In a recent InternetRetailing webinar, Learn how Shop Direct Measures, Optimises and Validates the Performance of its Website, Luca Collacciani, senior director of web experience EMEA at Akamai Technologies and Marc Farrell, site performance specialist at Shop Direct debated how to achieve the great web and mobile experiences that effectively engage consumers.
• Luca Collacciani started the webinar with a brief introduction to Akamai and the importance of performance. “When our website is performing well it’s good for our customers and good for us,” he said, before bringing in Marc Farrell of Shop Direct.
• Marc Farrell introduced himself and Shop Direct: the UK’s second largest purple retailer (Very largest brand, Littlewoods oldest brand), with 1.2m average daily visits, £1.9bn annual sales, with 73% sales made on mobile. Shop Direct now fully optimised for mobile.
Shop Direct on site performance
Choosing the right metrics
Moved from page load time to dom content load time (when the initial HTML doc has been fully loaded and parsed), first byte time (connection time) and HTML download time: gave a better insights into when the customer believed the page was useable.
Future metrics: start to create customer timers on key pages, used as time stamps to show when hero image loads, or an icon on the page. mPulse continuity plugin (currently in beta) that will give new insights into interaction measuring, for example, delays in interaction and rage clicks.
Measuring performance: how do you manage the complexity?
Use RUM tool (Akamai mPulse) as well as synthetic monitoring (Catchpoint) which is not as noisy as RUM.
Where did you begin with performance budgets?
• “We’ve always set performance budgets with our creative teams, to ensure that marketing banners are optimised as much as possible.”
• Set targets on specific banners: makes creative teams aware of the impact they can have by serving up images that are too large.
• Biggest change was setting targets for key pages, across channels. 32 pages that wanted to get below 2s dom content load time.
• Golden rule = less requests the better.
What worked and what didn’t.
Managing peak events
• Implementing HTTP 2.0 (H2).
First aim was multiplexing: ability for browser to start calling in all requests at the same time. •
Also interested in server push: get most critical content in front of customers asap. Know that’s difficult, and to distinguish what to push and when to push it.
• What did we find out about H2? After implemented it ran experiment with varying levels of concatenation. Found that still higher level of concatenation worked. Benefits; smaller, newer browsers but will that first variation really worked for us. A few things we learned from H2 after enabling and testing; those hacks should probably stay where they are. Less is better. Rlevant for the time being at least.
• Not all negative; new browsers better at leveraging the benefits. Different websites seeing different benefits, because of the sheer size and scale of change. Biggest change for 20 years, some time to get right. If you’re one of those sites that sees 10% to 20% in improvements when you switch on then go for it - and share with us so we can learn too.
• Testing important, monitoring, to ensure results we get are the ones we expect. What we’ll do with pretty much everything. Ensure changes we make benefit our customers on our site. Might work differently for other people but we’ll keep experimenting.
• Lessons from H2 - and what Shop Direct will do next.
• HTTPS: will boost security. Additionally, modern browsers restrict functionality such as push notifications and geo-location via HTTP: benefits of H2 only available over HTTPS.
How do you create a strong culture of performance?
• Aim is to get people to care, ensure people know performance is important and not just a job for IT.
• Cross functional product teams. Teams growing, changing too. There are more people and projects to affect site performance, but it also means we can do a better job. Really want to simplify site performance, provide with metrics and KPIs.
Things we know about performance
• We know it’s important. When Amazon famously said 100ms of latency cost them 1% of sales, 10 years ago, it felt like a real benchmark for performance. In 2016, Trainline said that by reducing latency by 3ms, customers spent an extra £8m a year.
• We wanted to know what it cost us, page browsing etc. What second actually cost us that money - is it 3s to 2s, or is it 7s to 6s. Led to a series of experiments that we want to run and learn from, injecting a second into the page to see what it does to revenues. It’s the opposite way round - comes at the beginning of the customer journey, injecting a delay - running in 70 tests across nine key pages on all our brands, different pages of the customer journey. See where it impacts the most.
• We can guess at it but we want to know exactly, scientifically. By running this series of tests we’ll be able to determine the price of performance, use it as a trade off with new developments. Make performance enhancements elsewhere on the page and optimise in key areas.
Visit the Akamai webinar page to view the webinar in full, including the slides and the final Q&A session.