Save content
Have you found this content useful? Use the button above to save it to your profile.
'Black box' AI forecasts lack the transparency CFOs need to tell stories
iStock_Brostock_Navigation display

Strategy: Use numbers you know to steer the ship

by

When it comes to business analysis, comparing actuals to the known logic of budgets and forecasts is more effective than predictive AI models, according to Unit4’s head of FP&A.

18th May 2022
Save content
Have you found this content useful? Use the button above to save it to your profile.

“Steering by numbers” is the guiding business strategy principle for Unit4 vice president of forecasting, planning and analysis (FP&A) Michael Lengenfelder.

It’s a simple idea, based on known business logic contained in business forecast models and budgets that allow finance managers to compare actual performance against expectations. In the fast-moving, data-shaped world of modern finance, however, there are two ways to generate those forecasts. And even though his employer has access to the latest artificial intelligence modelling and programming tools, Lengenfelder prefers using known numbers to the growing trend for AI-generated forecasts.

“Predictive AI forecasts are based on putting historic data into an AI black box. With zero effort, the system will deliver the forecasts for you. But then the issue is, do you buy the number? There’s no reason why the revenue forecast for Asia-Pacific is £1m. Only a data scientist can tell you. That’s the issue I see with super-automation – transparency and understanding,” Lengenfelder said in a wide-ranging conversation with AccountingWEB.

“There’s no question about the great use cases for AI. At Unit4, for example, we have the Wanda chatbot based on language recognition, along with smart invoice recognition and processing. Those are great applications of AI, because it learns from past accounts selections. But when you move into the planning area there are challenges: will you be able to understand and explain the data? Maybe you’re looking at a new market with special challenges that you don’t have any data for.”

The strategy formula

For someone who works with advanced technologies, Lengenfelder has a refreshingly simple approach to business strategy: “If you want to steer the organisation and be successful in a certain market, you set a target. Then what do you need to do to achieve that?”

Once the strategic vision of what you want to achieve has been defined, the finance team’s job is to break the objectives down into high-level goals and refine the financial inputs and outputs further down into a budget.

“Then you try to achieve the budget and compare it to the actual results: are we on track and how do we correct any deviations,” he explained.

The advantage of what Lengenfelder calls this “value driver data” approach is that it’s understandable, because it’s based on A+B+C logic. “You can model it and follow it,” he said.

Setting budgets at the right level

If only life were so simple. As anyone who’s been through a business planning cycle will know, it’s an imprecise and elusive experience. As Lengenfelder acknowledged, one of the hardest challenges for managers to master is forecasting and budgeting at the right level of detail to actually affect the numbers.

“For example, compare a top-level global group budget with no detail versus a company based on product plans and regional budgets. In the first case, if you compare actuals to the budget, there aren’t many stories you can tell. You can only tell what you have achieved on a high level. If you budget on a region and product level, you can tell more stories – for example to explain why you might want to invest more in a particular region or product, or explain why you failed somewhere else.”

How do you get to that data? 

Working within an integrated enterprise resource planning (ERP) software group, Unit4’s FP&A chief has an ongoing debate with specialist business intelligence software developers. It’s a discussion that will be familiar to users of accounting “ecosystem” apps such as Spotlight Reporting and Fathom who are seeing the forecasting software market consolidate  around Sage, Xero and QuickBooks at the small business level.

“The key point of all of this is integration with other systems,” Lengenfelder explained. “An FP&A system is empty – it needs data from other sources. An ERP system like ours is preconfigured with connections into those sources, so you don’t need to worry. You can load in data and push it back out.”

In higher education, one of Unit4’s core vertical markets, there are specific requirements that drive financial planning based on how many students are involved and what fees they generate. The FP&A software for the sector can be populated with standard planning logic based around all of the drivers that feed into those standard measurements.

The Excel factor

Much of this industry-specific logic is known to finance people working in different sectors. And many of them are perfectly content to build models and forecasts for themselves in Excel. This is exactly what Lengenfelder did himself when he worked in industry and it’s how he and his Unit4 FP&A colleagues go about creating new models they incorporate into their planning suite. 

“From a platform perspective, Excel is comparable,” he said. “You can build business logic with formulas and integrate them with source systems that let you output reports for storytelling. 

“But you always start from scratch. There’s a very generic financial planning capability that only starts to make sense when you connect up the balance sheet and P&L to your forecast models.”

The problem with Excel is that it opens up the option to tell too many stories, he continued. “You lose the single point of truth. If you download data to a local Excel file, you can create, adjust and build your own story explaining why something has happened. You can do what you like and come up with an alternative version of the truth. That’s not what financial storytelling should be.

“Analysts should be using a single source and looking at the data deviations to explain what happens, not pulling data into Excel and coming back with 10 different versions.”

Replies (3)

Please login or register to join the discussion.

avatar
By D V Fields
18th May 2022 20:32

“The problem with Excel is that it opens up the option to tell too many stories, he continued. “You lose the single point of truth. If you download data to a local Excel file, you can create, adjust and build your own story explaining why something has happened. You can do what you like and come up with an alternative version of the truth. That’s not what financial storytelling should be.”

That is not the problem with Excel; it is exactly its strength. It is a tool. Don’t blame your tools.

Thanks (2)
Replying to D V Fields:
avatar
By Hugo Fair
18th May 2022 23:22

Completely agree.
Might just as well say that the problem with a car is that it will take you to wherever you drive it - unlike a tram that only has one route. Not my definition of a weakness.

Thanks (1)
Replying to D V Fields:
Mark Telford Chartered Accountant
By Mark Telford
20th May 2022 14:14

Yep, sounds like 'user error'.

Thanks (0)