Celect

JOHN ANDREWS: My name is John Andrews. I'm the CEO of Celect. Celect is a predictive analytics technology company, focused on retail. The company was founded out of MIT by two MIT professors, Devavrat Shah and Vivek Farias, who'd been collaborating on a line of research over the last decade or so, roughly around the habit of customer choice modeling. So understanding how customers choose between an assortment of products, and understanding that in a much more granular and precise level. And then using that information to help retailers identify which products to put in front of which customers at any point in time by understanding how that customer is going to choose between a set of products.

So I met Vivek and Devavrat a little over three years ago. I was introduced to Vivek and Devavrat via one of the initial seed investors and advisors to select who was the former CEO and founder of a company that I spent about 10 years at called Endeca based here in Cambridge. And I was at that point, running the commerce product at Oracle. I had come into Oracle via an acquisition of Endeca in 2011.

And went Endeca was acquired, we merged with ATG, which is a commerce platform, another company here in Cambridge. And we brought the Endeca and ATG products together to form the Oracle Commerce product. And I took over leadership from a product perspective there.

So when I met Vivek and Devavrat and heard about what they'd been working on, in terms of the research and what they'd built with some of the early data customers, I got very excited about the science behind what they were doing, after building and selling product into retail for the past 15 years or so, and asking every retailer what's on your roadmap? What are you investing in and understanding what the needs were? And the state of the market from a technology perspective, in meeting those needs.

I realized that they had something very innovative that met a very real pain within retail. The core customer, who they would be selling into, a merchant and planner within the retail space, is actually the role that my wife plays, as part of her career. So I see what she works on on a day to day basis, and the challenges that she and her colleagues have, and realize that Vivek and Devavrat were onto something pretty amazing. And I find myself basically going to bed at night and waking up in the morning, thinking about Celect, what these guys had built, and just wanted to be a part of it.

JOHN ANDREWS: The Celect Choice Engine helps retailers identify how customers choose between an assortment of products. The context of every customer's decision in being able to look at not just what they bought but what was available to them when they made that selection, that context actually delivers very rich signal in terms of how an individual customers choose.

Now, one of the key challenges that a retailer has is around this problem of sparse data. Everybody talks about big data. Retailers have more data than they know what to do with. The problem is is that when you're trying to identify the buying patterns of an individual customer, you may only have one or two data points from that customer over the course of a year, right.

Think of how many times you've made a purchase at one of your favorite retail stores, right? It might be one or two, maybe three times over the course of a year. That's actually very sparse data in terms of being able to truly understand how that customer chooses.

But when you can add in the context of a customer's decision where we know what was in inventory and what was available to that customer, and identify with that inventory and that transaction information, that context, and also very easily layering in additional data, such as browse history, an individual customer may look at four or five products on the website but then put one or two of them into their shopping cart. That's context.

They may have three or four products in their shopping cart. They may abandon two of them but then purchase one of them. Again, that's context that helps you understand how that customer chooses.

By bringing all that together with every single customer and every single transaction, we now have a very robust model that allows us to predict based on customers' buying patterns what the likely behavior will be in the future and then be able to optimize on top of that to say, OK, based on this, what are the right products to put in front of an individual customer? Or more importantly, and what becomes much more complex for a retailer, is when I need to add constraints in, in terms of I only have so much space and I only have so much inventory available to me, what are the best products to put into an individual location?

We're helping them make those decisions that can lead to anywhere from 5% to potentially 15% increase in in-store revenue. And when every retailer these days is looking for ways to increase their margin and increase revenue, particularly the revenue within their stores, this gives them something that frankly gives them an ROI that they can't find many other places.

At the highest level, the problem that we're solving is what we refer to as the inventory portfolio optimization challenge, right? Retailers, the inventory number is the largest number. It's the largest number on their balance sheet. If you ask any retail executive what's on the top of their priority list, it's being more efficient with their inventory and inventory management, and increasing inventory turns, reducing markdowns, and reducing stock-outs of products. That all comes back to bringing the right inventory in.

Now, when merchants and buyers within a retail organization are going out into market to buy products, an average size retailer, a particular buyer may have-- they may have $500 million to spend on inventory, on product. If you think of something as simple as a shoe buyer at a department store with $500 million to spend, they might need to bring in 100,000 SKUs to put that money into play.

And when you think about what they need to buy, how much of an individual product, how that product is going to interact with other products, right, and the trade-offs, is it complementary, is it cannibalistic to another product, and how it fits in with the overall assortment, the complexity of that model can be baffling, right? And the results of that and the implications of that is the way that these decisions are made today is basically with gut instinct and Excel, right, because there's no real system that can handle the complexity to truly be able to manage within the constraints of I only have so much money to spend. I need to think about it in terms of the overall assortment that can handle that complexity.

And so Celect came along, right? So now we have a solution that can handle that complexity, that can help merchants and planners identify what products they should be bringing into their assortment, the buy quantification piece of it, which is how much of each product to buy, and then the allocation piece of it, which is where to put those particular products.

Where we've had the most success over the past couple of years has been within soft lines, and in particular, apparel retail. When you think about the challenge within fashion and apparel, a particular product that a merchandise manager may buy for their assortment, once you sell out of that product, it's gone, right? And then you're on to the next style, the next fashion, the new wave, and the new design.

And so identifying something that you've never sold before and predicting how well that you think that's going to sell, that's a hard problem to solve. And that's one of the core capabilities of the IP, being able to identify and predict something completely new based on all your past history and purchases to identify how well that product will sell in the future. And so fashion has been a really good market for us.

Now we've moved also into hard lines as well and more replenishment type items. But some of the customers that we're working with today, Urban Outfitters, Anthropologie, Free People, Aldo, which is the shoe manufacturer and retailer in Montreal.

