Mine-site laboratory contracts are evaluated on cost per sample, turnaround time, accreditation scope and mobilization schedule. Very few are evaluated on what happens to the data. Yet the decision to outsource a laboratory is almost always, and usually implicitly, a decision about who holds the analytical record of the operation.
This is not an argument against outsourcing. It is an argument for separating two things that contracts routinely bundle together: the service, which is genuinely worth buying, and the system of record, which is worth keeping.
What Outsourcing a Laboratory Actually Buys
The case for a contract laboratory on site is strong, and any argument that skips over it is not worth reading.
The first item is capital. A mine-site assay laboratory is a building with ventilation, water and effluent treatment, a preparation line of crushers and pulverizers, fire assay furnaces and cupellation, balances, ICP-OES or AAS, sometimes XRF or LECO, plus reagent handling and waste disposal. That capital competes directly with the pit and the plant, and it rarely wins the argument in a capital committee.
The second is people. Assayers, furnace operators and QA/QC chemists are scarce, slow to train and difficult to retain at a remote site. A contract laboratory group carries that labour problem as its core business, with a bench of staff it can rotate across operations.
The third is elasticity. Exploration campaigns, infill drilling programs and plant commissioning produce sample peaks that last months, not decades. A fixed in-house laboratory has to be sized for the peak and then paid for during the trough.
The fourth is accreditation. A contractor typically arrives with an ISO/IEC 17025 scope already maintained across many sites, with proficiency testing, method validation and an internal audit program that took years to build. For an owner starting from zero, that is a genuine shortcut.
And the fifth is accountability. One contract, one counterparty, defined turnaround commitments, and a service that can be measured and penalized.
None of that is in dispute. A well-run contract laboratory will frequently outperform an under-resourced in-house one.
The Clause That Is Not in the Contract
What is rarely negotiated is the system underneath. The contractor mobilizes with its own laboratory information management system, and in many commercial models that software arrives inside the service fee rather than as a separate line item. It is an efficient arrangement on day one: nothing to procure, nothing to configure, no implementation project competing with the ramp-up.
The consequence is structural rather than moral. A system supplied by the provider is administered by the provider, and that has four practical effects.
The database instance belongs to the contractor. Schema, backups, retention policy, user administration and access rights sit on the supplier's side of the fence, often on the supplier's global platform rather than on the mine's network.
History accumulates inside it. Every sample login, batch, control insertion, re-assay, approval, revision and released certificate is written into an instance the owner does not administer. Over a five-year contract that becomes the single richest description of the orebody the operation has.
Export formats are defined by the supplier. What the owner receives is a report or an extract — usually results, sometimes with a QC summary. That is not the record. The record is the layer underneath: sample states, batch structure, control charts, instrument raw imports, audit trail, who approved what and what changed afterwards.
Contract exit is a migration event, not a handover. It has to be scoped, priced, scheduled and technically validated at the precise moment when the commercial relationship is winding down and neither party has much appetite for the work.
To be clear about what is not in question: the certificates belong to the client, and no serious laboratory group disputes that. What does not transfer automatically is everything that makes a certificate defensible five years later.
The Moments When It Starts to Hurt
The gap stays invisible while the contract is running smoothly. It becomes visible in five recognizable situations.
A change of contractor. The incoming laboratory arrives with an empty database. Control charts restart from zero. Historical CRM behaviour, method bias, trueness by batch and duplicate precision — the accumulated quality memory of the operation — turns into an archive of PDFs instead of a queryable series. Trends that took years to establish are not transferred; they are re-started.
A commercial dispute over a lot. When a shipment grade is contested and an umpire assay is called, the question is rarely what the number was. It is how the number was produced: which pulp, which batch, which controls were inserted, which balance and which instrument file, who approved the release, and what was revised after the fact. That evidence lives in the operating system, not in the certificate.
