All Posts

A Nuanced Perspective on Harvard Business School’s Jagged Technological Frontier

Conor Doherty reviews a paper by Harvard Business School called Navigating the Jagged Technological Frontier. TLDR: Quality is subjective; cost is not.

Lokad passes 20,000 Followers on LinkedIn

Another milestone in the ongoing development and adoption of the Quantitative Supply Chain ethos is passed.

An opinionated review of the Yann LeCun interview with Lex Fridman

Joannes Vermorel reviews Yann LeCun's remarkable interview with Lex Fridman titled “Meta AI, Open Source, Limits of LLMs, AGI & the Future of AI”.

Distribution Network Analysis through Envision - Workshop #3

This third Envision Workshop offers students and supply chain specialists guided training on analyzing the distribution network of a retail company using Lokad's Quantitative Supply Chain perspective and tooling.

AI chatbot for supply chains and Lokad

Lokad’s own digital assistant has been trained as a supply chain expert. This chatbot has been trained based on the bulk of Lokad’s own supply chain materials. Access to this chatbot is provided free of charge, courtesy of Lokad.

Probabilistic Exponential Smoothing for Explainable AI in the Supply Chain domain

Read Antonio Cifonelli's PhD research on Probabilistic Exponential Smoothing for Explainable AI in the Supply Chain domain—another great study showing that half a dozen clever tricks can turn a model as basic the exponential smoothing into a racing car that beats state-of-the-art Deep Learning models.

Sales Analysis through Envision - Workshop #2

This second Envision Workshop offers students and supply chain specialists guided training on analyzing retail customers from Lokad's probabilistic, risk management perspective.

Selective Path Automatic Differentiation: Beyond Uniform Distribution on Backpropagation Dropout

The Selective Path Automatic Differentiation (SPAD) approach enhances Stochastic Gradient Descent (SGD) by adopting a sub-data-point perspective. This technique, implemented at the compiler level, trades gradient quality for gradient quantity, supplementing traditional SGD methods with a more nuanced view.

An opinionated review of Deep Inventory Management

A team at Amazon has published Deep Inventory Management (DIM) late 2022. This paper presents an DIM inventory optimization technique that features both reinforcement learning and deep learning. As Lokad went through similar path in the past, its CEO and founder Joannes Vermorel provides his critical assessment of the suggested technique.

VS Code extension for Envision.

Envision, the domain specific programming language dedicated to the predictive optimization of supply chains just got its first open source VS Code extension that enables code coloring and code validation via compilation.

Differentiable programming to optimize over large scale relational data

Paul Peseux's PhD research on differentiating relational queries - another under-researched area of supply chain - introduced TOTAL JOIN operator, Polystar and a mini-language ADSL to differentiate relational queries, all of which Lokad integrated into its DSL Envision as part of autodiff for optimizing daily inventory decision-making.

Supplier Analysis through Envision - Workshop #1

Lokad launches its first Envision Workshop, teaching students (and supply chain specialists) how to analyze retail suppliers using Lokad's probabilistic, risk management perspective.

Inventory management under the constraint of multi-reference minimal order quantities

Gaetan Delétoille's PhD research on MOQs - a surprisingly under-researched area of supply chain - introduced the w-policy, something Lokad integrated into its solution for daily inventory decision-making.

Classification algorithms distributed on the cloud

Matthieu Durut, second employee at Lokad, defended his PhD back in 2012 for his research work done at Lokad. This PhD paved the way for the transition of Lokad toward cloud-native distributed computing architectures, nowadays critical to deal with large-scale supply chains.

Large scale learning: a contribution to distributed asynchronous clustering algorithms

Benoit Patra, first employee at Lokad, defended his PhD back in 2012 for his research done at Lokad. This PhD brought radically novel elements to the supply chain theory, and set the stage for the future development of Lokad's probabilistic forecasting approach.

RFI, RFP and RFQ madness in supply chain

Over the years, Lokad has gained some notoriety, and we have earned the privilege of being part of an increasingly large number of RFI (request for information), RFP (request for proposal) and RFP (request for quote) related to supply chain optimization.