So when we first-- when we first started out, the focus really was on apparel and the reason for that is the dynamics of the apparel industry and how decisions are made in terms of the buy. Once you buy a particular product, if it doesn't sell well in the certain season, the customer-- the retailer needs to mark it down, right? Or write of that particular product.

So being able to identify the right products to buy and then putting them in the right location based on the buying patterns of customers so that you're not marking the product down, as well as making sure you have enough of it so you don't have stock outs. Right? So the customer is looking for the product and it's not there for them to buy. There is a very acute need within apparel and fashion.

So that was where we focused-- that was where we focused first. We are now moving into more hard line. Electronics, do it yourself, and home improvement, big box retailers. We've had a lot of success within department stores. Which, frankly, are really trying to find their way in terms of being able to present a very compelling assortment of brands to their customers and being able to identify and truly connect with what the preferences are for their customers, is something that's extremely important to them.

The longer term vision in terms of where we're going is really around the idea of inventory portfolio optimization. Where we've started today is really focused on the merchandising planning and allocation process within retail, but we see a much bigger opportunity across the entire supply chain, right? And through that value chain from brand and manufacturers to distribution all the way to retail across all different touch points and channels, whether it be online, direct, via wholesale, or retail in-store experiences.

And you think about what inventory is needed, how much of an inventory, in what assortment? And then at each step across that supply chain, identifying, OK, how much should I be buying? Which distribution centers or fulfillment centers should I be bringing that product into? How much of that product goes to each individual store, in what assortment?

When a customer buys a product online, every customer that-- every retailer who we work with today is starting to do what's called ship from store. They want to use their stores effectively as fulfillment centers. This allows them to push inventory into the stores, and then when an order comes in, be able to ship it from a store that might be near a particular-- near where that customer is buying the product from.

One of the challenges there is that what retailers are doing today is they're just looking for the closest store and they might be pulling that inventory out of a store where the product would actually sell on its own through that season, and it's sitting on the shelf in another store 100 miles away. They have an opportunity to optimize that and again, comes down to this inventory portfolio optimization. Which products where and then where to ship them from to get them to the end customer.

[MIUSIC PLAYING]

JOHN ANDREWS: Under the umbrella of inventory portfolio optimization, there's really two core solutions that we've been focused on and have had a lot of success with over the past year and a half, two years. One is around assortment optimization. And that actually breaks into three core modules when you think about the merchandising planning and allocation process versus what we call buy recommendation.

And this is used within the strategic planning and the merchandise financial planning where retailers are trying to identify, how much should I be spending on specific departments, or specific brands, or specific styles, and helping them understand what the demand is within their customer base and where they could actually sell those products based on all those different attributes and all those different factors.

The next is what we call buy quantification. And this helps the retailers figure out, OK, now I'm going to buy a particular shirt or sweater that I want to sell within my stores. How much of that product should I be buying? What's the demand going to be for that product? And helping retailers identify where they should be going big or where they should be buying small based on what the demand is going to be for that particular product, that's very early in the decision process. And getting that decision right is incredibly important in terms of what their final revenue numbers and markdown numbers are going to look like at the end of the season.

And then the final piece of the assortment optimization is the allocation piece. Now I've got an assortment of products. I know how much of each product I have. Where should I be allocating all of those products, get it into the stores in the right assortment and the right number of products in each store based on the buying patterns of customers within those stores? So that's the assortment optimization piece.

The other solution is omnichannel order fulfillment optimization. So this is with pretty much every retailer we're working with is looking at doing ship-from-store or is already doing ship-from-store. A customer buys a product online. The retailer is trying to push as much of the inventory into their stores, use those stores as fulfillment centers, and identify which-- then intelligently identify and optimize which store they should be shipping that product from.

Today, that's generally based on a rules engine out of a legacy order management system. But there's seven or eight different factors that go into the best place to ship that product from. And it's not a simple kind of rules engine that you need to go through and say, OK, what's the cheapest place if there's 10 stores that are the same price from a cost perspective? Then go to the next rule.

You need to optimize on all of those constraints. And when you can understand the demand for a particular product over the course of a season, you can make a much smarter decision in terms of what products to-- which store to ship those products from. So that's the omnichannel order fulfillment optimization solution.

Retail is clearly going through a transformation right now. People are-- you know, you hear in the news and in the press that x retailer is closing 100 stores, right? There's clearly a shift to online. But the reality is is that stores are not going away. Omnichannel, it's probably an overused term, but it's a very real issue and challenge for retailers to figure out how to leverage every interaction point, every channel with a customer, and be able to optimize across all of those different channels and provide the best experience to customers, whether they want to buy something online, return it to a store, maybe buy something online and pick it up in a store, buy something in the store and mail it back or return it back.

There is an enormous amount of complexity both from an operational perspective-- but at the end of the day, it all comes down to getting the right product in front of the right customer at the right time. And as part of this transformation in retail, it's going to evolve. It's going to change. Some stores you're going to-- some retailers are going to have less stores. Other retailers are opening up more stores.

The retailers who are able to truly understand how their customers are interacting with products, how the products are interacting with each other, and are able to optimize on that are the ones that are going to win. And you don't need to be Amazon to be able to use your data and make smarter decisions. And so every retail executive who we talked to, predictive analytics and leveraging machine learning to supplement the decision-making that their teams are making is at the top of their priority list. They quite simply want to understand how to use it and how it gets integrated within their environment.

And thankfully, over the past couple of years as we've been growing, we have more and more data points of how different retailers are using that information and using the science to help not take away the decision from the people but actually supplement that decision-making to help make better decisions to increase revenue, increase their-- and reduce stock-outs and reduce markdowns within the customer experience.

[MUSIC PLAYING]