Build Your App

Jet Forecasting

Jet Forecasting Logo

Customer Demand @Scale

Manufacturing companies, that forecast customer demand for their products, want to run their forecasts at the end of the month or beginning of the next month. They need the forecast results within a day or two for it to be useful.

Imagine that you own a cloud-based software solution that allows dozens or hundreds of manufacturers to run their forecasts, all at the same time.

Each company has hundreds of thousands of items they want to forecast. The demand on computing resources is extreme. How do you handle this? This is exactly the problem we had to face in developing Jet Forecasting.

We’ll address that issue in a minute, but the problem is actually worse than you think.

Machine Learning Algorithms

The customer, an expert in forecasting, required there be a 3-way method testing & validation of every forecast. In getting there, we run 20 different forecasting algorithms, or methods, that compete in order to find a winner.

That means there’s a lot going on for every single forecast.

This process uses data science tooling, and a proprietary machine learning process, to find the forecast that will most accurately predict future demand.

Demand by Customer & Locale

You might be inclined to think that each item a manufacturer makes gets 1 forecast. You’d be wrong. There’s a forecast for each:

Several thousand manufacturing items can quickly turn into a million forecasts!

The Solution - On Demand Computing

One of the most important benefits of cloud computing is the ability to scale demand up or down as needed. Jet Forecasting has this ability baked right into it.

Computing resources have a predetermined peak number of nodes per customer. This allows costs to be negotiated with the SaaS tenants and forecasts to be completed according to a predetermined schedule, usually within 1 to 2 days.

Utilizing the cloud provider’s API, we spin up predefined images that contain everything we need to distribute the load. This load is managed in order to achieve the optimum operating efficiency per node. A tenant running a forecast doesn’t interfere with the user experience of another tenant.

As a bonus to utilizing this architecture, we’re also able to make use of the computing languages and technologies that best fit the needs of each use case.

The Result

User Facing

The final product is a full user interface that allows users access to

Computational

On the backend, there’s a robust architecture for

This all happens without an administrator or user involvement aside from executing the forecast.

Our customer, the founder of Jet Forecasting, was able to achieve that which he set out to achieve. He knew, from day one, that the most important thing was the ability to process a million demand nodes in a couple of days, and we achieved that!

The platform excels functionally, with the needed performance to meet business needs. As noted on their website👇

We give you the power to generate intuitive forecasts with unprecedented speed and accuracy.

The Jet Forecasting platform is revolutionary because it learns your sales data and adapts forecasts based on your customized operational objective.

We are not a one size fits all solution to sales forecasting. We have over 20 different forecasting algorithms or methods competing to be the winner.

Jet Forecasting demonstrates how we can solve both tough business and technical problems. Let’s discuss how we can build your SaaS product - even if it seems like it is a difficult problem to solve!