A company manufactures too few units because the procurement department second guesses the sales plan. After all, the sales team missed its forecast the past few months, so procurement adjusted accordingly. Why should the company get stuck with excess inventory and the financial repercussions? But is it fair to blame the sales team when it hits its numbers, and nails the coordinated sales and marketing plan? Because of procurement’s decision to change the forecast, the company is now not only short of needed products, it's short of the materials it needs to make those products.

If the various departments along the supply chain, from procurement to manufacturing to logistics, don’t operate from a common plan, expect surprises. Yet coordinating those plans is difficult, especially in large operations. The difficulties that arise are good lessons for CFOs to heed and use as a basis for setting up a centralized process to ensure there's only one source of truth.   

"People tend to want to run with what is in their control, make their own plan, their own forecast," Greg Spira, principal at Oliver Wight Americas, told Supply Chain Dive in an interview. When that urge starts at the top, the results may be so far removed from the actual operational details that the forecasts are inaccurate. And without the drive to coordinate, the forecasts will be disconnected from other department plans. Planning in a silo can do that to a company.

---

"There’s so much fragmentation in the systems. There is no reference data available."

Shariq Mansoor

Founder and Chief Technology Officer, Aera Technology

---

Spira recommends eliminating the siloes by having one forecast, not individual department forecasts. The concept is delegating forecasting to a team serving all departments equally. They consider that the financial department needs these elements, and supply chain needs something else. Sales wants to know revenue by customer, while marketing prefers revenue by brand.

"The trick with a central organization is everyone must trust [the central forecasting team] or everyone will create their own views," Spira said. The group must be staffed with strong analysts who understand the business, not just great statisticians who spit out numbers and aren’t cognizant of how the company operates. "That won’t get you anywhere," Spira said.

The importance of baseline and assumptions

Spira also recommends all departments agree on a baseline and forecasting assumptions. After all, if everyone uses their own baseline numbers, the resulting forecasts will be skewed. Ideally, the various departments learn to trust one group to determine that baseline, building forecasts from that.

The forecasting assumptions have to be coordinated as well, otherwise it’s difficult to determine a decision’s impact. If groups offer different forecast assumptions, each will have its own bias. One group might look at what it did previously and change the forecast by 5-10%. "The problem is, you’re doing that at a high level, not with granularity," Shariq Mansoor, founder and chief technology officer at Aera Technology, told Supply Chain Dive in an interview.. Without forecasting at a SKU level, a company’s volume might look the same, but the categories have problems, like shortages in a 24-pack of one product, with an excess of a 10-pack of the same product.

Departments have to stick with their assumptions and not change them without agreement from other departments. If a new product is launching in June, the various groups must agree on the launch date, what percentage of their other products will be cannibalized by this launch and what sales and marketing will do to support the launch. If one group determines that the product will actually launch late, like in July, and changes its plan without getting other departments on board, the process will go off the rails, according to Spira.

That centralized forecasting group is responsible for facilitating and reconciling disagreements about the assumptions. If departments have differing assumptions, and they agree to just split the difference, there’s no accountability for the final number, and it’s not based on real data. "Reconcile assumptions, not numbers," said Spira.

Modern forecasting: Cut out the humans?

Add in different software programs, multiple distribution centers or manufacturing sites (possibly in different countries) and a dizzying number of multiplying SKUs, and it’s understandable that technology and data analytics are getting more popular. The planning process has become so complicated that humans can no longer do it, said Mansoor.

With the quickening pace of business and excessive amounts of data, the forecasting cycle can no longer keep pace. Mansoor’s solution is to cut out the humans, who he says can’t get to the granular level with so much data available. The plethora of data can actually hold an analyst back from creating forecasts.

---

"You can unbake the cake when looking at basic forecasting on Excel. When looking at machine learning, good luck."

Greg Spira

Principal, Oliver Wight Americas

---

In 2020, combining technology with human analysts means statistical planning has to be done at a high level. "As a human, that’s all we can manage," Mansoor said. Computers can plan and forecast, monitoring for changes at any time increment, from days to minutes. Humans can instead help where machines cannot: launching new products and SKUs, for example.

Spira said advanced analytics can make it near impossible for everyone at the table to reconcile their forecast or plan with someone else’s. Using an Excel sheet and basic forecasting techniques, different department plans can be compared in an unsophisticated, line-item way. However, in using advanced analytics for forecasts, reconciling between departments is nearly impossible, at least without some conditions. "You can unbake the cake when looking at basic forecasting on Excel. When looking at machine learning, good luck. It’s not going to happen," Spira said. While these analytics are incredibly powerful, companies will throw them out instead of leveraging them, he said.

A single source of data

While the approach of avoiding machine learning and complicated analytics sounds anti-technology, it’s not. It involves changing how businesses forecast. Mansoor suggests using software with one layer that encompasses fully processed and harmonized data from all systems, so that employees from all departments can work with the same data. "One of the biggest challenges in enterprise today is they have multiple systems, and there’s so much fragmentation in the systems. There is no reference data available, no single version of data available, so you’re limited to what you can plan," he said. These systems lack visibility, and it’s difficult to compare apples to apples.

By presenting enterprise data in one standardized layer, planners can access it across the board, with an understanding of all supply chain factors in real time. "You don’t have to wait a month for a collaboration plan. You can run the plan daily and adjust it," Mansoor said.

Automation and augmentation of the planning process that results in accurate forecasts can make a difference for a company. "Amazon has no product innovation. They’re using planning and supply chain as a competitive advantage. They’re doing it using machines, more dynamic and in real time," Mansoor said.