In short: With a discount d applied to a committed quantity Q and shortfall units paid at full contract price, the commitment only beats buying uncommitted if you consume more than (1 - d)Q, so a 12 per cent discount on 40,000 tonnes needs 35,200 tonnes to break even. Run the expected cost of each discount tier against your demand distribution using the loss function and the top tier frequently loses to a smaller one, and sometimes to no commitment at all. Two clauses move the answer more than the headline percentage: the deficiency payment rate, which in the worked example pulls break-even from 35,200 tonnes down to 27,500, and the measurement period, where switching an annual commitment to monthly buckets multiplies expected shortfall payments several times over on identical volume.
A supplier offers 12 per cent off if you commit to 40,000 tonnes for the year. The base plan says 34,000. Procurement books the saving into the budget, the contract gets signed in January, and by the end of September somebody in planning is being asked to find 4,000 tonnes of consumption from somewhere so the shortfall payment does not land in Q4. What gets bought is real material that sits in a tank until March.
That sequence is common enough to be boring, and the arithmetic that would have prevented it takes about ten minutes. Volume commitments are often worth signing. The error is sizing one against the base forecast when the contract's economics are asymmetric around it.
What a take or pay clause actually is
A take or pay obligation says you will pay for a quantity in a period whether or not you take delivery of it. The quantity puts a floor under your spend, whatever your consumption does. Four instruments sit near each other and behave differently, and contracts often combine them without saying so.
Straight take or pay. You commit to Q per period and pay the contract price for whatever you do not take. Common in gas, industrial gases, bulk chemicals and packaging.
Deficiency payment. The same structure with the shortfall priced at a fraction of the contract price, often somewhere between a fifth and a half. This looks like a softer version of the same thing and it changes the economics far more than the fraction suggests.
Retrospective rebate with clawback. You buy freely, and if you miss the annual band the rebate is recalculated and invoiced back. Economically similar to take or pay, though it usually lands as a one line credit note nobody in operations sees coming. Rebate mechanics have their own post.
Capacity reservation. You pay a fee for the right to call off volume up to a ceiling, with no obligation to take it. That is an option, and it gets priced as one.
The distinction that matters for planning is whether the payment is triggered by your consumption or by the calendar. Anything triggered by the calendar becomes a supply plan constraint the moment it is signed.
The break-even is below the commitment, and you can calculate it exactly
Let p be the price you would pay with no commitment, d the discount you get for committing, and Q the committed quantity. Take the strict case first, where shortfall units are paid at the contract price.
Your cost with the commitment is p(1 - d) times the larger of demand and Q. Your cost without it is p times demand. Set them equal and the price cancels out, leaving a break-even consumption of (1 - d)Q.
That result is worth sitting with, because it does not depend on the price at all. A 12 per cent discount on 40,000 tonnes breaks even at 35,200 tonnes. Below that figure, the deal you signed to save money is costing you money. And the base plan in the opening paragraph was 34,000.
Now put a distribution around it, because the point forecast is the wrong input. Say demand for the year is roughly normal with a mean of 34,000 and a standard deviation of 5,000, which is a reasonable spread for an annual industrial volume. The probability of exceeding 35,200 is about 40 per cent. So the commitment is more likely than not to be worse than doing nothing, and that is before anybody argues about whether the mean is right.
Expected cost is the number to compare. For a commitment Q, expected billed volume is the mean plus the expected undershoot, which is the standard normal loss function applied at z = (Q - mean) / sigma. At Q = 40,000, z = 1.2, and the expected undershoot is 5,000 times 1.2561, which is 6,280 tonnes. Expected billed volume is 40,280 tonnes, and multiplying by 0.88 gives an effective 35,447 tonnes at full price. Against 34,000 tonnes with no commitment, the 12 per cent discount has produced a 4 per cent cost increase.
Run the whole ladder before picking a tier
Suppliers quote a ladder. Take the same item with three tiers: 30,000 tonnes at 8 per cent, 35,000 at 10 per cent, 40,000 at 12 per cent. Same demand distribution, mean 34,000, standard deviation 5,000. Expressing each option as effective full price tonnes:
At 30,000 committed, z is -0.8, the expected undershoot is 601 tonnes, billed volume is 34,601 and the effective figure after 8 per cent is 31,833.
At 35,000 committed, z is 0.2, the expected undershoot is 2,535 tonnes, billed volume is 36,535 and the effective figure after 10 per cent is 32,881.
At 40,000 committed, the effective figure is 35,447 as calculated above.
With no commitment at all, the figure is 34,000.
The lowest tier wins by a wide margin, saving about 6 per cent against buying uncommitted, while the headline tier loses to doing nothing. That ordering is typical, and it survives reasonable changes to the inputs. Expected shortfall grows faster than the discount ladder does once the commitment passes the middle of the demand distribution, because undershoot risk compounds while the discount steps are linear and small.