The maintenance of the optimization

Lokad offers a flat monthly fee for software and expert services, ensuring optimal supply chain performance. Traditional optimization software fails due to inadequate maintenance. Lokad's supply chain scientists maintain automation, requiring ongoing engineering resources to prevent reverting to inefficient spreadsheet management.

Let’s try Lokad

The engineering teams at Lokad have spent the last decade refining Envision, a specialized programming language dedicated to the predictive optimization of supply chains. As it turns out, most supply chain optimization challenges are, well, weird (for lack of a better word) and mainstream software technologies fall flat on their faces when confronted with such challenges.

The role of IT

In many, probably most, large companies operating a supply chain, the IT department has years of backlog. The backlog is made of a myriad of glitches, inconsistencies, or fragilities that could be fixed but aren’t. Besides creating vast bureaucratic overheads, this endless series of pointless annoyances demotivates everyone.

Stochastic gradient descent with gradient estimator for categorical features

The broad field of machine learning (ML) provides a wide array of techniques and methods that cover numerous situations. Supply chain, however, comes with its own specific set of data challenges, and sometimes aspects that might be deemed basic by supply chain practitioners do not benefit from satisfying ML instruments – at least according to our standards.

LokadTV reaches 6000 subscribers

Almost exactly 18 months ago, I wrote a post to celebrate passing 3,000 subscribers on LokadTV. That post, in fact, was written almost 18 months after we celebrated passing 1,000 subscribers. Curiously enough, we passed 1000 subscribers about 18 months after we began actively promoting the channel.

On noise, NPCs, chatbots, and supply chain

The state of supply chain communication is dismal. By way of anecdotal evidence, for the last couple of months, my daily morning routine consists of attempting to find one post – just one – that qualifies as _worthy of interest_ to an audience of supply chain minded people – in order to have it shared on news.lokad.com. This seemingly simple exercise proves exceedingly difficult. Most days, I do not succeed.

Supply chain resilience requires bandwidth

The last couple of years haven’t been kind to supply chains. In fact, the only thing that hasn’t been in short supply appears to be forces of disruption':' lockdowns, wars, inflation, etc. As a result, a newfound interest in ‘resilient’ supply chains has emerged.

Pitfalls of supply chain for aeronautics

Aeronautics companies are fundamentally risk adverse – for very good reasons – and this reflects in the management of the companies’ activity in every area. However, this mentality can be a hindrance when it comes to supply chain, as avoiding risk and not integrating the very concept of risk at the heart of the supply chain means losing sight of what you are optimizing.

Differentiating Relational Queries

Supply chain data present themselves almost exclusively as relational data such as orders, clients, suppliers, products, etc. Those data are collected through the business systems - the ERP, the CRM, the WMS - that are used to operate the company.

Reproducible Parallel Stochastic Gradient Descent

The stochastic gradient descent (SGD) is one of the most successful techniques ever devised for both machine learning and mathematical optimization. Lokad has been extensively exploiting the SGD for years for supply chain purposes, mostly through differentiable programming. Most of our clients have a least one SGD somewhere in their data pipeline.

Launching Supply Chain News

The ambition of news.lokad.com is to create a vibrant online community. No terms to accept, no forced login, no ads, no monetization plan, just a plain discussion text-based board, with a voting system to let the most interesting contributions surface.

Forecasting Climate Change like a Supply Chain Planner

Transcript of the talk given by Joannes Vermorel at the Ecole polytechnique in Palaiseau (France) on Friday, June 3rd for the symposium 'Artificial Intelligence, Digital and Climate Change'.

Grand experiment on the TCO of a supply chain software

The price tag of a piece of software ranges from nothing, as it happens with open source, to quite a lot - enterprise software leaning heavily towards the latter. However, operating a piece of software always involves some degree of overhead. The notion of TCO (Total Cost of Ownership) precisely tries to capture the cost in full, taking into account both direct and indirect costs.

Control and bureaucracies in Supply Chains

Over the years, it has become increasingly frustrating to witness that most companies seeking to improve their supply chain performance are setting themselves up for failure through their own RFP (request for proposals) and RFQ (request for quotes) processes.

