How the numbers work

Average daily usage

What the container gave up over the last 14 days, divided by the time actually observed in that window — not by the calendar.

Time after the last reading is time nobody measured, so it stays out of the divisor. So does the stretch between two readings where a delivery landed: the level rose, so what was used across it cannot be read, and counting it would turn a truck into a day of using nothing. A gap in logging leaves the rate alone rather than quietly bending it toward zero — an unwatched chemical must never look safer than a watched one. How out of date the reading is shows separately, as a stale reading.

The level the forecast uses

Where the container is likely to be now: the last measured level, drawn down at the current rate for the hours since it was measured.

The level the screen shows stays the measured one — inventing a displayed number nobody read would be worse than showing a real one clearly marked stale. Only the forecast projects forward, and it never projects below empty.

Unusual usage

The last 3 days' rate against the 14 days ending 3 days ago, flagged when the two differ by more than 35%.

It needs roughly 17 days of history before it will say anything at all. It reports a change in pace, not a shortage — how much is left is the status color’s job, and the two move independently.

Stale reading

A chemical is marked low-confidence once its last reading is more than 36 hours old.

The forecast keeps running on the rate it already has. The flag is about how much weight to put on it, not a claim that the number is wrong.

Days until empty

The projected level now, divided by average daily usage.

It assumes today’s rate holds. It is a runway, not a promise — a process change moves it the moment the next reading lands.

Reorder point

How much runway you want still in hand when the order goes out: supplier lead time plus safety stock, unless a number of days has been set directly.

Days until reorder, and the order-by date

Days until the projected level reaches the order-by level set on the gauge — or, where no such level is set, days until empty less the reorder point.

A physical order-by level wins wherever one is set: it predicts the threshold actually marked on the gauge rather than an abstract day count. The order-by date then works back from the delivery day, so it can be earlier than the raw day count implies.

Next delivery date

The first day a truck can actually come, on or after the date the chemical needs it.

With no delivery days configured, every day counts as available. An order-by date already in the past is shown as today rather than backdated — it needs placing now, not explaining away.

Green, yellow and red

Red once the reorder point has been reached or passed. Yellow inside the last half of it. Green everywhere before that.

The band scales with the reorder point, so a chemical on a three-week lead time gets proportionally more warning than one that arrives next day. Red does not mean empty — it means the order can no longer safely wait.

Recommended order quantity

Your standard load, capped at the space actually left in the tank.

The cap is the point: no delivery can be proposed that would overflow the tank. With no load size on file the headroom is the recommendation. The figure is rounded to a sensible ordering quantity before it is shown.

Orders projected to year end

The container is run forward at its current rate to the end of your budget year, placing an order each time it reaches its reorder point. A cylinder rack is run forward one swap at a time, the same way its spare trigger is, so the spares on each row are whole cylinders. Where a year of past usage is on file, the rows past the next month follow last year’s pattern instead of today’s rate. An order already placed is counted on its delivery date instead of a projected one. One placed as a full load counts as your full load size as it is set now, so correcting that figure corrects the orders already out.

Where there is not enough on file to run the container forward, it falls back to how often deliveries have actually been arriving. Either way the dates move as usage does — a hot week pulls every one of them in.

Duty, standby and empty

A position reads empty at or below 5% of a full cylinder. Above that it is on duty if it has given anything up in the last 3 days, and on standby if it has not.

Empty is checked first: a bottle that has just run out has both a near-zero weight and recent consumption, and it is the weight that tells you what to do about it. 3 days rather than one interval, so a quiet night on a plant that batch-doses does not flip a working cylinder to standby and back.

Days until changeover

When the cylinder now on duty reaches empty, from its own weigh-ins and the rate it is being drawn at.

Each position is forecast from its own history, so a rack part-way through a bottle does not get the average of the whole rack applied to it.

Days until the spares run low

How long until the spare stack falls to the number you said to order at, counted one swap at a time.

Swaps that the weigh-ins can date are used as measured; past those the rack is running on bottles that arrived full, so the remainder is spaced at one cylinder’s worth of gas each. Spares are a single shared stack — the reorder math deliberately does not split them by what they are for.

Spend so far

What the deliveries logged inside the current budget year cost, plus anything recorded as spent before tracking began.

It counts deliveries, not orders — a load ordered and not yet arrived is not spend. That is what keeps this figure reconcilable against invoices.

Expected by now

The annual budget spread evenly across the budget year and taken up to today.

A straight line, deliberately. It describes the past and nothing else, and a chemical whose deliveries cluster early will read over it while still finishing the year comfortably under.

Projected year end

Spend so far, plus what the orders still expected this year are likely to cost.

Prices come from your own invoices, weighted toward recent ones so a price rise shows up rather than being averaged away against last year. Where a chemical has no delivery history to price from, the figure is spend so far and is labeled as such — it then answers "have you already gone over", not "are you heading over".

Following last year’s pattern

Past the next month, usage follows how last year moved from this time of year on, at this year’s level, so a busy summer and a quiet winter are both in the projection.

The next month always runs at today’s rate, and last year eases in over the month after that. Each past delivery is spread across the days since the one before it, because a truck replaces what went out since the last one. Where too little of last year is known, the projection stays at today’s usage. When the two year-end figures differ by 10% or more, both are shown.

The color on a budget

It keys on the projected year end, not the pace: green under budget, yellow in the narrow band just over, red clearly over.

This is why a chemical can be running ahead of pace and still be green. Pace is a description of the past; the color is a claim about the end of the year, and the end of the year is the question a budget actually asks.