Interview: How to use the quantum mechanics of retail to improve profitability.

PDF File: 

ASC7_wp_nechleba2.pdf

4R Systems, Inc.

Jiri Nechleba and Ed Rennemann explain how touse the quantum mechanics of retail to improveprofitability.

ASCET: What has changed in Crate and Barrel’s inventory management process since 4R and Crate and Barrel began working together?

Ed Rennemann (ER): We had not changed our store replenishing process since 1984. Historically, we have given our store managers the visibility to see what was available in the DC along with rates of sale and other essential information about the SKU, but it was at the discretion of the store manager to determine the quantity to order. This is very much in keeping with our philosophy that we like to see the store managers as merchants who know their customers. This gives them quite a bit of autonomy to make localized decisions.

Going forward, we didn’t want to lose that. What we have done with 4R is change to a profit-based model. 4R takes our information, does a sales forecast, applies a gross margin, profit-based, inventory carrying-cost model and recommends an order amount. The only change from a store perspective is that we have taken the ordering worksheets that the stores have used for the last 20-plus years and augmented them to include 4R’s recommendation. It’s still up to the store manager to review all the information in the worksheets; look at 4R’s recommendation; and use their gut feel to decide how to merchandise. It’s a blend of art and science.

Jiri Nechleba (JN): We looked at Crate and Barrel’s historical data to understand sales patterns in the stores and the variability or uncertainty of sales. We applied our models to that historical data, and estimated how better store inventory decisions could drive higher profitability and customer service as well.

Merchants understand that if they put in lots of inventory, customers will always find merchandise. But they’ll be awash with inventory costs. At the other extreme, lower inventory lowers cost, but customers can’t buy what’s not there, and sales are lost.

We bring mathematics, models, analytics and optimization technology to identify the optimal amount of inventory for each item, in each store and for a specific week that will generate the most profit. At Crate and Barrel when we analyzed past sales, we found their store-based, experienced merchant system to be among the most efficient we’ve seen. Nonetheless, there was still significant opportunity to be more profitable.

We aligned with Crate and Barrel’s store-driven culture. We saw that involving them and giving them an understanding of the information and ownership of the decision was key to driving profitability. It’s impossible for somebody to do a P&L on every unit of inventory, and it’s very hard to bring all the odds and payoffs to bear and come up with the best answer. The store managers understand very well that they should be raising or lowering inventory, but we help them with the difficult challenge of setting those amounts profitably.

Ed Rennemann (ER): A lot of it does come down to assessing the risk of having too much or too little inventory. One of the things I’ve heard from the store managers is that in the past, if they had a big spike in sales, they didn’t know if it was a fluke. Part of what 4R has done though their software and forecasting is help us understand the likelihood of a spike occurring and giving insight as to whether to bet the inventory on the possibility of a spike.

ASCET: How did you work together to find the right blend of art and science given Crate and Barrel’s unique culture? What was the training process like?

Ed Rennemann (ER): We did pilot stores, and we have had the Crate and Barrel IT people intimately involved in working with 4R in terms of the research and development of the solution. In the first pilot group, our IT people and 4R worked collectively for a full day with the pilot store management on concepts and goals and, most importantly, revised ordering worksheets and recommendations based upon their feedback. They talked through where the 4R recommendation did not align with their gut feel.

Jiri Nechleba (JN): Every week retailers makes the decision on how much inventory to carry on each item in each store. One of our customers makes over 5,000,000 of these “bets” on inventory every week.

We provide store management with reports that recommend 8 of SKU A, 2 of SKU B, etc. But before we showed them these numbers, we also gave them an economic framework to understand how inventory relates to profits. So they understand that the recommendation is the point at which, if they put in more inventory, the cost of the extra inventory is not likely to be covered by increased sales or vice versa – the point at which if you put in less you’re going to lose more margin dollars than what you save in inventory expense.

With Crate and Barrel staff, we did a lot of teaching and learning. They were very knowledgeable about this problem and had a very good subjective feel for it. We wanted to augment their skills and knowledge with an understanding of the economic model. We did not want to just give them a number; we wanted to get them to relate to that number, inspect that number and understand it. This builds on the culture of intelligent, responsible people in the store. There’s a really nice symbiosis between their view of the world and our ability to capture that in analytics.

ASCET: What does your software look at, Jiri, and how is it generating recommendations?

Jiri Nechleba (JN):Our analytics and optimization engines generate recommended inventory levels for each store for each item based on weekly gross margins, costs and current prices. It’s left to the store designers and managers to use their knowledge of their customers and local markets to keep the store fresh, inviting and interesting and drive traffic and sales.They appreciate getting a number because, previously, they had a rough idea of where to set inventory, but, in many ways, it was a guess; it was very hard for them to assess the odds and arrive at the most profitable answer. Our information, and the reasoning behind it, let them set inventory more efficiently.

ASCET: What are some of the new metrics that you put in place to analyze inventory?

Ed Rennemann (ER): Historically, we have used typical retail metrics such as weeks of supply, sales-to-inventory ratios and turn. We have shifted to more profit-based metrics. Currently, we are looking at a combination of inventory carrying costs and a measurement of lost sales.

