Insights

A new template will not fix a pack that arrives on the fifteenth of the following month. The board reads late numbers as provisional, even when the arithmetic is correct. Operators have already made the month’s decisions from bank balances and from whatever spreadsheet was nearest to hand. The pack then becomes a historical document. Beautyandhealthco treats the close calendar as the first artefact of an engagement, ahead of page design.

A usable calendar names two working days. Freeze day is when source files stop being replaced. Issue day is when the PDF is sent. We set freeze at T+2 and issue at T+4 for a single-entity close with stable extracts. Those stamps only hold if every source has an owner and a file-name pattern. If the POS export lands when the store manager remembers, T+4 is theatre. The source register is therefore part of the calendar, not an appendix.

Late files need a written treatment. The default we use is: the pack is issued with last period’s operational file and a flagged exception; the following pack restates that line. Restatement is visible in the appendix. Quietly inserting a late file after issue, and then arguing about which PDF is current, is how version control dies. The calendar also needs a reviewer who did not extract the files. A same-day self-review is how mapping errors survive into the board pack.

Templates still matter, but they matter after the clock exists. A locked page order helps a director compare months. A source note under a table helps an analyst rete the figure. Neither of those features will be used if the pack misses the meeting. When a diagnostic review finds both a weak template and a missing calendar, we sequence the work as calendar, then definitions, then template. Switching that order produces a handsome PDF that still cannot be defended on a stated day.

Public holidays in Singapore are treated as non-working days in the count. If T+4 falls on a holiday, issue moves to the next working day and the freeze moves with it unless the statement of work says otherwise. The rule is dull on purpose. Dull rules survive questions. Decorative dashboards do not.

Most operating companies that ask for a cash forecast already have a spreadsheet with weeks across the top. The usual failure is not the grid. It is the way receipts are entered as hoped collections, and payments as a smoothed average, with no tie to the bank or to the ledger. Beautyandhealthco builds the 13-week view from three imperfect sources and writes down where each source is weak.

The first source is the bank. Opening cash is the cleared balance on the last file date, plus a short list of unpresented items. If unpresented items cannot be listed, we still open from the bank and we label the gap. Inventing a “true cash” figure that no statement supports is how a model loses the room. The second source is the receivables ageing, multiplied by a collection-rate assumption on the register. Ageing is imperfect. Credit notes sit in the wrong bucket. Intercompany balances clutter the file. We still use it, and we keep a reconciling line for amounts we cannot place.

The third source is a payment calendar: payroll, CPF, tax, rent, known supplier runs, debt service. These dates are more reliable than receipt dates. They should be hard-coded to the week they fall, not spread. Smoothing a tax instalment across four weeks makes the cash low look milder than it is. Where supplier terms are informal, we use the last thirteen weeks of actual payment-run dates as the assumption and we say so.

Refresh day is as important as the formulas. A weekly overwrite of column one from the bank file is the control that keeps the view honest. Without it, the model becomes a plan that nobody compares with actuals. We prefer Monday 12:00 SGT using Friday’s bank file. If Friday’s file is late, last week’s actuals stay and the exception list records it. Forecasts remain scenarios. A collection rate of 85 percent of current-plus-30 is an assumption, not a prediction. When the rate is wrong, the cash view will be wrong, and the change log should show when the rate was last evidenced from actual collections.

Imperfect data is a reason to show the gaps. Empty cells with a note are preferable to filled cells with no source. Boards can decide when the gap is labelled.

Ask three people in the same company for contribution by channel and you can receive three answers that all add up. One person deducts only merchant fees. Another deducts merchant fees, shipping and paid media. A third deducts a share of warehouse rent as well. Each answer is internally consistent. None of them can be compared with last month unless the cost map is signed. Beautyandhealthco treats the map as the product. The table of channel margins is a consequence.

The map has three columns: direct, allocated, excluded. Direct costs move with the channel in a way the files can evidence. Marketplace commission, shipping on that channel, and payment fees usually belong here. Allocated costs are real costs that we spread on a stated key (orders, hours, floor space). Warehouse labour often sits here. Excluded costs stay in a central bucket: brand marketing that is not channel-tagged, head-office rent, one-offs. Putting brand spend into a channel because “it feels related” will swing the margin every time the campaign calendar moves.

Paid acquisition is the line that causes the most argument. If CAC is treated as a direct channel cost, contribution after ads will look severe on the paid channel and generous on organic. That may be the question the board wants. If CAC is kept in a growth bucket, the board is asking whether fulfilment economics work once demand exists. Both views can be produced. They cannot share a single cell labelled “contribution” without a footnote. We publish the definition in the register and we name the view on the pack page.

Returns and discounts belong on the walk from list or GMV to net revenue, before contribution is calculated. Burying returns in a cost line hides price realisation. Net revenue in the channel table must equal net revenue on the P&L. If it does not, the channel page is a separate model, and we treat that as a defect.

When a channel lacks a reliable cost feed, we leave the cell empty. The pack shows the gap. A guessed warehouse allocation that changes method every quarter will produce a story about improving contribution that is only a change of method. Method changes belong in the change log, with the first pack version that uses the new rule.

Commentary is often written as atmosphere. Sales were “resilient”. Costs were “well managed”. The board cannot act on either sentence. Beautyandhealthco writes commentary after the numbers freeze, to a three-question standard, in full sentences. The standard is part of the operating manual so a new analyst can follow it on T+3 without inventing a voice.

Question one: what moved. Name the line, the amount against budget and against prior year, and the driver. Volume, rate, mix, timing and one-off are the allowed driver labels. If two drivers moved, say so and split the amount if the files support a split. If they do not, say the split is unavailable. “Gross margin down” without a driver is a caption, not commentary.

Question two: whether it persists. Timing differences (a shipment delayed across month-end, a billing run pulled forward) are labelled timing and are expected to reverse. Run-rate changes (a price list, a roster, a supplier rate) are labelled run-rate. One-offs sit on the one-off schedule and are named. Mixing these three into a single adjective is how a board overreacts to a timing item or underreacts to a run-rate item.

Question three: what decision it informs. The list is short: price, hiring, collection, inventory, supplier terms, or no decision this month. “No decision this month” is a valid close. It stops the pack from manufacturing urgency. What we do not write is a motivational summary, a promise of recovery, or a percentage improvement that is not in the table above the paragraph.

Length is a control. Each material line gets three to six sentences. If twenty lines moved, the pack does not grow a twenty-page letter. Materiality is defined in the statement of work, usually as a dollar threshold and a percentage of budget. Below that threshold, the line can sit in the table without prose. Boards read faster when the commentary is scarce and specific. Analysts write faster when the questions do not change from month to month.