.svg)
We presented at the PI Apparel fashion spotlight global event on the what is future of fashion B & M, and what does fashion buying and merchandising for new normal look like. You can check out our exclusive panel at PI Apparel Spotlight here.
The full video recording and transcript below.
https://youtu.be/3Ea5QS5jjVUFull transcript of the talk on fashion buying and merchandising for the new normal below:
Ganesh Subramanian:
Hi everyone. Good evening and we are at a fantastic stage to look at how do we look at future of merchandise planning and it's nothing better than putting consumer at the center of it. Let me just start by talking about what are the key questions facing the fashion business planning right through the value chain. How can we be more efficient than even the previous panel talked about is, "How do we improve the quality of decision across design, buying and merchandising?" While I'm looking at the entire value chain, we'll focus a lot on buying and merchandising in the presentation today. How could we get better at fashion trend spotting and buying decision?
Because merchandise planning is very closely connected to design on the other side, buying on the other side for so look at holistically today. How could we better trend forecast and validate that better impact the KPIs of the fashion business? How do we take bets on new trends which we haven't done before? How can we decision making with better intelligence, not only fast, but also accurately, right? I'm sure you will have some more questions but if we are going to have some answers and unique approach to solving these challenges. Before I introduce. You know every transformation has always been through intelligence.
Whether it's technical revolution or sustainability of human and also businesses, achievement in businesses, and also in democracy cutting across it is intelligence that transforms any business. Now, this is something that you can look at from 1960s to till now. It's a period of industrial automation to business process automation, digital transformation. We are in an era of intelligent enterprise. How do we create a merchandise planning and buying merchandise planning in the intelligence space? Now, as you move through this progression you will see that the automation increases, repetitive tasks come down and high value tasks significantly increase.
Now we are actually giving you an approach today of how do we progress in this area significantly. Before I do that you have been hearing a lot about AI. I just thought I'll give a very quick one to slider on what AI is while a lot is being said about AI is it's a computer science technique, which helps you learn from experience and experience is all sitting in data. And also continuously adapting new inputs in a dynamic environment like we really need a technology which not only learns from experience but also adapts. Now what's very important is AI is all about data. We heard about it that data is all about people because all data is generated directly or indirectly by human and hence AI is all about people. AI is not an inanimate algorithm. For there is no AI success without good human input into the data which gets into these algorithms.
How does it help? It helps in various ways whether detecting diseases, driving a car, helping a restaurant or a retailer predict demand, optimized logistics, in agriculture, various use cases just throwing to you so that you can imagine what's possible in your area and we will quickly get into what's possible in the merchandise planning area? Why should you care about AI? Now potential is unimaginable. It can impact research, production or service whatever you do. Potential to rewrite the existing business models through an advanced way of planning. And more you learn and adopt and if you're the first one to adopt, you will have a leading edge in the marketplace.
AI will always need human irrespective of the case, be it medicine be a self-driving cars, human plus AI accuracies have always been far better than mission alone. Well, what I'm trying to say is it's very, very important for us to have a collaborative engagement with the AI models. But Let's come to the context of fashion. We're sitting with a humongous supply demand gap creating $750 billion of wastage. Over 50 billion garments not selling every year. Over 50 billion selling at discount. Now that's a huge gap irrespective of all of the solutions that we have at this moment. And what is causing this?
Either it's an intuition supply driven forecast, which is the demand planning side of the business, both from a quality and quantity side, there is no reliable way exist, right now to predict timely, both quality and quantity of merchandise. With and post COVID the challenges further. And how do we get out of this for, we really need a paradigm shift. What has not happened for decades? If you have to touch, we have to add something fundamentally. And today we are going to break that myth which is most of fashion planning and forecasting happens through supply data. Trend spotting and which is where the origin of all products start. It's happening through supply, right?