Jiri Nechleba (JN):Crate and Barrel had reasonably good ways of measuring the cost of inventory, but we gave them a way of understanding lost sales. With respect to typical metrics such as weeks of supply, we provided a better understanding of low-frequency events. For example, a certain level of inventory might only be needed once a year, but if the product has a very good gross margin, it’s worth stocking for that situation. Retailers place bets just like gamblers in a casino, and they place lots of them every week. So if you can bet that system more efficiently week after week, just 1 percent more efficiently, it adds up to a lot of money.

Ed Rennemann (ER): : I cannot emphasize how important the metrics are. As simple as it may sound, measuring lost sales and looking at things like the additional gross profit from having a little bit more inventory, has had a big strategic impact. It’s making us look at the business in a different way, and it’s definitely impacting our overall strategy.

Jiri Nechleba (JN): Retailers use a lot of inventory statistics, such as in-stock levels, weeks of supplies, etc. We’ve gotten very comfortable with those numbers to measure inventory. But we have to remember that one reason we measure those statistics is that they are easy to calculate; however, they are surrogates for things that we really care about. An old-line retailer will tell you that what they really want to know is how many sales they lost.

Our founders, professors from Wharton and Harvard, saw this problem and realized that if people really want to know lost sales, that while in-stock and outof- stock are related to lost sales, it’s not a linear relationship. They found basing decisions on lost-sales estimates led to better outcomes. For instance, if you’re in-stock 98 percent of the time, the profit impact for an item that sells once every four months is very different from an item that sells 40 a week. For the item that sells 40 a week, a 98 percent in stock leads to lots of lost sales and profit.

ASCET: Has profitability increased for Crate and Barrel?

Ed Rennemann (ER): It has, and we’re still in the roll-out stage. We’ve done enough measurement to know how much profitability improved at each store, and we have done extrapolations to the chain. We typically don’t share specific financial information, but the payback is less than a year.

ASCET: What are the most effective changes introduced by 4R?

Ed Rennemann (ER): From the executive level standpoint, it’s the introduction of accurate, dependable lost sale measurements. We’re really using that in our strategic thinking. The nuts and bolts is providing a number, a recommendation using a lot of calculations that nobody in the store could do.

Jiri Nechleba (JN): One of the things that I perennially see in retail is a cycle that starts with too little inventory and low service. So, we start building inventory and sales grow. Soon, service levels are great, but we’re awash in inventory, and costs are too high. So, we go on an inventory diet. It’s this constant oscillation between a Soviet shopping experience and a Turkish bazaar.

It’s really a lot of wasted energy. What retailers need to know is how to set the level of inventory to generate the most profit and then sail a steady course. We help our clients determine that level, in total and itemby- item.

We take into account many characteristics of a product – its margin, expected sales level, sales volatility, lead time to get more, where in the lifecycle it is – to name just a few. What we’re forecasting is actually not a number, but a pattern.

Everybody likes the part of the forecast that tells you how many units of a product you’re going to sell next week. But nobody likes the part that says that it might be as high as 700 with a certain probability, or as low as 300 with a certain probability.

We look at the error of forecast, and we use that as a way of understanding the pattern of demand. We then use that as an input into an optimization function that deals with the costs of inventory and the costs of lost sales to identify the most profitable inventory point. Forecast errors are large. We call it error because we can’t predict it. But I would argue that it’s a false word; it leads us to a false conclusion. The truth is, it’s not error; it’s variability. If we understand that what it represents is an estimate of the variability of demand - rather than our inability to predict - we can start to leverage that knowledge.

Ed Rennemann (ER): That approach is what makes the 4R solution unique. We looked at a variety of forecasting tools, and most everybody thinks about it as the number. The forecast of sales for the upcoming week is 500, and they may also provide you with a confidence in the number – our confidence is high or low that it will be 500. What 4R does is forecast that it could be anywhere from 400 to 600, and depending on the margin of the product or the cost to carry it in the store, we can make our decision. The bet is ours.

Jiri Nechleba (JN): Let’s say 500 is your forecast, and there’s a certain probability it’s going to sell 300 or 700. If the product has a 90 percent gross margin, every sale makes you 90 cents, at a cost of 10 cents on the sales dollar; you want to inventory the heck out of that product, right? You put up a dime, you get 90 cents back. Boy, I don’t care that the probability of selling 650 units in a given week is only 2 percent. On the other hand, if the product has a 90 percent cost or 10 percent gross margin, I don’t want to hold too much.

This notion of uncertainty permeates decisions throughout the supply chain. Most retailers have historically looked at supply chain deterministically. Gordon Segal, the founder and CEO at Crate and Barrel, resonated with our ideas about uncertainty. Gordon knows that some weeks you sell out and some you don’t. He’s probably scratched his head a million times trying to figure out why it’s high one week and low the next, and I think he understands that there’s a lack of credibility when you forecast an exact number. Somebody who’s really store-savvy doubts an exact forecast. They respect the notion that there’s variability. Therefore, a model of the pattern of variability offers a higher level of credibility and accuracy.

ASCET: What you’re describing then is a 20th century model of mathematics that incorporates probability?

Jiri Nechleba (JN): As someone whose degrees from MIT are in astrophysics, I use that analogy a lot. The old world is the Newtonian world where everything moves in predictable ways. When we look at a store’s data, we realize it’s bouncing randomly like new-world quantum mechanics. If we understand the quantum mechanics of retail and look at supply chain from that perspective, it allows us to improve the process by understanding uncertainly rather than pretending that it doesn’t exist. More importantly, it allows us to create increased profit for retailers by applying 21st century probability models to retail.