Notes on reporting, forecasting and finance data

A month has about twenty working days in Singapore. If the pack arrives on day 19, the operators who could have cut labour, delayed a purchase or chased a receipt have already made those decisions in the dark. The board meeting that follows is a review of a period that cannot be re-run.

Late packs are rarely a software problem. They are a close problem and a contents problem. The close is late because every account is treated as if it needed the same care as a statutory line. The contents are late because the pack is trying to be a complete history of the company rather than a short document for the decisions of the next fortnight. Both habits feel diligent. Together they produce a file that is accurate enough to defend and too late to use.

What a useful pack actually has to do

A management pack has three jobs. It must tell the reader whether the month made money, whether cash is safe for the next thirteen weeks, and which two or three operating lines have moved enough to need a conversation. Everything else is optional. Optional pages are the ones that get added after a single awkward question and then never removed. After a year the pack is forty pages and still late.

The first ten days after month-end are the window in which a figure can still change behaviour. A flash on day 5 or 6, even if two accruals are still estimates, is more valuable than a perfect pack on day 19. The full pack can follow by day 8 to 10, once the control accounts have been tied out. That sequence requires a written materiality line. Without one, the bookkeeper is being asked to finish every recon to the dollar before anyone is allowed to see the month.

A close calendar that people can keep

Calendars fail when they are a wish-list of dates with no owners. A working calendar names the person who extracts the trial balance, the person who signs the bank rec, the person who writes the commentary, and the meeting that will use the pack. It also names the accounts that may be estimated in the flash and must be finished in the full pack. Payroll, stock and revenue cut-off are the usual candidates. Everything else can be held to a materiality figure that the FD is willing to defend.

The first version of the pack should be thinner than the version in people’s heads. The list below is a starting cut that we have used on composite work of this kind:

• Flash profit and loss against last month and against the latest forecast
• Cash at bank, a 13-week summary, and the three receipts that would change the picture
• Working capital: receivables, payables and stock, with the tail older than 60 or 90 days
• Two operating measures the business already understands, such as labour as a percentage of sales or utilisation
• A one-page commentary that explains movement, not a restatement of the table above it

If a director wants a deeper cut, it can be an appendix on a quarterly rhythm. Putting it in the monthly file is how day 10 becomes day 19 again.

Why the late pack persists

People defend a late pack because it feels finished. A day-6 flash feels exposed. The remedy is to say, in the flash itself, which lines are still estimates. Readers will accept an honest gap. They will not accept a silence that lasts three weeks. Once the flash has been sent for two or three months, the full pack usually speeds up as well, because the arguments have already happened.

None of this is a reason to skip reconciliation. A fast pack that cannot be walked to the ledger will be ignored after the first challenge. The sequence is: decide the contents, set materiality, run the calendar, then decorate.

A 13-week cash forecast is a simple idea that is hard to keep honest. You start from the bank, you list the receipts you expect, you list the payments you have to make, and you look at the balance each Friday. The file dies when week one’s actuals arrive and the variance is large enough that nobody wants to open the sheet again.

The usual failure is a receipts method that treats an invoice date as a collection date. In a trading company that sells on 30-day terms, a meaningful share of invoices will clear on 45 or 60 days, and a handful will sit past 90. If the forecast assumes term, it will be chronically early on receipts and then chronically surprised. Payments have a milder version of the same problem: you pay some suppliers early to keep a lane or a product, and you delay others when the week is tight. The contract is not the cash.

Start from the bank, then from ageing

The opening balance must be the cleared bank figure, not the ledger cash line, unless you have just reconciled the two. Unpresented items belong in the first week as a known adjustment, not as a mystery. Receipts then come from the receivables ageing, grouped by how those customers have paid over the last two or three quarters. A customer who is on 30-day paper and pays on 52 days should be forecast on 52 days. That sounds obvious. It is rarely how the first spreadsheet is built.

Known payments go in by name: payroll, rent, the tax instalment, the large supplier run, the loan instalment. Recurring smaller payments can be a weekly average until they matter. The point of naming the large ones is so the Friday call can decide whether to move a date, not whether the model is “broadly fine”.

A forecast that is never marked against the bank is a story. Stories do not pay suppliers.

The weekly ritual is the product

The file is only as good as the meeting that uses it. Thirty minutes, the same day each week, with finance and the person who can delay a payment or make a call to a customer. The agenda is short:

• Last week’s actuals against last week’s forecast, and the three largest differences
• The receipts expected in the next fortnight, and which of them are late already
• The payments that could move without breaking a relationship or a covenant-style limit you already watch
• The lowest balance in the next thirteen weeks, and the week it falls

If the meeting becomes a general operations catch-up, the cash file will be demoted to an appendix and then abandoned. Keep the agenda on the first page of the workbook so a new attendee can follow it.

What “survives contact with reality” means

We look for a forecast that, after three or four weeks of marking, can explain its own errors. Early weeks are allowed to be ugly. What is not allowed is a silent overwrite of last week so the chart looks smooth. A change log of one line per week is enough: “Customer A paid on day 61, not 45; payroll clearing sat an extra day.” After a month the collection curves can be updated with evidence rather than optimism.

Multi-entity and multi-currency groups need a rate policy in the same note as the file. Translating each week on a spot rate that nobody recorded is how a healthy local balance looks like a hole in the consolidated view. Pick a method, write it down, and use it in the pack as well.

