The next wave of fashion planning will be through ‘Demand Intelligence’ combined with ‘Prediction’, powered by consumer demand signals at an Internet scale. All this is available for the decision maker on the cloud at a fraction of cost of the enterprise software of today…
Adriano Goldschmied, the Founder of Diesel and the Godfather of Denim, along with a dozen other experts met at a Bluezone Show, Munich in January 2018 to discuss how the future of denim will be shaped by Robotics, Artificial Intelligence and Big Data.
For example, in Alibaba’s Fashion AI Store, consumers get a completely personalised service from check in to the store to check-out. The environment between the consumer and the products is extremely interactive and the role of the sales staff comes largely from requests of consumers using the technology interfaces.
What came out from that discussion is extremely important for everyone in the fashion business who is looking to invest in Artificial Intelligence (AI). Key takeaways from the discussion were:
1.AI is giving a lot of opportunities to humans to make their jobs better
2.It helps predict human errors
3.The more data we have, the better it is
4.Machine Intelligence + Human Soul = Awesome Designs, it is a world of Man + Machine
5.AI will help in reducing the overall wastage (by predicting what to make, how much to make and when to make)
With these developments brewing in AI and Big Data, it brings the need of developing skills for fashion professionals from design to planning to operations in adopting and understanding how technology can enable their functions, thereby making the best use of man + machine.
How AI Will Affect Business Strategy
We have been focusing on retail developments in the West for a while, but here is an insightful and revealing look at what is happening in the East. China is leading in retail innovation, ahead of the rest of the world. In 2016, e-commerce giant Alibaba’s founder Jack Ma predicted a seamless merger of offline, online and logistics for a dynamic new world of retailing. China is no more the back end of the world but also at the leading edge of technology.
Alibaba CEO Daniel Zhang said recently, “Companies must use Big Data analytics to redefine the traditional core elements of retailing – Consumers, Merchandise and Stores – and the relationships amongst those elements to upgrade current formats and create new retail occasions.”
For example, in Alibaba’s Fashion AI Store, consumers get a completely personalised service from check in to the store to check-out. The environment between the consumer and the products is extremely interactive and the role of the sales staff comes largely from requests of consumers using the technology interfaces.
Inside Alibaba’s Fashion AI Store
At first glance, Alibaba’s Fashion AI Store does not seem so different from a regular shop. There are clothing items galore, neatly displayed on hangers. However, here clothing racks are RFID enabled and use gyro-sensors and Bluetooth low-energy chips.
This means that articles selected by shoppers automatically show up in the Smart Mirror.
Via the mirrors, shoppers can look at different size and colour options but also receive personalised mix-and-match options to complete the look. They do not have to search for the suggested items but simply follow the Smart Mirror’s directions to the right location in the store.
Shoppers who like to try before they buy can do so of course and do not even have to lug the selected items around with them while browsing. They just add them to their virtual shopping cart from where they will be delivered to the fitting room directly.
Then there is Intersport, Switzerland’s leading sports retailer, which is transforming consumer experience through its flagship store in Beijing. The store showcases everything from interactive windows to real-time product information, customer reviews and a cloud shelf too. This is combined with door delivery in under 2 hours.
Then there is Intersport, Switzerland’s leading sports retailer, which is transforming consumer experience through its flagship store in Beijing. The store showcases everything from interactive windows to real-time product information, customer reviews and a cloud shelf too. This is combined with door delivery in under 2 hours.
The store merges online and offline experiences as well as featuring custom-made fixtures such as three handcrafted wooden coffee bars, a 3-D printed tea bar, and a two-story, copper cask, engraved with traditional Chinese stamps that tell the story of Starbucks and coffee.
Developments in AI Drive Predictive Ability of Machines
Whether it is Harry Potter or Matrix, the human characters’ belief in predictions drive the plot. Predictions affect behaviour and they influence decisions. It is important to understand what prediction means. Prediction need not be just about the future, it can also be about the present moment. We predict whether a current credit card transaction is legitimate or fraudulent, whether a tumour in a medical image is malignant or benign, whether the person looking into the iPhone camera is the owner or not.