We observe. What does observation mean? We go and observe across the world. For we see what's in supply and we know two thirds of supply to not meet consumer demand then we need to move from supply signs to demand signs. So very quick introduction on what Stylumia is all about, we are based in Bangalore in India. And it's one of the fastest growing fashion technology companies in the world. Now we are serving fortune 100 clients right up to medium sized brands and retailers all over the world. We are presence in th Unites States, we are presence in United Kingdom, Australia and New Zealand. Our customers cut across sizes, shapes, categories across the world. And we have global with partnerships right from leading consulting companies to color experts like Pantone in technology partners. Moving into what's the need of a transformation right now in planning.
While the challenge in retail has always remained the same right of everything. But with technology, like the panel also discussed is we have a huge opportunity to improve precision, personalization which is localization, prediction and automation. And this is important considering the demand uncertainties and the challenges and shrinking margin and loss sales. We are sitting in an inventory epidemic today with over buy and under buy and the challenges more and more as we go forward. While the customer expectations have changed significantly, retail operations have changed product and inventory decision-making is something that needs to change significantly, we see a huge opportunity to change. For if you just look at the value chain, right from design, the process of design has remained the same for a long time.
But what we believe is that we can augment this area with information so that we can take informed consumer driven decision making. Moving into buying and merchandising. Data exponentially increasing. But the question is, what data are we using? Second? Are we removing the noise from the data? Now that's a lot of hard work. Now, unless you clean the data, remove all the noise, we are not going to get better outcomes from any of the machine learning or AI models. Looking at the store channel and distribution and allocation, local market has changed significantly. Consumer preferences are changing dynamically. But the way cluster stores, that's not changed dynamically. We're still working on rules. For rules are very, very static.
For rule-based merchandising systems are not going to cut the way forward. While, there is a need to change. We discuss significant demand and uncertainties, particularly post COVID. Hence, we need a new thinking. Now, the new thinking is a holistic view of the fashion value chain. Consumer trend spotting on one side to distribution on the other side and buying and merchandising in between. Now plan buy and merchandising is something sitting at the center of everything that we do. For what we solve for you is fundamentally three things answering three questions.
In other words, you need to ask these three questions. How are you solving the 'What?' problem, the 'How much?' And the 'Where?' and 'When?' problem. For there is no bad design if you bought the right quantity. Therefore, planning and merchandising makes a huge difference to the overall profitability and revenue growth of any brand. We have created huge impact across three key P&L variables and saving over 60 million garments and wastage thus far. As I told you before traditional retail planning platforms and rigid and limited by rules. What's the part to next level of retail performance, which is basically taking data driven decision-making in those five key areas, consumer insights, competitive market intelligence, forecasting, planning and allocation.
The next generation performance will deliver various metrics at different parts of the value chain one is how do we buy more of winning styles or how do we place bets on winning trends, moving into selling more and decreasing the markdowns and reducing the cost overall. Looking at all the parts of the value chain. Let me quickly take you into how do we get consumer intelligence at internet scale? We have an engine which is similar to what Google ranks pages, we products dynamically which means we sense demand not supply for we provide winning trends. You get winning trends across all geographies time-span categories, products very relevant to the DNA of your brand and your inspirations.
Moving into buy and plan optimization. Now in buying and planning context is very, very important. If I ask you, what's the mood of this person? Your answer will completely change when I show the context. How do we understand the context? As I was talking in that panel a while back where we know what customers actually buy that's the sales data, but they buy in the context of other products. For how do you predict demand of a product dynamically based on the assortment. The unconstrained demand is what typically we estimate. But the question is in the real world there is a lot of constrained demand.
For a lot of these practical questions, right? We sold out of a product in a particular store, we've never sold a product, but what do I send to that store? We are looking to add a new product and what could be the demand of that store? How do you understand all these? Therefore, you really need a new generation prediction model? What you're seeing on the screen is a very unique prediction model for fashion which reflects the human brain by using both images and textual attributes. Fashion cannot be narrated and hence attribute based prediction models will not have as much accuracy as an image plus text prediction model.
