Step by step to more intelligent forecasts

Why a collaborative, prove-it-as-you-go approach is critical to truly harnessing the power of advanced data and analytics.

Just about every business leader today understands the power of advanced data and technologies to deliver valuable new insights.

Artificial intelligence (AI), machine learning, terabytes of data, and predictive algorithms that train themselves to get smarter—these are now firmly established capabilities, not some science fiction vision of the future. They are powering more intelligent forecasting and generating bottom-line value for leading companies today.

But here’s what many business leaders still miss: Intelligent forecasting is much more accessible and executable than most realize. All that data and technology and the self-taught algorithms may sound like a massive enterprise planning initiative. But, by design, successful intelligent forecasting projects are highly iterative, step-by-step efforts that often show results in just a few months, as we outline in our thorough new report, Demystifying intelligent forecasting.

You don’t need to rebuild your entire forecasting platform in one giant project—and in our experience, that rarely works anyway. Instead, plan to go one step at a time, establishing capabilities and demonstrating results forecast-by-forecast, gradually expanding that recipe across the enterprise, and allocating resources and investment as you go.

It just starts with the first step.

One of the biggest barriers we continue to see with the move from traditional forecasting to more advanced predictive forecasting is that it sounds complicated.
From KPMG’s new report, Demystifying intelligent forecasting

Setting up for success

Iteration and steady expansion are key themes of effective intelligent forecasting initiatives, but that doesn’t mean “winging it” right out of the gate. Early alignment on both the short- and long-term goals is critical, with fierce collaboration across all appropriate teams.

Most important, the overall planning should be business-led and technology-enabled, with an invested function like finance or operations (or both) typically leading the way. And that concept of “fierce collaboration” is not simply corporate white noise: One of the most common mistakes that we see on intelligent forecasting initiatives is multiple teams testing multiple new tools without any coordination—often working from different data and technologies.

But a coordinated, step-by-step approach ensures that all are working from the same data, predictive tools, and business rules. Bonus: This iterative approach means the project’s required investment will scale as it proves value.

The roadmap for successful intelligent forecasting initiatives typically includes four foundational phases, as we outline in a related in-depth report, Intelligent forecasting: From proof of concept to connected capabilities. Here’s a quick look at each phase:


Plan
: Establish buy-in and collaboration, with early agreement on things like initial business rules, data sources, and forecast inputs and values. Early alignment is the key to developing forecasts that are accurate and trusted across the organization.


Pilot: Start with just a few forecasts to establish initial capabilities and learnings. This helps a company establish an initial framework of the signals, predictive tools and forecast outputs that perform best for them, pointing the way for the wider rollout.


Scale: Steadily extend the new forecasting capabilities and processes to other areas of the business, with adjustments for each function as needed. 


Govern: Monitor and manage on an ongoing basis to help ensure things like data integrity and regular evaluation of the predictive models. And don’t skimp on training and change management—intelligent forecasts are only valuable when they are used.

Essential components

For some companies, another barrier on intelligent forecasting is wrapping their collective head around the encyclopedia of technology terms. It can all sound, well, hard.

We’ve found that breaking things down into a few key areas can help, and it’s important to note: These are established tools and techniques that many companies are using today in their forecasting platforms. Here’s a closer look at a few key components:

Predictive models

This is intelligent forecasting’s “brain”—the algorithmic models that crunch the data and generate advanced insights. AI and machine learning enable the models to scan large amounts of data quickly—at speeds today that humans simply cannot match. There are hundreds of established predictive models publicly available, but most companies will drill down to just a handful that work best for their forecasting needs.

Data and signals

This is the engine room, powered by both internal and external data. Most companies can start with their existing data, and then can add new data over time, and especially as they are able to measure the increased value and accuracy. Incorporating new external data signals can be especially powerful—for example, using local weather and transit data to sharpen the accuracy of retail store revenue trends.

Visualization and interfaces

These platforms cover how users get their data—from spreadsheets to web dashboards to power-user apps—while ensuring that everyone is working from the same data and forecasts, no matter the output. These are typically integrated right into a company’s EPM system, allowing teams to collaborate and do sophisticated scenario-modeling.

Advance directly to “Go”

Step by step, steady iteration, and ongoing improvements. Done correctly, intelligent forecasting is the “project” that’s never finished. And that’s by design.

The overall business world will continue to change, and your competition—both the “known” of today and the “unknown” of tomorrow—will continue to come after your market share. But intelligent forecasting can help you maintain your edge with forecasts that are designed to continually evolve, improve speed and accuracy, and deliver powerful business insights that generate bottom-line value.

And, ultimately, it’s not just the algorithms that will get smarter over time.

Contact us

Brett Benner

Brett Benner

Partner, Finance Transformation, KPMG US

+1 267-256-2959
Peter Irwin

Peter Irwin

Principal, Lighthouse Data Analytics & AI, KPMG US

+1 212-872-7805
Colleen Schohl

Colleen Schohl

Director, Finance Transformation, KPMG US

+1 440-655-6258