Fraud detection accuracies used to be about 80 percent in 1990, moved upto 90-95 percent in 2000 and it is 98.5-99.9 percent today utilising AI capabilities.
Most of us are familiar with shopping at any online retailer, e.g., say Amazon. As with most online retailers, you visit its website, shop for items, place them in your cart, pay for them, and then Amazon ships them to you.
Right now Amazon’s business model is ‘Shopping-then- Shipping’. During the shopping process, Amazon’s AI offers suggestions of items that it predicts consumers will want to buy. The AI does a reasonable job. As per neutral studies, ‘With millions of product offerings, Amazon gets individual predictions right 5 percent of the time’. That is very impressive considering the width of offerings.
When this prediction accuracy reaches a threshold, it would make lot more economic sense for Amazon to first ship and then sell. In other words, consumers will get goods and they will keep it if they like it else return the goods. No wonder Amazon filed a patent in 2013 for ‘Predictive Shipping’.
With higher accuracies of prediction, the fundamental business strategy of Amazon shifts from Shopping-then- Shipping to Shipping-then-Shopping. If it adopts this new business model, Amazon then needs to build a network for getting goods back.
Prediction abilities hence will impact business strategy in the time to come.
Using AI capabilities accuracy of prediction has been improving significantly over the years. Fraud detection accuracies used to be about 80 percent in 1990, moved upto 90-95 percent in 2000 and 98.5-99.9 percent today.
Another area prediction has been successfully used in is language translation. We have come to a stage wherein the machine translation is a lot more accurate than even the best linguistic expert. We all see the benefit of this development in accessing information across the world.
More recently, the nature of prediction problems moved into cognitive areas like object recognition. The accuracy of predicting what is there in an image has also significantly improved over time and now the latest results outperform human benchmarks. The progression over time on image classification is evident below.Our businesses and personal lives are riddled with predictions. Often our predictions are hidden as inputs into decision making. Better prediction means better information which means better decision making. Prediction is, therefore ‘intelligence’ with us obtaining useful information. Better predictions lead to better outcomes.
How Are These Developments Relevant For Fashion Business?
A quick check on the way fashion predictions reveal that there is a fair amount of future validating questions that are asked to experts (wholesale partners or internal stakeholders).
What needs to be done to bring in a true ‘outside view’ into fashion decision making. With consumer information available a lot more than ever before, smart usage of data – both from outside and inside – to derive insights is one big step towards formally bringing in an ‘Outside View’ into the decision making process.
With the ability of machine vision, AI models are able to predict 40-50 percent better than existing fashion prediction processes. Those brands and retailers who equip themselves with a better view of the future, better than their counterparts in the market, are the ones who can win the consumers’ wallet in a sustainable way. With constant improvements in predictions, we could soon see shift in the fashion business paradigms like the one we saw above in the case of Amazon.
Better prediction means better information which means better decision making. Prediction is, therefore ‘intelligence’ with us obtaining useful information. Better predictions lead to better outcomes.
Here’s a glimpse of what the future Paradigm will look like:
As fashion is full of forecasts, it is important to measure the accuracy of the forecasts and ensure that this metric improves over time. The inspiration for this comes from the world of machine learning. When machine learning models predict, there must be a method to choose the right model. The metric used to measure the accuracy of the prediction is F1 Score.
Fashion forecasts are done by style using a numeric (1-5) or an alphabetical scoring (A to C) to indicate how good the potential of the style is. To simplify this, let the styles be categorised as Winners, Mediocres and Losers (pre-season) by each forecaster.
Once the actual results come in, plot the actual vs. prediction in the template below:
After plotting the prediction vs actual occurrences of a forecaster, F1 score can be computed as follows:
F1 Score = 2 * (Precision * Recall)
(Precision + Recall)
We need to compute Precision and Recall
Precision = True Positive
(True Positive + False Positive)
Recall = True Positive
(True Positive + False Negative)
To get clarity on TP, FP,TN & FN, we will take a Cricket analogy.
