Shop Direct’s Paul Hornby at CAS EMEA 2016

Good morning all. As Mahesh said, my name is Paul Hornby, and I’ve had to be commerce shop direct. Now contrary to what this photo would make you believe; I am also not the single most miserable man in the Western world. You gotta love your corporate come is brother. On a serious note, I’m here this morning to tell you a little bit about the personalization journey that we’ve been on over the past 12 to 18 months. Before I do that, though, I just like to bring shutdown, direct to life a little bit and tell you a little bit about our history. Shop direct historically was a catalogue retailer, we were actually made up of two big competitive retailers, which was Littlewoods, who were based in Liverpool, and the Goss Home Shopping group who were based in Manchester. Now those guys were big rivals, and they were brought together about 10 years ago now to form shop direct. Over the past 10 years, we’ve had a massive, massive business transformation. I think 10 years ago, when these organizations were brought together, we sent out 25,001,000 Page catalogues every single year. And I bet around 25% of our business online. At that point, we still had bricks and mortar stores, we had lots of mail order, we had lots of orders being taken from the telephone. But our focus was to try and move towards a pure play retailer model. Fast forward to today. And we’re the UK second largest pure play retailer. We trade across two main power brands are very cold at UK. And that woods.com. We also have a business over in the Republic of Ireland, which is called little diamond. And exactly a year ago this week, we launched very exclusive which is enjoying its first birthday as we’re here now. Very exclusive is a small female only accessible luxury brand, which has done really well and is offering a different proposition to the very Dakota UK brass. Very has been a massive success story. So very now it was launched in 2009. It’s an over 800 million pound business on its own, and is still growing at around 20% year on year, and has done so for the last few years. So we’ve seen really phenomenal growth, which is powered the overall group to total annual sales of 1.7 billion. It’s also worth mentioning that 63% of all of our sales now come through mobile as well. Centered around that growth has been a very consistent corporate purpose. So our purpose as an organization is to make good things easily accessible for more people. And as an E commerce team, we’ve derived from that. And we’ve set our own purpose, which is to make it easy to find us online. So via Google via retargeting via paid social, and then for our customers to enjoy the right experience at the right time on the right device. Now that ecommerce purpose, it screams, easy customer experience, friction free shopping, personalization, all of the other buzzwords that you’ll hear an awful lot. But it is really important to us. And it’s something that as a department will take very, very seriously. One other area that we’ve invested in very heavily is testing. So it’s all good standing up on stage like this and telling you that we have a clear ecommerce purpose. But if we don’t do something to ensure we stay true to that purpose, then it’s not worth the paper that it’s written on. So over the last few years, we’ve invested an awful lot in both qualitative and quantitative research. So this photo is of the UX Lab, which we built in the center of our head office, you may have heard about it before. But we basically said, if our aim was to become truly customer centric, and put the customer at the heart of the organization, then the easiest way of doing that was to physically put the customer at the heart of the organization. So every week or two, we have real customers who come in, we can test new ideas with them, we can get their genuine feedback, we can also put stakeholders in our observation room, and let them see how real customers resonate to some of the things that we may want to do. The more qualitative testing also fuels our experimentation program. So to keep us honest, and to make sure we understand the impact that we’re having on our customers experience. We’ve invested quite heavily in an experimentation program over the past few years, which has resulted in us now being able to run around 100 experiments a month. That’s been critical for us. It’s changed the culture, it hasn’t only been a technology change, we now have our CEO talking about hypotheses. You know, we have people on a very senior level talking about this concept of being brave and failing fast, or understanding the value of failure. So around 70% of our tests fail today. But the learnings that we get from that are arguably just as important as the learnings that we get from the things that do work, and that that has been critical. It’s also been critical as we’ve moved more and more into the personalization world.

