13,000 square metre distribution centre in Lidcombe, Sydney.
In 2015, Booktopia acquired Angus & Robertson's Bookworld business from Penguin Random House, but surprisingly, Baskin said there was very little overlap between their customer bases.
"We realised that [Angus & Robertson customers] are slightly more loyal, and more loyal to a brand that's been around for 130 years like Angus & Robertson. They're not jumping from website to website looking for the cheapest deal," Baskin said.
After analysing both customer bases, Baskin said they decided to develop two different strategies for the businesses.
The former is all about "letting the customer drive the experience," Baskin said. On the Angus & Robertson website, a feature called "Bookshelves" enables the customer to list all the books they have read, are currently reading, or want to read, and that information is then used to deliver personalised pages and search results.
"Angus & Robertson is really using that data-driven approach, saying a lot of people have this in their carts, a lot of people have read this, a lot of people have reviewed this well -- and it's serving up results that way," Baskin said.
Booktopia, on the other hand, combines customer data -- such as purchase history, browsing habits, shopping cart content, and wish list content -- with recommendations from its community of book experts.
Interestingly, Baskin said the top 10 bestsellers on both sites are, a lot of the time, quite similar.
"What our customers are telling us and what our experts are telling us are the same. The data-driven approach is being validated by the experts on Booktopia, which is very interesting," Baskin said.
Despite becoming better at utilising data over the last decade, Baskin admitted that Booktopia has "barely scratched the surface". Two key challenges the company is currently grappling with are the sheer abundance of data and the shortage of data scientists.
"The problem used to be where do you store data, how do you store data? The cloud has made that easy ... We're using Amazon Redshift as the data store for our business intelligence team," Baskin said.
"But knowing how to work with large datasets, knowing how to interpret data, these are not skills they've been teaching at university until recently. It's a new skill and a self-taught skill that people are bringing to market.
"We just have to find those people."