How many people does it take to change a supply chain light bulb?

Supply chains involve a patchwork of enterprise software. These software layers have been gradually, and sometimes haphazardly, rolled out over the last four decades. The venerable EDI (Electronic data interchange) may sit next to a blockchain prototype.

Envision VM (part 4), Distributed Execution

The previous articles mostly examined how individual workers executed Envision scripts. However, both for resilience and for performance, Envision is actually executed across a cluster of machines.

Envision VM (part 3), Atoms and Data Storage

During execution, thunks read input data and write output data, often in large quantities. How to preserve this data from the moment it is created and until it is used (part of the answer is on NVMe drives spread over several machines), and how to minimize the amount of data that goes through channels slower than RAM (network and persistent storage).

Envision VM (part 2), Thunks and the Execution Model

Like most other parallel execution systems, Envision produces a directed acyclic graph (DAG) where each node represents an operation that needs to be performed, and each edge represents a data dependency where the downstream node needs the output of the upstream node in order to run.

Envision VM (part 1), Environment and General Architecture

A Supply Chain Optimization pipeline covers a wide range of data processing needs':' data ingestion and augmentation, feature extraction, probabilistic forecasting, producing optimal decisions under constraints, data exports, analytics, and dashboard creation.

Beyond time-series

When all you have is a hammer, everything looks like a nail. The hammer long favored by the supply chain community has been time-series and, as a result, in supply chain circles all problems look like time-series forecasts. The hammering temptation is compounded by the extensive literature that exists on time-series forecasting beyond supply chain use cases.

Incrementalism is the bane of supply chains

Incrementalism in supply chain includes improving the forecasting accuracy, improving the service level, reducing the stock level, reducing the lead time. Despite its prevalence among large companies, this approach rarely yields any tangible benefit.

Supply Chain as a Service

In terms of predictive optimization, most supply chains are stuck in the early 1990s. As we started to address the root cause of predictive optimization failures, Supply Chain as a Service emerged as our business model.

Tidbits of supply chain terminology

The emergence of a terminology is, at best, a haphazard process. Supply chain is no exception and, in hindsight, a sizeable portion of supply chain terminology is inadequate. Confusing terminology hurts newcomers and seasoned practitioners alike. Newcomers struggle more than they should with accidental complexity. Practitioners may not realize that the premise of their field ismore shaky than it appears.

How many SKUs should a demand planner manage?

Lokad employs a team of around 30 supply chain scientists. The scientist is responsible for the data pipeline, the forecasts, the economic modelling and the end-game decisions. But how many SKUs do they manage?

LokadTV reaches 3000 subscribers

18 months ago, LokadTV had reached 1000 subscribers. Since that time, our audience has now tripled reaching 3000 subscribers. We remain committed to bringing topics of genuine interest to the supply chain community.

Why FTP instead of REST

Most web apps feature web APIs styled as REST, yet Lokad features FTPS and SFTP, which may appear surprising. However, this choice is intentional, why did Lokad choose to go this route?

Supply Chain Story: The Bullwhip Job

My first professional supply chain experience happened back in 2004. At the time, I was a computer science student at the Ecole Normale Supérieure (ENS), a university in Paris. My interests covered a wide range of wholly theoretical subjects, yet, I was also intrigued by the idea of testing out those theories in the wild.

The Adventures of Captain Obvious in Supply Chain

Captain Obvious has been working overtime in supply chain. His natural leadership has inspired many who are now following the same path. Yet, underneath the uniform, there is little to be found but a great deal of confusion.

42 tricks to achieve Supply Chain greatness

Many domains are complex, irredemiably so, and supply chain is certainly one of them. So what could an itemized list of tricks help you to achieve?

On the accidental complexity of supply chain systems

A single smartphone could technically run a predictive inventory allocation for a very large retail network, yet why do so many companies have such slow systems in place?

Supply Chain Lectures, but why?

I would like to share some insights into my motivations, as a CEO, for undertaking such a series of supply chain lectures. It boils down to the four key challenges faced by Lokad.