Now this is something that we use to predict unconstrained demand of new product that which introduced for the future. We have-demonstrated accuracy lift anywhere between 10 to 40% using these models with our customers that has a potential of almost doubling your profit. Now Merchandise planning through in season at various levels of the hierarchy this is an in-COVID prediction accuracy over 95%. After sensing the demand for the first eight to 10 weeks our models are able to predict in an Omnichannel environment, demand accuracy at various clusters at that level of accuracy.
For what you're seeing on the screen is our merchandise assortment optimization planning tool which helps you plan your business at various levels of the hierarchy and, it gives you instant heat map on how should you invest or reinvest money depending on where you're using the tool? Either pre-season or in season? For dynamic allocation and reallocation of money within different hierarchy of your business. Very simple to use. You can decide and what metrics that you would like to plan and, like to the size of the squares is the size of the business and the color tells you good, bad, ugly with respect to the plan numbers. So you can instantly focus on that area.
You not only have a heat map view, you have detailed view and you have trend view where you can see the moment of the business over time. And it also helps you predict the over stock and under stock situation as you run through in season of that business. For you're completely on the dot in terms of what decisions you need to take? And it will help you take decisions at the level of the style SKU store level whether you want to do marketing or deals or discounts or return or replacement. That's the buying assistant for you. And the last one is all this is good if you don't allocate the merchandise in the right place and allocates are actually are the custodian of all our consumers.
Now, one of the biggest challenges and allocation is how do you understand the dynamic demand of stores, not the traditional way of clustering. An allocate them goes through various challenges through the day and they need to answer all these questions how do I handle your products? Short life saving products or basic and seasonal products. Our allocation model cuts down the rigor from 12 steps to just five steps where the human element is involved. It automatically optimizes the allocation by store by style. It also has an opportunity for you to override in case you want to have a judgment on the optimal allocation. This system uses the taste model which I talked about which understands the predicted demand of a style in a dynamic changing assortment within a store for it doesn't work on any fixed clustering of stores.
You have various options to look at store level, style level allocation or into a style level, store level views, multiple views. It also applies the relevant size serve across various products to the stores. And it does optimization on your behalf. Therefore, it does the grueling task and reduces time taken for an allocater on a daily basis, improving the inventory done and key KPIs for your business. And the last one, I just want to leave you. This is a cutting-edge innovation. Nothing to do with merchandise planning, but I taught it's very, very important because all merchandise starts with creation. What you're seeing on the screen is a machine generated fashion design which is a design assistant and I'm no designer what is created is something real time using generative adversarial networks.
The idea is to enable the designers to come out with winning design ideas and we can curate the ImaGenie with that data sources which are very relevant for creation based on the DNA of your brand. For that's the opportunity from creation to distribution. It's a short time. I just wanted to leave you with a lot of thoughts of what are the possibilities of using cutting-edge technology to increase the accuracy and automate and minimize the grunt work and guesswork from the retail merchandise planning. For some of the recognitions and what our customers say that's most important for us.
Now, why Stylumia or any other tool today is with the dynamic environment it's very important that we need to put consumer at the center and we need to have a demand sensing view which is a consumer driven approach to whether create products, spot trends or plan merchandise or allocate merchandise to our customers.
With that. I would like to ask David if there are any questions from the audience.
David Wilcox:
Thanks, Ganesh. Yeah, look, we do have a few questions. So let's get straight into them. Have you seen an increase in demand for your solution because of the pandemic?
Ganesh Subramanian:
Yeah. So can you repeat, David?
David Wilcox:
Sorry Ganesh. So have you seen an increase in demand for your solution because of the pandemic?
Ganesh Subramanian:
Absolutely. In fact, we've more than doubled our business during the pandemic period. This is the time what we see is that the technology adoption, particularly understanding consumer demand side of the business has exponentially increased. This is across geographies, While retail is the most suffering during that period for this is one area that people are actually investing, not cutting back. Because they're preparing themselves to take informed decisions for the future.
David Wilcox:
Okay. Do you feel this new found interest in AI and being more predictive is here to stay? I imagine you do.
