Why Your Spreadsheet Stops Being Trustworthy

Almost every early-stage company in Southeast Asia runs its finances on a spreadsheet somebody built in a weekend. That is not a criticism. A spreadsheet is the correct first tool. It is free, it is flexible, and it forces you to think about your own business rather than fill in someone else's template.

The problem is not that spreadsheets break. It is that they do not break. There is no error message the week your model stops describing your company. It keeps opening, keeps calculating, keeps producing a number you keep quoting. The drift is silent, and it usually surfaces in the worst possible room: a board meeting, or the second call with an investor.

Here is how it happens.

The model and the books stop agreeing

Month one, your spreadsheet matches your accounts. Month four, revenue was restated after a credit note, two invoices moved between periods, and a subscription was refunded. None of that flows back into the model, because updating the model is a manual job somebody has to remember.

By month six the model and the books are describing two different companies, and there is no alarm because both look internally consistent. The test is simple and worth running now: take the last closed month, and check your model's revenue and cash against the actual accounts. If they differ by more than a rounding error, every forward number built on top is already off.

Drivers get typed over

This is the most common failure and the hardest to see. Someone needs the numbers to work for a deck, so a formula gets replaced with a typed figure. The cell looks identical. It just no longer responds to anything.

Now your model has a section that does not react to reality. You change the price, and revenue does not move. You add a hire, and costs shift in one place but not the place downstream. The model still calculates, so nothing announces the break. A quick check: pick your three most important assumptions, change each by ten percent, and watch whether the bottom line moves the way you expect. If a change lands nowhere, you have found a hardcoded cell.

It cannot answer a question in the room

The real job of a financial model is not to produce a forecast. It is to answer the question somebody asks you, while they are still in front of you.

What happens to runway if we hire two engineers in October instead of January. What if the enterprise deal closes a quarter late. What if we raise the price by fifteen percent and lose one customer in ten.

A model that can only tell you what you already typed into it fails exactly when you need it. Most founder spreadsheets can produce a number. Very few can produce a different number in under a minute, which is the only speed that matters in a live conversation.

There is no single current version

Model_v4_final_updated.xlsx. Everyone has this folder. There is one file on a laptop, one in a shared drive, one attached to an email an investor still has, and each of them is right about something.

The cost is not the confusion. It is the credibility. When an investor's number does not match the one you say in the meeting, the conversation stops being about your business and starts being about whether your numbers can be trusted at all. That is a hard mood to recover from in diligence.

Currency and timing get flattened

This one is specific to operating here. You bill in MYR, some contracts settle in SGD, a customer in Manila pays in PHP, and your cloud bill is USD. Most founder spreadsheets carry one blended rate typed in once at the top, so a flat revenue month reads as a demand problem when it was a currency move.

Timing is the same story. Booking a sale and receiving the cash are not the same event, and in this region enterprise and government-adjacent buyers can sit sixty to ninety days past invoice as a matter of routine. A model that treats revenue and cash as the same line will tell you that you are fine in a month you cannot make payroll.

What to actually do about it

You do not need to abandon the spreadsheet tomorrow. You need three habits.

Reconcile monthly. Once your books close, tie the model back to them before doing anything else. Fifteen minutes, every month, non negotiable.

Separate inputs from calculations. One clearly marked area holds the things you decided: prices, hire dates, growth assumptions, exchange rates. Everything else is formulas that reference it. If a number is typed into the middle of a calculation, it is a bug.

Model decisions, not just periods. The useful question is never what next year looks like. It is what this hire, this price change, this new market does to you. A model built around the decisions you are actually weighing survives contact with a board meeting. One built around calendar columns does not.

Takeaway

Your spreadsheet will not tell you the day it stops being true. It will keep opening and keep calculating, and you will keep quoting it. Reconcile it to the books every month, keep your assumptions in one place where you can see them, and build it around the decisions you are about to make rather than the months ahead of you. That is the difference between a file that reports the past and a model you can steer with.