The vendor did not go bankrupt. That is the part I had not planned for. They simply stopped replying, over about five weeks, while still technically trading and still invoicing. Our contract had an insolvency clause and no insolvency ever happened. Nine months of work sat on machines we could not reach, and the clause we were relying on did not apply to the situation we were actually in.
What follows is the method I use now. It is dull on purpose. The whole value of escrow lives in four or five details that are easy to skip at signature and impossible to fix on the day you need them.
Step 1: Run the Triage Test Before You Ask for Escrow
Most escrow arrangements I see should not exist. They cost money, they add a negotiation, and they protect systems the company could rebuild in a fortnight. Meanwhile the one integration nobody else understands has no protection at all, because it was never a formal project.
Score the engagement out of four. One point for each yes:
- Does this software run a process where a month of downtime would cost real revenue, or breach an obligation you have made to a customer or regulator?
- If you had the code tomorrow, could your own team or a replacement vendor realistically take it over? If the answer is no, escrow is not your gap, documentation is.
- Is the vendor small enough that one person leaving or one bad quarter ends your relationship with it?
- Is the code genuinely unavailable to you today, as opposed to sitting in a repository you could simply be given access to?
Two or more, set up escrow. Fewer than two, and the cheaper instrument is a repository access clause plus a documented handover obligation, which costs nothing and covers the common cases. That fourth question eliminates roughly half the requests I see. A surprising number of companies pay for an escrow arrangement to protect code the vendor would have given them read access to if anyone had asked.
Our expert take
Escrow is a continuity instrument, not a trust instrument. If you are considering it because you do not trust the vendor, the arrangement will not save you, because every failure mode you are worried about will arrive faster than a release process can resolve. Fix the vendor choice. Use escrow for the vendors you do trust, against the risk that something happens to them that neither party chose.
Step 2: Write Release Triggers You Can Actually Evidence
This is where my own arrangement failed. The trigger was insolvency. The reality was a company that was still alive and had simply stopped working. Insolvency triggers are standard, easy to agree and almost never the thing that happens to you.
Keep the insolvency language, then add operational triggers that you can prove from your own records without needing the vendor to admit anything:
- Non-delivery. Failure to deliver a contracted release or milestone within a defined number of days of the agreed date, after a written notice period.
- Non-response on incidents. Failure to acknowledge a severity one incident within the contracted response window, a defined number of times in a rolling quarter. Your own ticketing system is the evidence.
- Loss of key personnel. Departure of the named individuals in the contract without a replacement proposed within a defined window. Name people, by role at minimum, or this clause does nothing.
- Abandonment of the product line. For product vendors, a public end-of-life announcement or withdrawal of support for the version you run.
Two drafting details decide whether these work. First, every trigger needs a number of days attached, because “unreasonable delay” is an invitation to argue. Second, specify who decides the trigger has fired and on what evidence. An arrangement where release requires vendor agreement is not protection, it is a formality. Expect a notice and objection window, which is fair, but it should be measured in days with a defined end.
Step 3: Define the Deposit as the Whole Build Chain
Ask for “all source code and related materials” and you will receive a zip of source code and nothing that lets you use it. Source is the least valuable part of the deposit, because it is the part you could partly reconstruct. What you cannot reconstruct is how it was assembled.
Put a schedule in the agreement that names artefacts individually:
- Source for every component, including scripts and tooling that are not part of the shipped product
- Build scripts and the build configuration, with pinned dependency versions, not version ranges
- Infrastructure definitions, container files and deployment manifests
- Database schema, migrations and any seed data needed to reach a working state
- Environment variable names, their purpose and expected format, with the values held separately by you
- An inventory of every third party service, what it does, and which account owns it
- A written build and deploy runbook, tested by someone who did not write it
- Cryptographic keys and certificates only by reference, with a documented rotation path, never as deposited secrets
The runbook and the pinned versions do most of the work. I have watched a competent team spend eleven days failing to build a deposit that contained every line of source, purely because a transitive dependency had moved on and nobody had recorded which version the build had ever worked against.
Step 4: Match the Arrangement to How the Software Is Delivered
Three shapes, and picking the wrong one is how companies end up paying for protection they do not have.
| Shape | Fits | Watch out for |
|---|---|---|
| Single beneficiary deposit | Bespoke software built for you by an agency or dedicated team | Costs more per arrangement, so triage properly first |
| Multi beneficiary deposit | A product vendor offering the same deposit across its customer base | Deposit scope is written for everyone, so check it covers your configuration and modules |
| Continuity terms, no code deposit | Anything the vendor hosts and operates for you | Code alone restores nothing here. You need data, configuration and a handover path |
That third row is the most expensive misunderstanding in this whole area. If a vendor runs the software as a service, receiving its source code on a trigger day leaves you with code and no data, no tenant configuration, no infrastructure and nobody who has ever operated it. The right ask is a tested export of your own data on a defined cadence in a documented format, a copy of your configuration, the infrastructure definitions, and a written handover path with a notice period. The question for a hosted vendor is never “do you offer escrow”. It is “what would I have on a Monday morning if you stopped answering”.
