Back-to-Back Is an Intention, Not a Mechanism

The commercial manager is confident about the package because the subcontract was written back-to-back. Same scope wording, same rates basis, same retention, same programme. Whatever the client does to the main contractor flows down; whatever the main contractor is entitled to flows up. On paper the risk is passed through.

Then the final account arrives and the two sides do not agree. The subcontractor's account carries eleven variations; nine of them have a matching claim against the client and two do not. Retention has been released to one subcontractor whose defects period ran from a date that has nothing to do with the main contract's. Three back-charges were agreed in a site meeting and never appeared in either ledger. And the net position — what has been certified to us against what we owe out — cannot be produced on one page for anybody, which is one of the standard symptoms that a contracting business has outgrown its systems rather than a failure of any one person.

None of that is a drafting failure. The subcontract is genuinely back-to-back. What is missing is the thing that makes a back-to-back clause operate: a system in which every downward event is joined to the upward event it mirrors, and any event that has no partner is visible as an exception the same day rather than at the final account.

What back-to-back does and does not mean

It is worth separating two things that get conflated in the same sentence.

The contractual intention is that the subcontractor's entitlements, obligations and timings correspond to the main contractor's. Whether a particular clause achieves that — and in particular whether payment conditional on the employer's payment is effective — is a question of law and of the specific wording, and it belongs to your counsel rather than to an article about systems. Nothing below turns on it.

The operational reality is that the correspondence is performed by people, event by event, across two sets of documents governed by two contracts with two counterparties. The clause states the intention. The correspondence either happens or does not, and by default it does not, because nothing in an accounting system knows that a subcontract variation is the reflection of a client instruction.

That gap is the whole subject. A back-to-back clause with no mechanism behind it is a statement of what should have happened, produced at the point when it is too late to make it happen.

The five places the mirror breaks

They are always the same five. In most contracting businesses two of them account for nearly all the loss.

Variations instructed downwards with no claim made upwards. A main contractor is on both sides: the client instructs variations to you, and you instruct variations to subcontractors. The expensive pattern is a variation issued down without the corresponding entitlement recorded up — the subcontractor's account grows, the cost lands, and there is no matching claim against the client, either because the upward notice was missed or because nobody joined the two events. The governance for this is set out in how variations have to be handled in both directions, and the joining is the part that has to live in the system.

Retention that does not actually mirror. Retention runs in both directions and the two are not the same instrument. Your rate and cap come from the main contract; each subcontractor's rate and cap come from that subcontract, separately. More importantly, the release trigger differs: yours is your completion and your defects period, and theirs is each subcontractor's completion and each subcontractor's defects period. Treating these as one figure is how a contractor ends up releasing money out before receiving it in, and the mechanics of holding both sides are worked through in how retention behaves as two separate positions.

Notice periods that are shorter downwards than upwards, or the reverse. Two contracts, two notice regimes. If the subcontract requires notice from them in seven days and the main contract requires notice from you in fourteen, you have a window. If it is the other way round, you have a liability, and it is invisible until somebody misses a date. This is a calendar problem before it is a systems problem, and it needs an owner and a reminder rather than an understanding.

Back-charges and contra-charges. Costs you incurred on a subcontractor's behalf, coming back the other way. On the client-facing certificate these are rare; on the subcontract certificates you issue they are constant, and they are the single most commonly ungoverned document in contracting. They are agreed verbally on site, they are remembered differently by both parties, and they surface at the final account as an argument. The remedy is that a back-charge is a document with a number, a date, a value and an acknowledgement — the same standing as anything else in the certificate chain.

Measurement that diverges. Your quantities against the client are remeasured; the subcontractor's quantities against you are measured on their own basis, sometimes on a different bill and occasionally on a different drawing revision. Once the two measurement bases drift apart, the mirror is broken arithmetically rather than commercially, and no amount of contract wording repairs it.

The one field that fixes most of it

The mechanism is smaller than the problem suggests. Every downward commercial event carries a link to the upward event it mirrors, or an explicit statement that it has none.

A subcontract variation points to the client-side variation it belongs to. A back-charge points to the cost event that caused it. A retention release points to the completion certificate that triggered it. An advance payment recovery points to the advance it recovers.

And crucially: an event with no partner is not blocked, it is flagged. Plenty of downward events legitimately have no upward mirror — work you got wrong, a subcontractor's own inefficiency, a scope decision you took commercially. The point is not to prevent them. The point is that "unmirrored" becomes a state with a reason and an owner, reviewed weekly, rather than a discovery made in the final account by whoever is reconciling.

This is the whole trick, and it is why the problem survives good drafting and good people. The information exists on both sides. Nothing joins it, so nobody can see the gap until it has stopped being fixable.

The report that ends the argument

One page, per contract, showing both directions at once.

Certified to us, to date. Certified out, to date. Retention held against us and retention held by us, separately, with the release triggers dated. Variations claimed up and variations instructed down, with the unmatched ones listed by name. Back-charges raised and acknowledged. Advance received and recovered against advance paid and recovered.

If that page can be produced on demand, most of the arguments above never reach a final account, because the divergence is visible in the month it happens. If it cannot, the business is running on the assumption that the paperwork mirrors — and the assumption is checked once, at the end, when the answer costs money.

It is the same discipline that turns work already done into a number you can defend, which is the subject of the gap between what has been done and what has been billed. Both are failures of the same kind: a real commercial position that exists in documents and does not exist as a figure.

Where to start if the paperwork is already behind

Take one live contract, not all of them. Preferably one that is roughly half complete, with at least two subcontract packages and some variation history. Fixing everything at once is how this becomes a project nobody finishes.

Reconstruct the join backwards. For each subcontract variation on that contract, find the client-side event it mirrors, or record that there is none and why. This is tedious, it takes a day or two per contract, and it is the only way to know the size of the exposure rather than fear it.

Then hold the line forwards. From a stated date, no downward commercial event is raised without either a link or a flagged reason. That is a rule about how documents are created, so it belongs to the commercial team rather than to finance, and it needs one person who reviews the unmatched list every week.

Put the notice dates in the same calendar as everything else. Both directions, per contract. A missed notice is the cheapest loss in contracting to prevent and one of the most expensive to argue about afterwards.

Leave the guarantees alone until the rest works. Advance payment guarantees and performance bonds discharge on dates rather than on progress, and they need their own register — but they are a separate discipline with its own failure mode, described in why guarantees behave unlike every other contract document.

The short version

A back-to-back subcontract states an intention. Making it true is an operational job that nothing in a standard accounting system performs, so by default it is not performed.

The mirror breaks in five places: variations instructed down with no claim up, retention whose rates and release triggers genuinely differ on each side, notice periods that do not align, back-charges agreed on site and recorded nowhere, and measurement bases that drift apart.

The mechanism is one link. Every downward commercial event points to the upward event it mirrors, or carries an explicit unmirrored flag with a reason and an owner. Unmatched events are reviewed weekly, not discovered at the final account.

Then build the one page that shows both directions per contract — certified in and out, retention both ways with dated triggers, variations up and down with the unmatched listed, back-charges, advances. Start with a single half-complete contract, reconstruct its joins backwards to size the exposure, and hold the rule forwards from a stated date. That sequencing is ordinary implementation work on a contracting business rather than a legal exercise, and the contracts you already signed are fine.