Large organisations have spent decades improving their financial systems.
ERP platforms record transactions with extraordinary precision. Treasury Management Systems provide sophisticated tools for managing cash, funding and financial risk. Banks provide increasingly capable APIs and real-time information. Business Intelligence platforms can analyse enormous quantities of data.
And yet, in many organisations, answering a very basic question remains surprisingly difficult:
What money do we actually expect to come in and go out over the next few days and weeks?
For something so fundamental to Treasury, it is remarkable how often the answer still involves spreadsheets, extracts, emails and manual intervention.
Why?
I don't think the answer is that Treasury needs another forecasting model, I think there is a more fundamental problem.
Before we can forecast the future, we need to make better use of what the organisation already knows.
Let's start with what we know
Consider the most immediate Treasury horizon.
Today, tomorrow, next week.
The organisation's accounting systems already contain substantial information about expected cash movements.
Accounts Payable contains money that the organisation expects to pay.
Accounts Receivable contains money that it expects to receive.
There may be thousands, or millions, of individual transactions distributed across multiple legal entities, ERP systems, currencies and countries.
This isn't speculative information.
- An invoice exists.
- An amount is known.
- A currency is known.
- A counterparty is known.
- A due date exists.
Before attempting sophisticated prediction, there is an obvious question:
Can Treasury see all of that clearly?
Surprisingly often, the answer is no.
The information exists, but not where Treasury needs it
A multinational organisation may operate several ERP systems.
Individual subsidiaries may run different accounting platforms.
Some businesses may have been acquired and never fully integrated.
Different systems may represent currencies, entities, suppliers and payment dates differently.
Treasury therefore faces an odd situation.
The organisation knows that €3.2 million is due to suppliers next Thursday.
It knows that $4.7 million of customer invoices are due for collection.
But the people responsible for ensuring the organisation has sufficient liquidity and managing the associated FX positions may not have a simple consolidated view of either.
The information exists. Visibility of that information doesn't.
This matters for more than cash
Expected payables and receivables aren't merely accounting records.
They are potential cash movements.
And once currencies are involved, they are potential foreign exchange exposures.
Imagine that over the next fourteen days an organisation expects:
- £8 million of Sterling receipts;
- €5 million of Euro payments;
- $4 million of Dollar receipts;
- CHF 1 million of supplier payments.
Treasury needs to understand not only the overall liquidity position, but the position by country, bank, currency, entity and date.
Does a subsidiary require funding?
Can expected receipts cover upcoming payments?
How confident are we that expected receipts will be paid on time?
Is there a natural currency hedge?
Does Treasury need to buy Euros?
When? How much? Where does it need to be?
These decisions become much easier when expected operational cash flows are visible together.
Due date is not the same as cash date
Even this apparently simple picture contains an important complication.
Accounting systems generally tell us when something is due.
Treasury cares about when it will settle.
Those aren't necessarily the same thing.
A supplier invoice due on Friday might be included in Tuesday's payment run.
A customer may routinely pay twelve days late.
Another customer may consistently pay three days early.
An invoice may be disputed.
A payment may be split into instalments.
One receipt may settle several invoices.
The accounting record is correct, but its due date isn't necessarily the best estimate of the resulting cash event.
This is where the problem begins to move from financial data towards Financial Intelligence.
Don't start by trying to predict everything
There is a temptation when discussing forecasting to jump immediately to Artificial Intelligence, predictive models and sophisticated scenario analysis.
I think that's backwards, the first objective should be much simpler.
Give Treasury the best possible view of the financial events the organisation already knows about.
Bring together Accounts Payable and Accounts Receivable.
Normalise them across systems, entities and currencies.
Present them in a way that enables Treasury to understand upcoming liquidity requirements and FX exposures.
That alone can be extremely valuable and importantly, it doesn't require replacing the organisation's ERP, accounting or Treasury systems.
It simply makes their information more useful.
Then start learning
Once expected transactions can be observed alongside what actually happens at the bank, a much more interesting opportunity appears.
We can compare expectation with reality.
An invoice was due on the 15th, when did the cash actually arrive?
A supplier payment was expected on the 20th, when did it settle?
Do particular customers consistently pay late?
Do particular entities process supplier payments on predictable days?
Does behaviour vary by currency, invoice size or business unit?
Over time, the difference between due date and likely settlement date becomes something we can understand rather than simply guess and that understanding can improve continuously.
Then look further upstream
Accounts Payable and Accounts Receivable are not the beginning of the financial story, they're simply a very useful place to start.
Before the invoice there may have been an order.
Before the order, a contract.
Before the contract, a sales opportunity.
On the expenditure side there may have been a purchase request, procurement process or contractual commitment.
Each stage potentially provides earlier information about a future cash event, but earlier information comes with greater uncertainty.
A receivable recorded in the ERP may provide strong evidence of a future receipt.
A signed contract may provide earlier evidence.
A sales opportunity may provide earlier evidence still, but with considerably less certainty.
That suggests a progression, not a giant transformation project, a progression.
Start with the information closest to cash and make it useful.
Then, where there is business value, move progressively further upstream.
From forecasting to understanding
This changes the nature of cash forecasting. Instead of periodically asking departments: "What do you think will happen?"
We can increasingly ask: "What does the evidence we already possess tell us is likely to happen?"
Those are very different approaches. The first produces a forecast. The second develops understanding.
And that understanding can change continuously as new evidence appears.
An invoice is raised.
A delivery slips.
A customer disputes a charge.
A payment arrives.
Each event changes what the organisation knows and a financial view of the future should be capable of changing with it.
A practical beginning
There is a much larger opportunity here.
Eventually, financially significant information could be connected across Sales, Procurement, Legal, Operations, Finance, Treasury and Banking.
But organisations don't need to solve that entire problem before obtaining value.
The starting point can be remarkably practical:
Show Treasury the payables and receivables already contained within the organisation's accounting systems, consolidated by when, where and in which country, bank and currency cash is expected to move.
That provides immediate information for liquidity and FX management.
Then improve the expected settlement dates as evidence accumulates.
Then add other information sources where they provide sufficient additional value.
Each stage improves understanding without making the previous stage obsolete.
Perhaps that's the real reason cash forecasting has remained difficult, we've been treating it as a problem of predicting the future.
The first challenge is simpler.
We need to see the future the organisation already knows about.
And only then should we start predicting the part it doesn't.
