Does it work on our data?

We check your data thoroughly, every month. Collecting and cleaning data is our core expertise.

Our demand is intermittent, with a long tail.

Our engine is specifically designed for products that sell only a few times a year. It learns from similar products and from the usual ordering patterns of your customers.

Our data is not ready.

We start from the data you have. Cleaning it is part of the project.

Can we start with one region or one product line?

We advise against it. With a pilot on one region or channel, we would have to clean the data twice, once for the pilot and again for the rest. A result on one region also convinces only that region. The other regions may argue that “their business is different”.

We already pay for planning software.

Great. It keeps running your planning, now with better forecasts from us.

Can you help us set up the forecast in our planning software?

No. That setup asks your planners to select models, tune parameters, clean outliers and sort products into ABC classes. We advise against all of it. Our model runs without tuning, and its forecast loads into your planning software. Your planners then make sure the model gets the right data. They enrich the forecast only when they know something the model cannot see.

Do we need a new platform?

The forecasts arrive as files or load into your current planning software.

We have our own data scientists.

Great. We can coach them, so they build faster and forecast better.

How should planners use this model?

We train and coach your planners. They learn what goes into the model and how to enrich the forecast where they add value.

Do you implement software?

We build and run the model, and deliver the forecasts to you as a service.

We reduce forecast error by up to 30%

One machine-learning engine forecasts your whole portfolio, new products included. It learns from your promotions, prices, shortages and sell-outs.

Start a proof of concept See what clients measured

What clients measured

The forecast engine created by SupChains allowed our demand planning team to improve our forecasting accuracy by 20 points in a matter of months, an objective we have had for a few years.

Sesh AddankiCOO, Vantage · specialty chemicals

…in the first 2 months of using SupChains’ model, over 50% of our forecasts were 100% ML-based…
Read the full quote
We worked to redesign our monthly forecasting process, including the machine learning baseline forecast in our forecasting tool. And in the first 2 months of using SupChains’ model, over 50% of our forecasts were 100% ML-based, with the remaining products having value added to the baseline by the planners, or being assessed to understand what data issues are driving significant differences vs planners’ assessment of the situation.
Still, machine learning is much less work than our previous experience implementing and maintaining statistical models. The initial results are strong across large parts of our portfolio.

Alex SugdenGroup Supply Chain Director, Animalcare · veterinary pharmaceuticals

…he delivered both a tailored inventory optimization model and a machine-learning-based forecasting model for Bridgestone Europe, which resulted in considerable improvements from our previous projects.
Read the full quote
Nicolas makes sure he understands well the business context, so any solution is properly applied when it makes sense. He focuses on what can make the difference, not wasting time on buzzwords that do not bring any relevant improvement. This has been my experience working with him, where he delivered both a tailored inventory optimization model and a machine-learning-based forecasting model for Bridgestone Europe, which resulted in considerable improvements from our previous projects.

Jon San AndresDirector Center of Excellence Supply Chain Planning, Bridgestone · automotive parts

Start with a proof of concept

You see what the model delivers on your own data before you commit to a service.

2–8 weeks

From your data to the result. The time to clean your data sets it.

Start a proof of concept

What we need: Your sales history, product hierarchy and product transitions, for your whole portfolio.
Read the details
We start with online or on-site workshops. Together, we review your demand planning process, agree the granularity and the horizon your supply decisions need, and set your objectives. Then we collect and clean the data and insights you have: the promotion calendar, product status and launch dates, sell-out, customer stock and forecasts, price lists and price changes.
What you get: Our forecast against a moving average, and against yours if you share it, on months the model never saw.
Read the details
We hold back your latest months and forecast them. You see error and bias per product group. You also get a data quality report, a readout with your team, and one month in the App to investigate the results in detail.
What sets the price: The number of products, the complexity of your data and the days on site.
Read the details
You decide on the service only after you see the result of the proof of concept.
What comes next: Forecasts as a Service with support for your planners, or a licence.
Read the details
We deliver the forecast every month, in the App or as files for your planning software. We also help your planners clean the input data and improve their Forecast Value Added. If you choose to buy a licence of our model, your team runs the engine.

