eProcureAI / Platform / For FP and A

Role

Forecasting on invoices
is forecasting on history

Committed spend is known weeks before it invoices. Building a forecast without it means rediscovering decisions that were already made and could have been planned around.

Written for the people building the model. Committed spend is the input you are missing.

Forecast inputsWhat you have today
The situation

The decisions have been made. You just cannot see them yet.

Financial planning built on invoiced actuals is planning around decisions that were taken weeks ago, by people who did not think to mention them.

A department approves a purchase in March. The order goes out, the goods arrive in April, and the invoice posts in May. Your forecast learns about it in May, at which point it is not a forecast at all, it is a reconciliation.

The information existed in March. It simply was not carried anywhere you could see it, because the budget only moved when an invoice arrived to force it.

Commitment is the missing input

Approved orders that have not yet invoiced represent decisions the business has already made. Including them in a forecast converts a chunk of your model from estimate to fact.

It also changes the conversation with budget holders. Instead of explaining a variance after the event, you are discussing a position they can still influence.

And variance stops being archaeology

When a line moves unexpectedly, the useful question is which transactions caused it and who approved them. Drilling from a variance to the underlying purchases with approver and reason attached turns a week of investigation into a few minutes.

The forecasting gap

Five points where information exists but does not reach you

Point 1

The request is raised

Intent, not commitment. Reasonably invisible to a forecast at this stage.

IncludeNot yet
1
Point 2

It is approved

The business has committed. This is the moment a forecast should know, and usually the moment it does not.

IncludeYes
2
Point 3

The order goes out

External commitment, difficult to unwind. Still absent from invoiced reporting.

IncludeYes
3
Point 4

Goods arrive

Received but not invoiced. Relevant to both accruals and timing.

IncludeYes
4
Point 5

The invoice posts

Where most forecasting inputs begin, which is several weeks too late to be useful.

IncludeAlready too late
5
What changes

Three inputs you did not previously have

Committed spend as a forecast input

Approved orders with expected timing, available at the moment of approval rather than when the invoice arrives.

  • Committed value by charge code and department
  • Expected timing informed by lead times
  • Available continuously rather than as a monthly extract
  • Converts estimate into fact for a chunk of the model
Committed viewLive
Approved not invoicedMaterial
By departmentBroken down
Expected timingFrom lead times
AvailableContinuously
A decision already madeRather than a guess

Requests in flight as a leading indicator

Requests raised but not yet approved are the earliest signal you can get. Not commitment, but a genuine indication of where demand is building.

  • Pipeline of requests by department and category
  • Earlier signal than any committed number
  • Useful for pressure rather than precision
  • Shows where a budget is about to come under strain
PipelineLive
Requests in flightVisible
By departmentBroken down
StatusPending approval
SignalWhere pressure is building
Earliest indicatorDirectional rather than precise

Variance you can explain in minutes

Drill from a moving line to the transactions behind it, each carrying who approved it and why, so the explanation is a lookup rather than an investigation.

  • Drill from summary to transaction
  • Approver and reason on every decision
  • Category and supplier breakdowns underneath
  • Explanation available before the meeting
Variance drillLive
LineMoved
TransactionsListed
ApproverNamed
ReasonRecorded
Time to explainMinutes
TraceableRather than reconstructed
Three sources of signal

Each one arrives at a different moment

Using all three gives you a picture that starts earlier and firms up as decisions are made.

SignalWhen it appearsConfidenceBest used for
Requests in flightBefore approvalDirectionalSpotting pressure building on a budget
Committed spendAt approvalHighForecasting the next few periods with facts
Goods received not invoicedAt deliveryCertainAccruals and period cut off
Invoiced actualsWeeks laterCertain but lateReporting rather than forecasting

Most planning functions only have the last of these, which is why forecasts feel like they are always catching up.

Explaining variance

Six questions that become quick to answer

What moved

Which line and by how much

Drill from the summary rather than starting from a general ledger extract.

PathSummary to transaction
Who approved it

The named person

On every decision, with the reason they gave at the time.

RecordedAt the decision
Was it in budget

Or an exception

Exceptions are recorded as exceptions, so the answer is on the record.

VisibleSeparately
Is it recurring

Or a one off

Category trend across periods rather than a single month in isolation.

ComparedAcross periods
Is more coming

Committed but not invoiced

The part that turns an explanation into a forecast adjustment.

IncludedCommitted
Which supplier

And at what price

Supplier and price adherence underneath the category number.

UnderneathEvery figure
Who does what

The short version of your inputs

Better forecasting here is not a modelling improvement, it is an information timing improvement.

What you do

Build the model and explain the movement. Both get faster with better inputs.

The human partLive
Build the forecastWith committed included
Explain varianceFrom the record
Model scenariosOn current numbers
Advise budget holdersWhile it matters
AnalysisNot data gathering

What eProcureAI does

Supplies the inputs at the moment they exist rather than when they invoice.

The automatic partLive
Record commitmentAt approval
Track expected timingFrom lead times
Surface requests in flightBefore approval
Attach approver and reasonTo every decision
Build accrualsFrom receipts and commitments
CurrentRather than a monthly extract
0committed spend available as a forecast input
0requests in flight visible as an early indicator
0to explain a variance from the transactions behind it
0current rather than a monthly extract
FAQ

Questions people actually ask

Why does committed spend matter for forecasting?
Because it represents decisions the business has already taken. Approved orders that have not invoiced yet are facts, and including them converts part of your model from estimate to certainty.
How is expected timing worked out?
From the order and its lead time where one is recorded, and from historical patterns for the supplier and category where it is not.
Can we see requests before they are approved?
Yes, as a pipeline. It is directional rather than precise, but it is the earliest signal available and it shows where a budget is about to come under pressure.
How does variance analysis work?
Drill from a summary figure to the transactions underneath, each carrying who approved it and the reason they gave. That turns an explanation from a week of investigation into a few minutes.
Does this replace our planning tool?
No. It supplies procurement inputs earlier and in more detail. Most teams continue modelling where they already do and export from here into it.
Can budget holders see their own position?
Yes, scoped to their own charge codes, which tends to reduce the number of questions that reach you.
How does this help with accruals?
Goods received but not invoiced and open commitments are both recorded, so accruals are calculated rather than assembled at month end.
How much history do we need before it is useful?
Committed spend is useful immediately. Category trend analysis and supplier patterns need a quarter or two of clean data behind them.

Bring a forecast that missed

We will show what the committed position looked like at the time and what you would have seen.

Book your free demo

Related: All solutions and Spend Analytics