Reconciliation between mine, plant and port. Reconciliation works by comparing measurements that were taken independently. If the assay data on which several sides of the balance depend sits in one system administered by one party, the exercise can still be performed — but it cannot be re-run independently, which is a different thing.
An audit that asks you to reconstruct a three-year-old result. Assessment bodies, internal audit and technical reviewers all do this. It is a reasonable request. It becomes an awkward one when the reconstruction path runs through a company whose contract ended two years ago.
Due diligence. A buyer, a lender or a technical reviewer assessing an operation will ask for the analytical history behind the resource statement and the production record. "We can request it from our laboratory provider" is a materially weaker answer than "here it is, with its audit trail".
The Alternative: The Owner Provides the LIMS, the Contractor Operates Inside It
Nothing in the logic of outsourcing requires the contractor to also own the system of record. The service can be outsourced while the record stays at home. Making that work in practice takes four things, and the fourth is the one most often ignored.
Roles and segregation by contractor. Each external laboratory operates within its own scope: its methods, its work orders, its instruments, its batches, its analysts. It should not see another contractor's work or the owner's commercial layer. Done properly, rotating a contractor becomes a permissions change and a training week rather than a migration project.
Service levels written on the data, not only on turnaround. Which fields are mandatory at login, what QA/QC must accompany a batch before it can be released, how long a result may remain unapproved, what qualifies as a valid re-assay, and how a non-conformance is recorded and closed. These belong in the contract next to the turnaround commitment.
Interchange formats agreed before mobilization. A contractor should not be forced to abandon the instrument parsers and internal tooling that make it efficient. It should be required to land the result in the owner's system with its metadata intact — sample identity, method, batch, controls, instrument and operator — and to be able to read back what it needs.
The external laboratory has to be comfortable working in it. This is where owner-provided systems usually fail. If the owner's LIMS is slower, more rigid or less instrument-aware than the one the contractor brought, the bench will work around it, and the record quietly degrades into a copy of a copy typed in at the end of a shift. Barcode-driven reception, direct instrument capture, real batch handling, tolerance for a poor site link, and response times that survive a full shift are not conveniences. They are what keep the record honest.
This is the model OnLIMS is built for. OnLAB handles reception, batch control, preparation states and result capture on site, with direct instrument integration so that data arrives from the balance, the ICP, the AAS or the XRF rather than from a keyboard. OnQMS carries the quality-system layer around it — controlled documents, non-conformances, competency and training records, calibration and the clause-level evidence an ISO/IEC 17025 assessment asks for. Deployment is on-premise, on the owner's SQL Server, on the owner's network, which is the specific property that lets the record outlive any single contract.
Five Questions Worth Putting in the Tender
None of this requires a change of strategy. It requires five questions in the request for proposal, asked before signature rather than after.
1. Who owns the database? Not the results — the instance. Where does it run, who administers it, who can grant and revoke access, and under whose jurisdiction is it hosted.
2. What happens to the history when the contract ends? Ask for the deliverable, the timeline and the price now. A cost that is not defined at tender will be defined at the moment your negotiating position is weakest.
3. In what format, and with what metadata? A spreadsheet of results is not the record. Ask specifically for sample identity and states, batch structure, QC insertions and their outcomes, instrument raw imports, the audit trail, approvals and revisions, and the method and version behind each determination.
4. Can I audit live? Read access to the operating system during the contract, not a monthly summary. Can your own QA lead open a batch and see its controls on the day it runs?
5. Can a certificate be reconstructed on demand? Pick one at random and ask the bidder to walk it backwards to the raw instrument file and the controls that released it. The answer separates a client portal from a system of record faster than any capability matrix.
None of this makes outsourcing the wrong decision. It makes it a decision with a second half. The laboratory service can be bought, renewed, re-tendered and replaced. The analytical history of an orebody cannot be re-created. Keeping those two things on separate contracts is the cheapest continuity an operation can buy — and it is far cheaper to arrange before signature than after.