Logiciel Contact Us
Success Stories Tech News Contact Us
whitepaper

Why A 6.4-Day Close And 9% Forecast Error Are Not ERP Problems.

Twenty years of system upgrades bought two days off the close and changed almost nothing about the work inside those days. This report sets out the five things actually wrong, the line between what a machine drafts and what a human signs, and a 90-day sequence that starts with your chart of accounts.

In depth

The Close Got Faster. The Work Got No Better.

01

The trap most finance functions walked into: buy the next system, take two days off the calendar, and leave the reconciliations, accruals and flux commentary exactly where they were, so 70% of period-end effort still goes on mechanics and the controller analyses last, at nine in the evening, tired.

02

What the four-day functions do instead: time the current close task by task, force one group chart of accounts with a named owner who can reject local exceptions, fix the broken process before automating it, then start on nightly reconciliation rather than the accrual model.

The detail

What Separates A Four-Day Close.

Zone · 01

One Group Chart Of Accounts

The spread between a four-day close and a ten-day close sits inside functions running comparable software, which makes the variable process discipline and master data. Entity-level variations in account structure and intercompany mapping are the single most common reason a working pilot in one ledger fails across four.

Zone · 02

A Written Machine-Human Split

Every finance process worth automating divides into a machine-produced draft and a human judgement that carries the accountability. The machine runs nightly matching and posts clearing entries. A named human signs dispositions and write-offs. Write that split down per process, with the evidence an auditor will accept, before anything is configured.

Zone · 03

Accuracy Scored By Line And Horizon

A single blended error figure hides everything, because netting a revenue overshoot against a cost overshoot reads as precision. Score weighted absolute percentage error by line item and by horizon, then backtest against frozen vintages: what the forecast said the day it was made, against the actual as first reported.

By the numbers

The figures that make it a board-level conversation.

70%
of period-end effort goes on reconciliation, journal preparation and data gathering rather than analysis
9%
median absolute error on next-quarter revenue forecasts. Boards commit capital as though the number were accurate to two or three percent
6.4
working days is the median monthly close. Top-quartile functions land near four, bottom-quartile ones take ten or more, and 44% still need more than five
Inside the report

What you'll take away.

01

Time the close task by task

Across two periods, record who did each task, how long it took and what they waited for. Most functions discover that two thirds of the elapsed time is waiting rather than processing, which is not a software problem.

02

Force one chart of accounts and one owner

A named master data owner with authority to reject local exceptions. A recurring manual journal usually signals a mapping error or a stale policy, so fix the broken process first. Automating a workaround makes it permanent.

03

Start with nightly reconciliation, not the accrual model

Matching is mechanical, easy to evidence, and it surfaces the data quality problems that would otherwise poison the forecast work later. A mismatch surfaces on the eleventh, while the person who caused it still remembers the transaction.

Questions

Frequently asked.

Do we need to replace or upgrade our ERP before any of this works?

No. The spread between a four-day and a ten-day close sits inside functions running comparable systems, so the variable is process discipline and master
data. Reconciliation, anomaly scoring and accrual estimation all read from the ledger you already have. An upgrade programme mostly delays the work that produces the days.

Will our external auditor accept a journal entry that a model drafted?

Auditors accept it when four things exist per entry: a reproducible derivation that regenerates the figure from the same inputs, an immutable snapshot of
those inputs, a named approver, and a documented control with a stated frequency and an evidenced exception process. Under SOX the accrual is a management estimate.

Can a model actually forecast better than our FP&A team can?

On some series, reliably. Collections timing,transaction volumes, unit demand, churn and days sales outstanding have hundreds of observations and stable
drivers. On new market entry or an unprecedented pricing change it cannot, and there the honest output is a few scenarios with named assumptions. Eighteen monthly points produce a confident line and no information.

We already run quarterly reforecasts, so is a rolling forecast worth it?

The value is not frequency. It is removing the annual number that every reforecast negotiates against. An annual budget locks assumptions in October and asks the business to defend them for fifteen months, which is why bias is directional and stable. A rolling six to eight quarter view refreshed monthly on drivers takes away the reason to bias the estimate.

How much of the close can we realistically automate in one quarter?

One process end to end, evidenced. Typically nightly sub-ledger matching for the entities already on a common chart of accounts, with the control
walked through by internal audit. Accruals and flux commentary follow once the reconciliation data proves clean. Attempting three processes at once produces three half-built ones.

Who is this report for?

CFOs and group controllers who have already been through an ERP programme and still close in six days or more. It assumes you own the forecast that funds capital allocation, you suspect it is wrong by roughly a tenth, and nobody grades it honestly.

Get the whitepaper

Have it emailed to you.

Drop your details and we'll send Why A 6.4-Day Close And 9% Forecast Error Are Not ERP Problems straight to your inbox - no spam, unsubscribe anytime.

Download whitepaper
Next step

Prove it on one reconciliation.

Bring one reconciliation and two closes of task-level timing to a working session with our engineering leads. We will tell you where the days actually are.

Book a close diagnostic