Settlement timing and your cash forecast

A store's cash position is not a function of its sales. It is a function of when each sale settles, and a Malaysian store selling through several payment methods and channels has sales settling on four or five different rhythms at once.

Most cash forecasts collapse that into a single average lag. It is a reasonable simplification in a steady month and a misleading one around every event that matters.

The clocks you are actually running

Cards on your gateway's cycle, typically the most predictable.

FPX on bank rails, so weekends, public holidays and cut-off times intrude — see FPX settlement and what it does to your cash.

E-wallets on their own cycle, sometimes through the gateway and sometimes direct — see e-wallet payments and how they settle.

BNPL on the provider's cycle.

Marketplace payouts, if you also sell there, on schedules unrelated to any of the above and with their own escrow rules.

Less whatever is withheld as a rolling reserve, released later — see payment gateway holds and reserves.

Six streams is normal. An average lag across six streams describes none of them.

Where the average breaks

Public holidays. Malaysia's calendar includes state-varying holidays, so a lag that holds in one week does not in another. Bank-rail settlement is affected more than card settlement, which means the mix shifts the effect.

Weekends before a long break. A Friday sale ahead of a holiday Monday can wait significantly longer than the same sale on a Tuesday.

Sales spikes. The worst case, because three things happen together: volume rises, reserve withholding rises with it, and risk-triggered holds become more likely. Your best trading week has the slowest and most withheld conversion to cash.

That combination is what turns a successful promotion into a cash squeeze, and it surprises merchants every time because the sales figure looks wonderful.

Growth generally. A rolling reserve is a percentage of revenue, so a growing business has a growing amount of its own money held. Growth consumes working capital before it produces it.

Building a forecast that survives contact

Forecast by settlement stream, not in aggregate. Each stream has its own lag and its own calendar sensitivity. Six simple forecasts beat one sophisticated average.

Use trading-day lags, not calendar days, for anything on bank rails. Then apply the actual holiday calendar rather than assuming five working days a week.

Model the reserve explicitly. As a percentage of revenue withheld and released on a schedule, so a growth month shows the working capital it will absorb.

Forecast promotions separately. Do not apply your normal lag to abnormal volume. Model the spike with its own slower conversion.

Reconcile the forecast against what happened. The gap between forecast and actual settlement, measured for a couple of months, tells you which lags you have wrong. This is the step that turns a guess into a model, and it is the one nearly always skipped — see measuring whether AI accounting worked.

What automation contributes here

Not prediction. The value is currency and granularity.

A forecast built on reconciled data three days old, with each stream's actual observed lag, is genuinely useful. One built on a spreadsheet updated last month with a single assumed lag is decoration — see AI cash flow forecasting for businesses.

What no system can do is tell you when a gateway will change its cycle, when a reserve will be increased, or how a public holiday will affect a specific provider's processing. Those are questions for your provider, and the answers belong in your model as inputs rather than being inferred from history.

The one number worth having on hand

Money earned but not yet available. Settled sales awaiting payout, plus amounts held in reserve, plus marketplace balances in escrow.

For most growing ecommerce businesses this figure is considerably larger than expected, and it explains the recurring puzzlement of a profitable month with an uncomfortable bank balance. It is not a problem to be solved so much as a structural feature to be planned around, and knowing its size is the difference between planning and being surprised.

Common questions

Why is one average settlement lag not enough for a cash forecast?

Because a Malaysian store typically runs several settlement streams at once — cards, FPX, e-wallets, buy now pay later and marketplace payouts — each on its own cycle with different sensitivity to weekends and public holidays. An average describes none of them accurately, and diverges most around exactly the events where cash accuracy matters.

Why does a successful promotion cause a cash squeeze?

Because three effects coincide. Volume rises, rolling reserve withholding rises proportionally with it, and risk-triggered holds become more likely on unusual transaction patterns. The strongest trading week therefore converts to cash both more slowly and less completely than a normal one, while the sales figure looks excellent.

How should public holidays be handled in a settlement forecast?

By using trading-day lags rather than calendar days for anything settling on bank rails, then applying the actual holiday calendar including state-specific holidays. Bank-rail methods such as FPX are affected more than card settlement, so the impact depends on your payment mix rather than being uniform.

What single figure best explains a profitable month with poor cash?

Money earned but not yet available — settled sales awaiting payout, plus amounts held in reserve, plus any marketplace balances still in escrow. For most growing ecommerce businesses this total is larger than expected, and knowing its size turns a recurring surprise into something that can be planned around.

Related: AI cash flow forecasting for businesses · payment gateway holds and reserves · managing cash flow across multiple channels