Watchfinder calls time on AWS, shifts to Microsoft Azure
- Summary:
- The online/offline seller of pre-owned luxury watches is using the cloud-based platform to explore new technologies and new ideas
As the business has grown, Gill and his small IT team have developed in-house the vast majority of the software needed to run it. Today, the website receives around 1 million unique sessions per month and the company sells between 1,200 and 1,300 watches per month at an average purchase price per watch of £4,200. He says:
Quite early on, I realised that finding off-the-shelf software and adapting it to what we needed wasn’t the way we needed to go. We’ve written all our software, right down to the point-of-sale [POS] system running in our stores. The only bit missing in store now is the link to the credit card terminal, which should be coming on board in the next eight to ten months. In a company of our size [it has revenues of around £70 million] that’s generally unheard off but we like to do things differently and to challenge the norms.
For many years, the company relied on a dedicated server hosted by Rackspace, later expanding that to two servers. But that situation brought various issues: Gill didn’t always know how much capacity he needed to provision and the company was, until recently, entirely self-funded, with most of its cashflow needed for buying and servicing inventory. Gill decided on a ‘lift and shift’ onto virtual servers hosted by Amazon Web Services (AWS) for a few years, but has more recently revisited that decision, opting for Microsoft Azure.
A rich environment
The decision made good sense, he says, given that Watchfinder’s growing collection of stores contain a lot of technology: customers use Windows 10 touchscreens to browse stock as well as larger 42-inch screens and Microsoft Surface devices provide in-store staff with a mobile POS.
We built software that manages our service centre, we built the software that prices the watches we buy from the public and we’re starting to collect a lot of data everywhere, so suddenly, just running virtual servers on AWS was beginning to look a little restrictive. In February 2015, we started to reinvestigate our options: would we go all-in with AWS or see where Azure’s heading. Everything we’ve written is in .NET, so naturally, I was going to look at Microsoft’s technology. And when I popped my head above the parapet, and took a closer look at what the Microsoft Azure team was doing, I saw straight away that it was worlds apart.
I have a relatively small IT team - six of us support all internal infrastructure, engineer the platforms, run the website and everything else. If I don't have to run a server, I won’t. I'll utilise cloud technology. Azure's got the Web Apps platform, so all the websites I was running on AWS I now move into Web Apps and, at the last count, that's 13 servers gone from my management schedule. That would previously require around four to six hours a month just on patching, as well as other care and feeding machines need to stay up to date and secure.
Just as importantly, he says, Microsoft Azure provides a rich environment bristling with new technologies that will support some of his most ambitious future plans:
I’m beginning to look into big data and machine learning, so Microsoft’s work on Project Oxford, for example, is pretty interesting to me. All the things I’m planning to do would have cost us an absolute fortune on paying consultants, because we can’t know everything in-house, but with Azure, I can leverage new technologies easily and probably get around 95% of the way to where I need to be, purely with our own resources and skills.”
We already have algorithms built internally that price the watches, but we also have a wealth
of data on how quickly a watch sells, how many people view it, how long it takes to service it. We’re going to be using machine learning elements in Azure to work out what watches we need to buy, when we need to buy them and in what order the service department needs to service them, so that we can get the right product onto the site in front of the customer, at the right time and at the right price. But that’s just the first phase. As we explore these ideas further, we’ll probably start to think about recommendation engines, too.
All that exploratory work is opening up a wide range of learning opportunities for he and his team, he adds.
When someone comes to work for me, and when they eventually move on, I strongly believe they should leave in a better position, skills-wise, than when they joined us. That’s my job done. All the developers have their own Azure subscriptions and, if they want to, they can go and play with bot frameworks, bring me back their ideas and we’ll get a version into the platform. They can use Microsoft products or open source ones, like Kibana for live event streaming. They can use Azure to fire up Linux servers, configure them and have a play around. And when we come to productise our brightest ideas, we can make the decision to go platform-as-a-service or script it out ourselves. The best thing about all this, for me, is that we get to leverage the brainpower of thousands of people working for Microsoft to help Watchfinder do what it needs to do - and I find that mind-blowing.