Now, personalization is very, very important to us. We spoke about this in the past. But the reason that we are really obsessed about the concept of personalization is because we think it’s the best way of making the customer’s experience easier. We started personalization a while ago, along with an awful lot of us. And then about 18 months ago, we got a bit more serious. We launched an acceleration team whose task was to focus specifically on personalizing key elements of the journey across all devices. And early on, I’d be the first to admit we did get seduced by some of the sexy things that didn’t really add any value. You know, we would put weather in the header and say, it’s raining today, because it probably wasn’t because the AP matching was probably terribly wrong. But we tried. It’s raining today, why didn’t he go and buys clothing that is that is appropriate for it raining isn’t really helpful. It’s raining now you need to understand what the weather is going to be like in the future. And we tried a few bits like that. And then over time, we focused on data driven and evidence led personalization, and have really obsessed on thinking about whereabouts in the journey, we can do things that makes her experience easier, and ultimately tries to create an online experience that feels as if she’s walked into a bricks and mortar store and had something merchandise just for her. A shopping assistant who knows the payment methods that she likes and the delivery options that she prefers. So over the last 12 to 18 months, we played around with lots of nice stuff in it specifically in the personalization space. And I’m going to share a few of those stories now. So first one was personalizing gallery sort order. Now we’ve got a team of data scientists up in Liverpool, you can actually get data scientists in Liverpool, it’s a bit more difficult than getting them in London. But we do have some of them up there. And what they’ve been working on is crunching lots of algorithms to understand how we can curate that gallery page. It was to a question that was mentioned before. But we’ve taken away to a certain extent, in this trial. Anyway, the merchandiser curating it. And we’ve done that allowed the brain to curate the gallery page. We’ve done this using propensity analysis to understand you know, what, what areas does she browse, where does she buy, where is she where she returns, and trying to get a picture of, of what works for her. Now we’ve done this category level as well. So for example, if I went into the men’s shirts category, and I have a real propensity to buy into branded goods, then it would know that and it would boost certain branded goods that I browsed or purchased in the past. But if I then went into the dresses category to buy a dress, my wife, I may be slightly tighter on the purse strings when I’m spending something for my wife and for me, so I may go for some of our own brand items rather than national brand items. And it would understand that a category level and would curate that gallery, according to my previous behavior. We’ve seen some interesting wins with this, but not across the board. I think one of the challenges we’ve seen in a number of these initiatives is it’s quite different to traditional conversion rate optimization, where you make a change, it looks good for everyone. You remove some friction and you roll on. We’ve been seeing specific segments either under or over indexing in terms of the performance. And so it’s that balance in terms of what do you do there and what is your next step? We genuinely genuinely believe that personalizing gallery sort orders has got lots of cash in it, and lots of benefit for the customer. And so it’s an area that we’re going to continue to invest in. Similarly, we personalized our top nav. Now, our top navigation, we are a department store, it’s quite cluttered, we know the customer reads from left to right. And so we know that customers would come in and shop with us because the real electrical loyalists or the real beauty loyalists, their items are on the right hand side of that top nav. So using the same propensity analysis, we started to work out what zones as we call them, customers are more likely to shop into and then started to curate that top nav experience for them across desktop and mobile. And we did see some wins and we specifically seen some wins in the customers who have more of an affinity for Home and Garden which is kind of lost in the noise in the middle of the nap. And again this is an area that we do really really think is interesting and wants to extend into our gallery page navigation. personalizing the order facets based upon the type of fastest customers use and really getting under the skin of that type of behavior. Probably the most visible change that we did was on our homepage.

Now, lots of retailers in the room will be accustomed to the frequent Royal Rumble when it comes to deciding what content needs to go on your homepage and shop direct we are no different. There’s definitely still a perception in some of the non technical teams that if my content isn’t on the homepage We’re not gonna sell. If it’s not on some of the category pages, nobody’s gonna see it, nobody’s gonna buy it. And so what we did to challenge that behavior was put it all in the hands of the brain, rather than in the hands of our training guys. We created a huge raft of available promos, we wrap some context around them all. And then we ran some algorithms, which basically defined what promos each customer could see. And we did it at a one to one level, which resulted in us serving about 1.2 million different variants of our desktop homepage. Now, again, common theme, we’ve seen some real wins in specific segments, and genuinely do believe once we can overlay the retail specialisms of we need, we need sale to be on here, because we’ve got 17 million pounds worth of stuff that could go terminal. You know, once we’ve got that type of involvement, we think this will be a real winner. And again, it’s an area that we are definitely going to look to continually invest in.

Now, there are a few of the highlights from our personalization journey, the most successful one. And I’m not just saying that because I’m stood on this stage poor, RichRelevance(now Algonomy) has without doubt, been our biggest personalization success story over the last 12 to 18 months. To context that prior to prior to working with the guys, we had a recommendations engine, but it was not personalized. So it was a recommendations engine that was based predominantly upon the product catalogue, and didn’t really understand anything about the customer, anything about the recession behavior, and use that data to really curate what we were serving via the recommendations. The journey with the RichRelevance(now Algonomy) guys, for us started back in the September 2014, when we made the decision that we were going to plough on not just prior to peak isn’t the normal time to make decisions like this. But we genuinely see the value of recommendations and believe that we could get something in to influence that Christmas period. So there was an awful lot of hard work from both sides. mid November 2014, we’ve got a client side implementation live on very cold at UK desktop website. And it was there it was stable and perform very well, commercially, it was great. And we’ve got over the peak period. Now we do a third of all of our business in a quarter of that year. So it was getting something live, which can be such a game changer was massive for us. We got through Christmas, everyone was really happy. We then started to roll out that client side implementation to our other devices and our other brands. And we continually seen an uplift in performance as we roll RichRelevance(now Algonomy) out against our previous incumbents. Then we moved to the full service side API based solution. I’d say end of summer last year. And then September, I think it was last year, we put our specific RichRelevance(now Algonomy) product feed in which enabled us to get a richer set of attributes across to the guys for them to utilize with their algorithm and obviously make wiser wiser decisions. Now, well, I personally think we’ve only scratched the surface with RichRelevance(now Algonomy), we put it in, we let the algorithms do things. But due to our testing mantra, we haven’t really rested on our roles. We we’ve tested lots on top of what we got, you know, we ran a test adding price, which feels like a no brainer. But when we first bought RichRelevance(now Algonomy) in we didn’t show the prices of reviews, sorry, recommendations, that was a big winner. We, we started to test recommendation slots on an empty basket, because previously, a customer would just get an empty basket staring at them. And this proved to be really, really successful. We ran a number of lab testing sessions on our home pages on both mobile and desktop to understand what type of content resonates with customer. The feedback was that you get banner blindness from banners, that won’t be a shock. But customers really do resonate with products as high on the journey as possible. And the heat maps were just off the scale. Even if there was product content further down the fold. Customers really resonate with a human face. And so our aim was let’s test different placements that get a contextually relevant set of products for that customer. As high off the journey as possible, which we tried in a number of ways on mobile. We tried it on our desktop homepage as well. And we also tried it on all of our top level zone or category pages.