The 13-week view does not replace a twelve-month plan. It answers a different question: can we get through the next quarter without a surprise that should have been visible on a Friday.

Ask two competent finance people to show margin by outlet and you can receive two opposite rankings. Both can be arithmetically correct. The difference is almost always the allocation of rent, shared labour, delivery commission, marketing and the central kitchen or warehouse. Until those choices are written down, a conversation about a “loss-making site” is a conversation about a method that nobody has agreed.

The clean first step is contribution: revenue, less the costs that move with the site. Food or goods, site labour, site-level delivery commission, and the utilities that are billed to the unit. That view tells an operator whether the site is earning a keep before anyone discusses the lease. It is also the view that can be updated weekly from the same extracts that feed the pack.

Then write the allocations

Fully loaded margin is still useful. Boards ask for it. Landlords and buyers ask for it. The error is to treat it as a fact of nature. Each allocation is a policy:

• Rent for a shared commissary: by case volume, by revenue, or by a fixed share agreed when the kitchen was opened
• Head-office labour: equally, by revenue, or kept unallocated and shown as a group line
• Marketing: by the campaign’s target sites, or as a group cost if the brand work cannot be split honestly
• Delivery platform fees: to the site that made the sale, not to a central “digital” bucket that hides a channel problem
• Waste and staff meals: to the site, unless a central kitchen is the true source

Any one of those choices can move a mid-ranking site across the break-even line. That is why the file must let you turn an allocation off without breaking the rest of the sheet. Operators should see contribution in their weekly meeting. The board should see contribution and the loaded view, with the method on one page. If the method is hidden in a journal that appears on day 17, nobody will trust either number.

Labour is usually the line that matters

In F&B and in multi-site retail, labour as a percentage of sales is the operating measure that changes behaviour. It has to be timely. A monthly labour figure that arrives with the day-19 pack is a history lesson. A weekly figure, even if overtime is still estimated, is something a manager can roster against. The finance job is to make the definition stable: which allowances sit in the percentage, whether agency staff are included, and how a central trainer’s week is treated when they are on a site.

Illustrative example, composite of typical work: a group with 11 outlets had a consolidated labour ratio that looked acceptable. Site-level contribution, after labour and delivery commission, showed two suburban sites below the line for three consecutive months. The loaded P&L had hidden them because a strong central kitchen allocation had been spread by revenue, which favoured the busy sites and punished the quiet ones twice. Once the rule was written, the conversation moved from “the suburbs are weak” to “this allocation is doing work we did not intend”. That is a better conversation.

What to refuse

Refuse a request to produce a single “true” outlet profit that buries the method. Refuse a dashboard that shows loaded margin in green and red before the allocation note exists. Accept a request for both views, a written policy, and a quarterly revisit when a site is opened or a lease is renegotiated. Margin work is a set of choices. The analytics job is to make those choices visible, comparable and dull enough that the meeting can be about the site.

Dashboards fail in predictable ways. The refresh works. The colours are fine. The figure on the screen does not match the pack, or it cannot be cut the way the meeting needs, or it changes when someone posts a journal to a clearing account that was meant to be temporary in 2026 and is still open. In almost every case the chart of accounts, and the keys around it, were not ready. Building the screen first is how you pay twice.

These six questions are the ones we ask before we agree a dashboard scope. They are also a useful self-test if you are about to buy another licence.

Can revenue be cut the way the business talks?

If the company talks about outlets, channels, products or contracts, those cuts must exist as postings or as reliable keys that join to the ledger. A single “Sales” account plus a memory of what the month contained will not survive a dashboard. The test is simple: can last month’s revenue be split the way the last board paper described it, without opening a side workbook.

Do the control accounts actually control?

Bank, receivables, payables and stock should tie to subledgers. If they do not, every cash and working-capital view will be an argument. Clearing accounts that never clear are a related problem. They become a place where difficult items go to wait, and then they leak into margin.

Can inter-company lines be mirrored?

Multi-entity groups often have a sales line in one company and a cost in another that do not match. Consolidation then depends on a spreadsheet that one person understands. A dashboard on top of that structure will show a group margin that nobody can walk through. Fix the mirrors, or keep the dashboard at entity level until you can.

Is payroll posted in a way that matches how you manage people?

If labour is the operating measure, it cannot live in one account called “Wages”. Department, site or function has to be in the posting or in a join that is stable. Agency and overtime need a home that the definition of “labour ratio” can see. Otherwise the dashboard will show a smooth payroll line that operators do not recognise.

Do customers and products have keys that survive a join?

Profitability work dies on duplicated customer names, products that were re-coded in the stock system, and invoices posted to a miscellaneous debtor. A dashboard will not heal that. A clean-up will. The question is whether the keys are good enough for the first cut you care about, even if the long tail is still messy.

Is there a written map from the chart to the pack?

Someone has to own a mapping file: this account sits in that pack line, this exception is treated like this, this currency is translated like that. If the map exists only in a working paper from last year, the dashboard and the pack will drift within two closes. The map is part of the numbers layer. It should be versioned when the chart changes.

A practical sequence follows from the answers:

• If more than two of the six are “no”, commission clean-up and a pack before any dashboard.
• If the cuts exist and the controls tie, a small dashboard of the pack views can be built in the same engagement.
• If only the keys are weak, build the pack from the ledger and keep profitability in a controlled file until the keys are repaired.

None of this is an argument against good visualisation. It is an argument for sequence. The chart is the language the company uses to talk to itself. Teach that language first. Then put it on a screen.