Ganesh Subramanian:
Sorry?
David Wilcox:
Do you feel like this new found interest in AI is here to stay? I think we are not behind any technology we are behind. How do we take informed decision making for the consumer, wherever AI definitely helps because it can handle a lot of unstructured and semi-structured data and give real time predictions. Because in an industry like fashion while it's a lot more challenging from a data perspective but in my view, AI is all about pattern recognition and pattern recognition is here to stay. Because we have to continuously recognize patterns from all the data and take key business decisions.
David Wilcox:
Okay. Next question is, does your solution help to remove any noise from the data before it can be used as actionable insights? Do you kind of understand the core of that question?
Ganesh Subramanian:
Yeah. Our philosophy here is data is 80%, 20% is the AI model. For we spend lot of time in moderately fast understanding the data of our clients and cleansing the data and validating the data back with our clients where the brands or retailers. Very, very important to get that right before you even feed that into the model. For even in our implementation time, 80% of the time goes in data and then 20% in the model and prediction.
David Wilcox:
I see. Okay. What about your clients? So have you noticed that they're investing in new skill sets when it comes to data? There's been a lot of talk about training and digital skills deployment in the industry. Have you noticed that clients are investing in that training?
Ganesh Subramanian:
Yeah, now, In fact, some of our clients have even asked us saying that, "Can we have an orientation or a AIs for starters?" It's very important that before we make people adopt and of course we have clients at various levels of adoption but I'm just giving you shot in the middle. It's very important to create awareness, And a lot of our clients are investing in equipping the existing team to say, "Why should they adopt?" Because, everybody understands without adoption there's no value of any technology. You need to believe that it will lie to value. And at sometimes people... There are some areas where you feel that will it removes some jobs? Or will it reduce the work? In fact, what we say is it's augmentation. In fact, how can we do more and better, faster, right? From that perspective to bring all of that on the ground, there is a fair amount of skill augmentation, training and many of our clients are actually doing that. And we do contribute in that with some of our clients.
David Wilcox:
Yeah. Has that been something that you've now taken on that duty of training. If everyone's asking you to do it, it must be an opportunity there?
Ganesh Subramanian:
Oh, absolutely. Absolutely. We believe that it's important for us to do that so that we carry the why for them to understand why should I... Because sometimes AI becomes a black box, right? So why should I believe the black box? Is this true or isn't this true, right? For, we really need to prove and tell them the fundamentals and then say, "Yes, we will prove with the data." And then showcase that it's actually improving. Right? Because unless you showcase the performance that you will not be able to create the trust. For it's very, very for us to create trust in the technology, trust in the models that we are building whether it's relevant for them? And is it working in their real environment? And it's very important for us to do that before getting into large scale adoption.
David Wilcox:
I imagine so. It must be interesting. I've got a personal question here when you talk about new technology introduction into a company and certainly AI and predictive planning is reasonably new within the industry, you could argue that, but who do you approach within the company? Is it a top-down or do you have to find certain champions within different departments to kind of champion this adoption within the companies?
Ganesh Subramanian:
Yeah. David, I think it's a very interesting question right now. How do you get that option right? And we worked with various types of clients. But I can tell you one model that works really very well is that the on the ground people who are working actually doing their role, they're very busy and we know retail is... People get so busy, very, very less of breathing time for you to experiment with anything new. And that's on the one side. And second is people have something called an innovation department if there's a large company, right? Typically we go through innovation and a large company set up. And what that innovation does is innovation actually incubates a new idea in the company, hand holds with the mainline function and ensures that there is a smooth transition. And that has to be an orchestrator. Very important to have an orchestrated between I would call the artist and the people who are actually on the line execution, right?
For if you delicately put an innovation on the line from day zero, it's difficult for anybody to have that time and breather to implement something new, particularly bring some new challenges from a data perspective in particular, not all our solutions have that. They're not some off the shelf solutions like a consumer intelligence tool. It's a SAAS model, one can get started in just a week's time. But when you're really getting into prediction, quantitative prediction of demand, there is a lot of data play there.
