The Hedging Process: Five Stages from Contract to Settlement

Most conversations about hedging start with whether to hedge at all, and at what ratio. What actually decides whether hedging survives contact with the business is something else: whether contracting, pricing, execution, attribution and settlement connect end to end. When the line is closed, exposure is real. When it is broken, a hedge ratio on paper is only a hedge on paper.

In most of the companies we have seen, all five stages exist — and every handover between them runs through a person. Contracts are registered by sales, positions sit with trading, limits are written down on paper by risk, pricing requests arrive in a chat group, and when fills come back, operations tags them by hand before they can be attached to a contract.

The direct consequence of working this way is not "slow". It is books that do not match reality: the exposure in the system and the exposure the company actually carries are not the same number.

Stage 1: Contracting — where exposure starts, and the step most likely to distort it

Once a contract is signed, the company holds three things at once: a fixed quantity, an undecided price, and a delivery that will happen on some future date. Together, those three things are the exposure.

In practice, the contract ledger and the futures account are two entirely separate books. Sales knows how much the customer bought, trading knows how many lots it holds, and nobody can say, at any moment, how much exposure is left on a given contract.

There is one technical point here that many companies overlook: different pricing forms create different exposure shapes. A fixed-price contract fixes its price at signature, so its exposure is largely a quantity question. A deferred-pricing contract has no price at all until the pricing date, so its exposure covers quantity and time together. Average-price and basis contracts sit between the two. If the system aggregates exposure only by contract value, the risk characteristics of all four are flattened out into one number.

The first thing to fix is to make the pricing form a primary dimension of exposure calculation, not a note field attached to a contract.

Stage 2: Pricing requests — where most of the "could have been done right" sits

Pricing is the moment exposure actually starts to move, but many companies treat it as pure administration: the customer says how many days to price, the trader relays it, the trader executes.

The problem is that a pricing request should carry two checks, and in practice often carries neither. The first is authorisation — is this customer, under this contract, allowed to price at this time in this way. The second is limit — after this pricing, is total exposure still inside tolerance.

Without those two checks, pricing becomes a process where the loudest voice decides. And once pricing has happened the exposure has already changed materially, which leaves very little meaning in approving it afterwards.

Another fact that gets underestimated: pricing requests are clustered by session. Across the companies we have worked with, night-session pricing regularly accounts for around 35% of the total. A company that staffs day hours only therefore starts with a third of pricing requests unhandled — they usually pile up until the next day, and by then the price is no longer the one the customer expected.

Stage 3: Execution — the underestimated night-session problem

The cost of manual coverage is not just payroll. It is response delay. In the traditional approach, during the window when requests concentrate, waiting time per request often exceeds 10 minutes. Long waits produce repeated submissions, calls from sales to chase, and eventually duplicate or missing orders.

Two things have to be kept apart: automating routine requests, and using automation to replace risk judgement. The first is safe — requests with clear rules, sufficient limit and limited risk can take a standard channel. The second is dangerous — over-limit, near-delivery and abnormal-basis requests must stay with a person, and must leave a record of who handled them.

The mature dividing line is: routine goes automatic, exceptions go to a person and get logged. Mixing the two (automating everything, exceptions included) or keeping everything manual (no night cover) both fail.

Stage 4: Attribution — the core of hedging, and the easiest stage to fudge

The entire effectiveness of hedging rests on one thing: every futures fill must map one to one onto a physical contract. If it does not, the P&L cannot be computed, the exposure cannot be computed, and the hedge ratio cannot be computed either.

The reality of manual tagging: dozens to a hundred trades a day, 8 to 12 minutes of human time on each, several hours of intense repetitive work every day. Under that load, missed tags and double tags are not a capability problem — they are a certainty. The error and omission rate observed in industry practice usually sits between 1% and 3%.

More troublesome is the amplification: one attribution error mis-states the exposure on that contract, which then distorts the hedge ratio judgement, the margin estimate and the final assessment of hedge effectiveness. A 1% error rate can turn into a number that has to be explained to a board when it lands in the quarterly report.

There is one more cost that is easy to miss: correction is expensive. If every correction requires a full recomputation and a manual re-check of the whole chain, correcting a single historical record can take half an hour. Once a system makes people afraid to correct, errors become permanent in the books.

Stage 5: Settlement and archiving — decides whether the system survives three years

This is the stage most often skipped, and it is where many companies' systems eventually collapse.

A hedging system that is to run for years has to answer three questions from audit and internal control: who made this decision and when, what data was it based on at the time, and has it been reviewed since.

The fatal flaw of the traditional approach is definition drift: each month's report is exported by a different person, at a different time, with different filters, and when the numbers do not reconcile, everyone goes back to email attachments and compares them one by one. Retrieving history beyond 30 days routinely takes a dozen minutes — and that is exactly the period in which review matters most.

The real marker of a closed loop is a small but difficult metric: any operation from three months ago can be found and explained within seconds.

Five stages side by side: what is genuinely managed, and what is only recorded

We suggest working through this table row by row. The question is not whether a stage happens, but whether the way it happens is verifiable — that gap is the real gap in hedging maturity.

A self-check you can run today

If you do not want to rebuild the entire process at once, answer these three questions first. None of them needs a system to verify, and the answers surprise most people:

1. Add up the exposure on every live contract right now and compare it with the net position in the futures account. Do they tie? If they do not, what is the difference?
2. Over the last 30 days, how many pricing requests were rejected automatically by system rules, and how many were executed first and authorised afterwards?
3. Pick one fill from three months ago at random and try to find the contract it belongs to, who attributed it, and the limit check at the time. How long does that take?

The answers usually say more about where your company actually stands than any hedge ratio recommendation.

One qualification: many breaks do not mean you must go to a system immediately. If you have few contracts, a single commodity and single-digit daily volume, a well-designed spreadsheet plus a fixed reconciliation rhythm can reach an acceptable level. But with hundreds of contracts, hundreds of trades a day and night-session exposure, the cost and the error rate of a manual approach cross the threshold at the same time — and what you need then is not a better spreadsheet, but rules written into a system so that they execute identically every time.

Data sources and notes

The practice data in this article — the five process stages, the time-of-day distribution of pricing requests (night session around 35% of the total), manual pricing and attribution time (8-12 minutes per trade), the error and omission rate range (1%-3%), and the target levels after systemisation (attribution within 1 minute per trade, error rate of 0.3% or less, history search within 3 seconds per 10,000 records, night-session automatic processing coverage of 90% or more, and manual watching hours reduced by 60% or more) — come from RMC's own observations and statistics gathered in commodity risk management diagnostics for trading and manufacturing companies. Figures are given as ranges rather than single points, so readers can calibrate against their own situation. Nothing here constitutes an endorsement or evaluation of any specific software product.