The general shape holds: the right commitment usually sits below the mean of your demand distribution, often around the 25th to 35th percentile depending on how steep the ladder is. Commit to the volume you would consume in a bad year, and buy the rest at whatever the uncommitted price turns out to be.
Two clauses that move the answer more than the discount does
The deficiency payment rate. If shortfall units are paid at a fraction f of the contract price rather than in full, break-even consumption becomes Q(1 - d)f divided by (d + (1 - d)f). At f = 1 that collapses back to (1 - d)Q. At d = 12 per cent and f = 30 per cent, the same 40,000 tonne commitment breaks even at 27,500 tonnes rather than 35,200. The top tier goes from a bad deal to a comfortable one on the strength of one number that is rarely the focus of the negotiation. Negotiating f down is worth more than negotiating d up, and suppliers concede it more readily because it costs them nothing in a year where you perform.
The measurement period. An annual commitment lets a weak March net against a strong October. A monthly commitment does not, and the difference is large. Split the 30,000 tonne annual commitment into 2,500 tonnes a month against monthly demand averaging 2,833 with a standard deviation of 1,443, which is the same annual variability spread across twelve independent months. Expected undershoot is 424 tonnes per month, or about 5,090 tonnes across the year, against 601 tonnes for the same volume measured annually, which is roughly eight times the exposure on an identical nominal commitment. If a supplier proposes quarterly or monthly measurement in exchange for a slightly better rate, price it before agreeing.
Two more terms deserve a read. Make-up rights let you take units in a later period that you have already paid for, which converts a loss into a prepayment and is standard in gas contracts. Assignment and resale rights decide whether excess volume can be sold on or transferred to a sister site, which is the difference between a shortfall and a redeployment.
What happens to the plan once it is signed
A commitment is a supply floor, and most planning systems have nowhere to put one. MRP consumes what demand requires, netting against existing stock, and it will happily under-consume a contract for eleven months without any complaint, because nothing in the netting logic knows the contract exists.
The monitoring that works is a cumulative one. Each week, record volume taken to date against pro rata entitlement to date, and project the year-end position by carrying the current demand plan forward. Alert when the projection crosses the commitment minus a tolerance you set, with enough lead time that a decision is still available. On a twelve month contract, the last useful decision point for most industrial volumes is around month seven or eight, because that is the last moment a genuine demand response can move the number.
When the projection does breach, the choice is between paying the shortfall and buying material you do not need yet, and the deficiency rate decides it. At f = 1 you have already paid for the goods, so taking them is better than not taking them as long as they keep and you have somewhere to put them. At f = 0.3, taking 2,000 tonnes costs 88 per cent of price against 26 per cent for walking away, and only a strong view on future consumption justifies the difference. Working that comparison at the point of decision beats working it in the year-end review.
One accounting note, because it changes when the pain becomes visible. Under IAS 37 an onerous contract provision is recognised when the unavoidable costs of meeting an obligation exceed the benefits expected from it, with the IASB's May 2020 amendment clarifying which costs of fulfilment count, effective from January 2022. In practice, a shortfall your team can see coming in month eight may have to be recognised by finance before the invoice arrives, and finance would rather hear it from planning than from the auditor.
There is also an enforceability question worth knowing exists. In English law contracts, the UK Supreme Court's 2015 decision in Cavendish Square Holding v Talal El Makdessi reframed the penalty rule around whether a clause protects a legitimate interest proportionately. Take or pay obligations are usually drafted as primary payment obligations, which sits them outside that test, while a clause framed as punishment for a breach can fall inside it. Where this matters, ask your legal team rather than a planning blog.
Where this stops
The whole calculation assumes the uncommitted alternative exists. In a tight market it may not, at any price, and then the commitment is buying availability, with the discount a secondary benefit. Shortfall arithmetic cannot value that. What values it is a scenario where the volume is unavailable for a quarter and you cost the lost production, which is a different exercise producing a different answer, often in favour of committing to more rather than less.
The distribution is the other soft spot. Volume commitments run twelve to thirty-six months, and forecast error at that horizon is poorly described by a standard deviation measured on recent history, particularly where the demand itself depends on a customer contract that may or may not renew. Where that is the case, replace the normal with two or three explicit scenarios and weights, run the same expected cost comparison across them, and read the output as a ranking of the tiers. The ranking is usually stable even when the inputs are not, which is what makes the exercise worth doing at all.
And every input here is negotiable. The tiers, the deficiency rate, the measurement period and the make-up rights all move. The calculation is an input to that conversation, and its main value is knowing which of the four to spend your negotiating capital on.
Take the largest live volume commitment you have, find the deficiency rate and the measurement period in the contract, and compute the break-even consumption against your year to date run rate annualised.