Now we’ve also tried some stuff that was probably a bit too far. So we put RichRelevance(now Algonomy) slots all over our product page. It looks like a bit of fruit machine at times. But we just wanted to understand where that natural level was for the customer. This This test was quite interesting. Putting a full horizontal slot of six products did not work. But bringing it down to a scrollable block of three products really did work. And it’s just subtle differences like that, that we’re still trying to get under the skin off. But again, making sure as long as the right product is in there and then the UI and UX is right Is the hierarchy MSDN is right, we are seeing and responding to it really possibly again and again.

There’s been some other bits as well that have helped. So with our previous tool, there was just basic merchandising and retail and pieces that we couldn’t do. We used to struggle in the past to coordinate menswear, you know, so the men’s tailoring suits, jackets, waist culture to tie trousers, shoes, we sell them all separately, our previous engine wasn’t smart enough to enable us to pull all those together and present it to the customer in a compelling way. RichRelevance(now Algonomy) enables us to do that fight in the advanced merchandising, basic things like upselling, we’ve struggled with, that really, really stepped on an awful lot in the last 12 months. collections. So we’re a department store, we have an awful lot of furniture that we sell an awful lot of is old brands. So nobody else sells it. And we used to struggle to do very, very simple things like bring various collections together, because again, they will all set up as very individual products. This has enabled us to do that. And our cross selling real basic stuff, like, let’s get the right HDMI cable for a TV, that’s all stepped up massively. And the results of all of those things are that RichRelevance(now Algonomy) in the first full year performed 40% better than our previous incumbent, and now accounts for 5% of all of our sales. Now, that’s, for me, it’s very encouraging. Because we haven’t even scratched the surface yet. There’s lots that we want to do, we need to improve the UI and UX from our side, we basically just put in a default implementation of RichRelevance(now Algonomy) from an customer experience perspective. And, and so there’s still an awful lot of legwork that we can do to match the great work that the RichRelevance(now Algonomy) guys are putting in. And so our focus for the next year is, let’s improve the UI and UX, let’s make it make it look nicer. Let’s make it easier to use. Let’s look for new placements for existing strategies. So we are definitely seeing an awful lot of the existing strategies working very well, but feel as if they could work very well elsewhere. We think there’s opportunities for new placements and completely new strategies, we haven’t even scratched the surface with the variety of strategies available. We really like the idea of when you go into bricks and mortar store and you have your tail tat, which is there by the tail, and you pick it up, it’s a kinda, it’s a very easy, cheap, no brainer additional purchase. We think there’s an opportunity for starting to introduce some of that, add a specific line level within the basket, but then wrapping that in a really clean user experience of enabling the customer to add it to basket with one click. So we love the idea of that type of stuff. And then just making it easier for customers to interact with these recommendations. So if a customer sees a recommendation on the product page, it’s likely that the hero product on that main page is the one that we want them to obsess about. Let’s open the recommendations in a quick view window, so they don’t have to leave the page.

We’ve built a Multi Product Quick View window so that we can build outfits within RichRelevance(now Algonomy). And a customer can see all of the items within that outfit and add them all to basket at the same time. That also works for our furniture or complete the look. Or for a group of related electrical pieces. I can said we’ve interacted with, we’ve developed a single SKU add to basket for your real quick purchases like HDMI cables. And we really want to drive this entire experience on whilst the RichRelevance(now Algonomy) guys and our data science guys continue to drive the algorithms. So for me, we’ve been on a real personalization journey. The exciting part is we’re only just learning what actually doesn’t work. We’ve got an awful lot more to do. And I look forward to coming back to come back and tell you more about in the future. Thank you very much.