David Wilcox:
Hmm.
Ganesh Subramanian:
But I would say that the structure of an orchestrator plus an innovation and a champion from the function be it planning or merchandising would really help. And that has helped us in many of our client relations.
David Wilcox:
Excellent. Thanks. Thanks for asking that. What do you think is the turning point with a brand that drives their investment in a tool like yours?
Ganesh Subramanian:
Yeah, now, it's a great question. Basically we were asking about how do you sense the ROI from an investment, right? I think every business wants to see what's the return on investment I put on a technology like this? Our experience, of course you need to have a minimum gestation period of six months because in fashion you take the decision you don't see results immediately. Depending on where? Allocation decisions are weeks, but if you're taking planning decisions and if your supply chain is six months, then you need to wait for that cycle to get completed. But we have been able to deliver a 10x ROI in 12 months with most of our clients-
David Wilcox:
What was that? Sorry, Ganesh, 10-X-R-Y?
Ganesh Subramanian:
10x, 10 times return on investment in 12 months, right?
David Wilcox:
Right.
Ganesh Subramanian:
... From the time of investment and we do measure by... First, we asked our clients to do implementation in a thin slice and say, "Don't take full scale." Right? Implement in one part of your division. If it's a medium-sized brand, maybe you've got multiple brands, maybe apply one brand. If it's a very large brand then say, "Take one slice of a category." Create a success in a thin slice, therefore land there, make it working, right? And actually those users help us scale within the organization. It's easy for the organization not to actually carry the full load of full scale implementation initially. That approach has also helped us scale within multi brand retailers and also aggregated brand companies or large brand setups.
David Wilcox:
Okay, fantastic. And how would you respond to... I guess concerns might be the right word, but the AI limits creativity within the fashion sector. What, what do you say to those concerns that are out there? There's always people that are concerned with, with AI-
Ganesh Subramanian:
Absolutely. Yeah. I must tell you that our core users of one of our tools, consumer intelligence tool is mostly creative people. They're all designers. And designers love us. The reason is we don't sell AI, we sell consumer intelligence. Right?
David Wilcox:
Mm-hmm (affirmative).
Ganesh Subramanian:
Nobody's against consumer. Tell me which designer in the world saying that, "I'm going to go against consumer." Now I think it's very important. It helps them take some important decisions. Now the important decisions are... Many times I will give you one example maybe quickly is to say, designers come with some great ideas sometimes and when they present it, somebody will say, "I don't think it will work." Some great ideas are shot down. Now Stylumia is a platform, helps them take a risk and also convince others within the system and say, "While I'm saying that this is a great idea, there's also winning and I'm spotting the trend."
And this trend is actually winning in the areas that we are observing. It helps them to validate a fair amount of their creative decisions. Sometimes because it's subjective, ideas might get shot down. It could be the other way around, right? Sometimes they are favoring an idea. It may not get through. But it helps to validate and it enables them to take a risk. It's not going safer for it's a friend and that's why we calling it a designer's assistant.
David Wilcox:
Mm-hmm (affirmative)
Ganesh Subramanian:
Right? We are assisting designers, we are assisting buyers, we are assisting merchandise plan. Therefore, they are on the top. It's just an enabler. For just a rocket engine to fly them to wherever they want to go.
David Wilcox:
Yeah, absolutely. It makes me feel like the term or the word artificial is really misplaced here-
Ganesh Subramanian:
Absolutely.
David Wilcox:
... Because there's nothing artificial about it.
Ganesh Subramanian:
It's amplifying intelligence or you call it augmented intelligence. It's augmenting to and I think any success of any of these tools is I think 80 to 90% goes to that person, which is the designers, which is the buyer, merchandiser, because they got the belief, they tried it and they feed the data. They give all the inputs, full credit to the adopters of there and I really want to thank all the clients and anybody who's adopting the technology leave us anywhere else in the world. And it needs an experimental mindset and managing the daily chores of their jobs and experimenting. It's not easy.