True Positive: Umpire gives a batsman NOT OUT when he is NOT OUT
True Negative: Umpire gives a batsman OUT when he is OUT
False Positive: Umpire gives a batsman NOT OUT when he is OUT
False Negative: Umpire gives a batsman OUT when he is NOT OUT
If F1 score is calculated every season and forecaster, the brand will have a clear idea of the strengths of forecasting in the organisation.
This is where the power of technology comes in to enable every decision maker with predictive powers of machines.
With the morphing of retail landscape, what is important to you changes and hence what to measure.
With the connected consumers demanding for “Control”, there is a need for Brands and Retailers to fundamentally shift their approaches to win consumers.
Here are some of the imminent key shifts:
The fundamental shift is from a pipeline of value delivery to a participative value delivery to consumers and their community. It is selling memories and not just products. Winners in the future will shift their approach signicantly.
The Future of retail is not ‘in a box’ any more, and brands will create ‘Ritual-tailments’ There is a need to ‘Learn and Go’. If you wait for 100 percent certainty, you will be 100 percent late. Speed is key.
How Stylumia Helps
Stylumia’s tools come with the potential to enable informed decision making by all stakeholders. You need tools that provide relevant metrics in a form to take quick actions pre-season, in-season and postseason.
We converted the changing paradigms into measurable metrics:
Fashion Forecasting of the 21st Century Is Similar To 19thCentury Medicine
We can learn a lot across domains. A close look at how medicine has evolved tells us a story. When we are ill these days, it is an automatic response to call a doctor or go to a pharmacy to fetch medicines. Just about 200-300 years back, this did not exist. We have reached far ahead in medicine now, thanks to getting to the bottom of the causation, with the discovery of ‘Cell Theory’ by Robert Hooke. It is the advances in microscope technology in the 19th century helped scientists observe live cells and we reached a tipping point of modern medicine.
We are at a similar tipping point of discovering and ú nding ways to cure and prevent challenges which have been there in the fashion industry for decades.
Dwelling further in the area of forecasting, itself, reveals another story.It is Philip Tetlock who did decades of research on forecasting ability of experts and came with the theory that many experts forecasts were as close to chance as throwing a dart by the common. That is not to say there are no expert forecasters. They are a very few and he studied in detail what made those Super Forecasters perform superior to many others.
Since Tetlock’s enlightening interview in 2015, technology in deep learning has advanced signiú cantly and also the ability of machines to predict. We still believe best outcomes in forecasting will happen through:
1. Improving the ability of the human to forecast better; and
2. Humans using inputs through technology, collaborating to update the forecasts constantly and validating the forecasts with machine generated ones.
What Makes Someone a ‘Super Forecaster’?
What kind of personalities would a media company get to talk about a predictive subject for their show?
a) Charismatic personalities with bold, certain views (a Hedgehog); or
b) Someone who comes with moderate perspectives, analyses both sides and comes with a central view? (a Fox)
The obvious answer to the question should be hedgehog, but study on forecasting accuracies over the years has revealed that it is those who t the Fox prole are Super Forecasters. There is an inverse correlation between forecasting accuracy and media presence.
- Some key traits of a Super Forecaster are:
- Be cautious, for nothing is certain
- Be humble, for reality is infinitely complex
- Be intellectually curious and open-minded; think probabilistically
- constantly updates relevant information
- Value diverse views and synthesize them into your own
- Believe it’s possible to get better
Brands & Retailers who realised this tipping point and the power of Man + Machine have started working with us and are outgrowing the market and their competition.
The Next Wave Of Fashion Planning
The next wave of fashion planning will be through ‘Demand Intelligence’ combined with ‘Prediction’, powered by consumer demand signals at an Internet scale. All this is available for the decision maker on the cloud at a fraction of cost of the enterprise software of today.
This is the transformation that we at Stylumia are powering. We enable the three key questions of AI, Big Data and fashion business that any fashion Brand using Machine Intelligence by combining the power
of image and text analysis of both outside-in intelligence and inside intelligence (of the retailer’s own data)
Here’s a fashion risk decision tree with solutions for retailers to get an idea of the possibilities of future-proofing their businesses with a consumer-driven actionable intelligent system.