What better forecasts are worth

Better forecasts cut inventory and shortages at the same time. Here is how much, according to three studies and our own simulations.

−2.7%

finished-goods inventory for every 1% of forecast improvement, with 3.2% lower transportation costs and 3.9% less obsolete inventory.

Gartner, “Win the Business Case for Investment to Improve Forecast Accuracy”, 2017. Consumer products.

+3%

pre-tax improvement or more, from a 15% forecast accuracy improvement, in the experience of the Institute of Business Forecasting.

Institute of Business Forecasting, “How Much Does Forecasting Software Cost, & How Much Will It Save?”, 2018.

−5%

inventory costs, and 2 to 3% more revenue, from 10 to 20% better forecast accuracy.

McKinsey, “Most of AI’s business uses will be in two areas”, 2018. Advanced manufacturing.

−4%

inventory, or 6% fewer shortages, from 10% better forecasts.

SupChains simulations, “What’s the Business Impact of 10% Extra Forecast Accuracy?”

I read the first three as orders of magnitude. Our own simulations give smaller numbers, and the gain is still large.

Test it on your own numbers

Impact 1

We increase service level while keeping same inventory

–Extra revenue thanks to higher service level – computed based on lost sales
–Extra profits thanks to higher service level – computed based on lost sales
–Extra revenue thanks to higher service level – computed based on service level
–Extra profits thanks to higher service level – computed based on service level

in m$/year

Impact 2

We reduce inventory while keeping service level constant

–Cash saved because of inventory reduction in m$
–Savings thanks to inventory reduction in m$/year
–Savings thanks to obsolete reduction in m$/year
–Total savings in m$/year

in m$/year

Nicolas Vandeput teaching at a whiteboard

About half of the changes planners make worsen the forecast

That is what Fildes and Goodwin found.

So we coach your team to do better.

The engine

One global machine-learning engine forecasts every product. It uses your business drivers: promotions, prices, shortages, sell-outs, client inventory, confirmed orders, even the weather.
It forecasts new products before they reach the market.
Hundreds of checks run before and after each forecast, and a report lists every data inconsistency.

Your planners

Your planners enrich a forecast only when they get specific insights that the engine is not aware of.
The SupChains App measures what each change adds (Forecast Value Added).

Your data team, your planning software, a consultancy, or SupChains?

SupChains

Right when you want:

one team of forecasting experts to review your demand planning data, process and objectives
a model built and run on all your data and insights
forecasts every month in your planning software
planners trained and coached to enrich forecasts

Your own data team
Right when your data scientists are forecasting experts, with time to build and test their own engine, and you want to own the intellectual property of the model.
Your planning software’s forecast
Right when it beats a moving average by 20% or more and uses all your business drivers by itself. Your planners then spend their time on insights, instead of hunting for promotions and stockouts.
A large consultancy
Right when you need a team on site for months, for a new planning process or a software rollout.

Questions clients ask us

Does it work on our data?: We check your data thoroughly, every month. Collecting and cleaning data is our core expertise.
Our demand is intermittent, with a long tail.: Our engine is specifically designed for products that sell only a few times a year. It learns from similar products and from the usual ordering patterns of your customers.
Our data is not ready.: We start from the data you have. Cleaning it is part of the project.
Can we start with one region or one product line?: We advise against it. With a pilot on one region or channel, we would have to clean the data twice, once for the pilot and again for the rest. A result on one region also convinces only that region. The other regions may argue that “their business is different”.
We already pay for planning software.: Great. It keeps running your planning, now with better forecasts from us.
Can you help us set up the forecast in our planning software?: No. That setup asks your planners to select models, tune parameters, clean outliers and sort products into ABC classes. We advise against all of it. Our model runs without tuning, and its forecast loads into your planning software. Your planners then make sure the model gets the right data. They enrich the forecast only when they know something the model cannot see.
Do we need a new platform?: The forecasts arrive as files or load into your current planning software.
We have our own data scientists.: Great. We can coach them, so they build faster and forecast better.
How should planners use this model?: We train and coach your planners. They learn what goes into the model and how to enrich the forecast where they add value.
Do you implement software?: We build and run the model, and deliver the forecasts to you as a service.