Not sure which shape your engagement needs
Tell us how your development is delivered today, in two or three sentences. We will tell you whether a code deposit protects anything in your case, and what we would ask for instead if it does not.
Get startedStep 5: Pay for Verification, the Line Item Everyone Cuts
Verification is the difference between an arrangement and a comfort blanket. It is also the first thing removed when someone is trimming a quote, because it is the only part with a visible recurring cost and no visible recurring benefit.
Choose a level deliberately and write it into the agreement:
- File listing and integrity check. Confirms something arrived and has not been corrupted. This is the floor. It does not tell you the deposit is usable.
- Compile and build test. An independent party builds the deposit from its contents alone and reports whether it succeeded. This is the level I would put on anything that matters.
- Full rebuild into a running environment. The deposit is stood up and exercised against defined criteria. Expensive, slow, and correct for a small number of genuinely critical systems.
The strongest argument for level two is that it converts a silent future failure into a present, fixable finding. A failed build test today is a two week fix while the vendor still cares about you. The same failure discovered on a trigger day is a project.
Our expert take
If budget forces a choice between a cheaper escrow provider with build verification and a premium provider with a file listing, take the build verification every time. The provider brand matters on the day of a dispute. The verification level matters on every other day, and it is the only part of the arrangement that tells you the truth about whether the thing you are paying to protect can actually be restored.
Step 6: Tie Deposit Cadence to Milestones, Not the Calendar
An annual deposit against a fortnightly release cycle means your worst case fallback is up to a year behind production. Technically compliant, practically useless.
Attach deposit updates to events that already exist in your delivery process, so nobody has to remember a date:
- Every major release, with deposit confirmation as a condition of release sign off
- Every change to pinned dependencies or infrastructure definitions
- Every change of named key personnel
- A quarterly floor, so a quiet period does not leave the deposit stale
Then make the confirmation visible to someone who is not the vendor. A deposit obligation nobody tracks decays within two quarters, in my experience without exception. One line in your release checklist and one recurring reminder with the deposit receipt attached is enough. The metric worth watching is the gap between your current production version and the deposited version. If it exceeds one release cycle, the arrangement is drifting.
Step 7: Rehearse One Release, Then Diarise the Review
Run a restore exercise once, inside the first quarter, with your own engineers. Give them only what is in the deposit. No access to the vendor, no asking a friendly developer a quick question. Time how long it takes them to reach a running build.
That number is the real value of your escrow arrangement, and it is almost never what people expect. I have seen a well-constructed deposit restored in under two days and a badly constructed one defeat a strong team for a fortnight. Whatever the number is, write it down, fix whatever blocked them, and tell the vendor what you found, because a good vendor will want to fix it too.
Then attach the escrow review to the contract renewal date rather than to a random month. Renewal is when you have commercial leverage, when scope has usually changed, and when raising verification level or deposit cadence is a normal part of the conversation instead of an accusation.
The 4 Things Escrow Will Never Fix
Worth saying plainly, because escrow gets sold as a general answer to vendor risk and it is a narrow instrument.
- Knowledge. Code plus a runbook is not the same as a person who knows why a decision was made. Escrow buys you a starting point, not continuity of understanding. The defence against that is documentation and more than one person touching each area, during the engagement.
- Ownership questions. A deposit says nothing about who owns what. Ownership is settled in the contract, and it is a separate piece of work we cover in intellectual property ownership in UAE developer contracts.
- Third party components. Your deposit can be perfect and still be unusable if a component inside it is licensed to the vendor rather than to you. Ask for the component inventory with its licence terms at step 3, not at trigger time.
- A slow vendor. Escrow fires on failure, not on disappointment. If the real problem is that delivery is late and quality is poor, the instrument you need is a different commercial structure, which is the subject of fixed price against time and materials contracts in Dubai.
How to Raise It Without Souring the Relationship
Three sentences that work, and I have used them often enough to be confident in them. Raise it at contract or renewal stage, not mid-project. Frame it as business continuity rather than trust, because nobody argues with continuity and most vendors have been asked before. Offer to pay for it yourself and keep the triggers narrow and evidenced, which removes the vendor’s real fear, which is a customer seizing the code during a routine dispute.
A confident vendor agrees quickly, and some will already hold a multi beneficiary arrangement you can join for a fraction of a bespoke one. Genuine resistance is itself useful information, and the shape of the objection tells you a lot. Resistance to the cost is normal. Resistance to verification is the one I pay attention to, because it often means nobody is confident the code builds anywhere except on the machine of the engineer who wrote it.
For the wider vendor and delivery picture, our guides to contract management systems in Dubai and enterprise software development in the UAE cover the surrounding decisions.
Frequently Asked Questions
Is source code escrow worth it for a small Dubai company?
What is the most common reason escrow fails when it is finally needed?
Does escrow help if the vendor runs the software as a service?
Who pays, and how do you raise it without souring the relationship?
Lance-toi: start with the triage test
Score your engagement out of four using step 1. If you land on two or more and want a second pair of eyes on the deposit scope before you sign, send it over. We also help UAE employers build the in-house capability that makes a deposit usable in the first place.
Get started