Construction Materials Testing Software Problems: The 6 That Cost Real Money, and How to Avoid Them

Why does field capture get scoped as a form so often?

Because the visible artefact is a field report, so the specification says let technicians submit field reports on a tablet. That gets built, it demos well, and it fails on contact with a jobsite because the report is not the point. The sample is.

Everything downstream depends on knowing what a cylinder is: which placement, which truck, which structural element, which mix design, which slump and air content, cast by whom, at what time, cured how. That information originates in weather, often by a technician who is simultaneously watching a pump and a finishing crew. Two design decisions determine whether it survives.

• The identifier has to be created on site at the moment of casting, printed or scanned onto the sample, not assigned later at lab intake when someone reads marker off the side of a mould.
• Every custody transfer has to be an event: technician to vehicle, vehicle to lab intake, intake to curing tank position, tank to break.

With those in place, a crate arriving short surfaces at intake rather than on day 29. Without them you have digitised the paperwork and kept the failure.

The second scope failure in the same area is assuming connectivity. The jobsites where this matters have no signal, and a field application that requires a network will be abandoned inside a month and replaced by paper. Offline capture with a local queue and sync on return is not a nice to have in this industry, it is the difference between adoption and a wasted budget. Test it by taking the application into a basement, not by reading a specification.

What goes wrong when you migrate open samples and project history?

The migration problem here is unusual and it catches people out: you are cutting over with physical inventory in a curing tank. A concrete cylinder cast three weeks before go live has a 28 day break that falls after it, and it exists as a mould with a marker label and a row in whatever you ran before.

Three failures recur. Open samples get imported without their true casting date, so the break calendar generates wrong dates for exactly the samples closest to breaking. Historic results arrive without the specification they were assessed against, which makes them useless for trend analysis and dangerous if anyone re reports from them. And client names are typed differently in four places, so a firm with three hundred projects discovers it has four hundred and forty clients.

The approach that works is to migrate physically as well as digitally. Take a full inventory of what is in the curing tanks and on the shelves, reconcile it against the incoming data, relabel every open sample during the count, and run the two systems in parallel only for results, never for scheduling. The break calendar needs a single owner from day one, because two calendars is worse than one whiteboard. Import historic results as read only records with their original report attached. Budget two to three weeks and pick the quietest part of your year.

Why do the integrations that matter here break after launch?

Testing firms rely on three connections and each degrades in a way you will not see in acceptance testing.

Machine integration is the most valuable and the most variable. Reading a compression machine, a nuclear density gauge and a maturity logger are three unrelated problems with different interfaces, and some equipment only exposes a serial output or a file on a memory card. The failure mode is an interface that works on the machine you tested with and not on the older model in your second lab. Name the specific machines in the contract, and decide what happens when the link fails, because the answer must be manual entry with a flag rather than a stopped lab.

Invoicing is the second. Time and materials billing in this industry depends on actual on site hours including standby and cancellation, and those arrive from the field application rather than from the lab. When the field data and the accounting system disagree about a visit, someone reconciles by hand and the value of the integration disappears. Reconcile daily rather than at month end and alert on visits with no billing outcome.

The third is a client portal, which is not really an integration but behaves like one, because the moment clients can see live status they will notice every gap in your data that was previously invisible. That is a benefit, and it is also a reason to launch the portal after the internal spine is trusted rather than alongside it.

What happens when non conforming results and accreditation evidence are not covered?

These two are deferred more often than any others in this category and they are the two that carry professional risk rather than inconvenience.

A break result below the specified strength must never leave the building as an unremarked number inside a batch of reports, because that is precisely the document that gets read out later in a dispute. The control is a chain: automatic flagging against the mix design specification, a hold on release, review by a qualified person, and a linked investigation or retest. If a system prints a failing result the same way it prints a passing one, the firm's signature is being applied without a decision behind it.

Accreditation evidence has the same shape. It depends on documented equipment calibration and verification, technician qualification, method compliance, proficiency sample participation and internal quality checks, all of which exists but sits scattered across a calibration binder, a training folder, an email chain and one person's memory. Attaching equipment to tests, so every result records which machine performed it and whether its verification was current that day, turns an assessment into a report. The preventive effect matters more: an expired calibration blocks result entry rather than being discovered afterwards on results you already issued.

The same logic applies to dispatch. Model technician certifications with expiry dates and make assignment refuse anyone whose qualification has lapsed, because assigning an uncertified technician is both a quality failure and an accreditation finding.

Should you build custom or configure what you already own?

If you have fewer than about eight technicians, a single lab and work in one or two jurisdictions, buy. QESTLab is a serious laboratory system built for this industry and MetaField covers field to lab workflow credibly. At that size a well configured package beats a build on cost and on time to value, and the difference is better spent on technicians and equipment. Firms in that position who commission a custom system usually end up with something that does the same job later and costs more.

The honest reading of the packaged tools is that they handle sample workflow and laboratory records well. What firms keep improvising around is the intersection of the break calendar with technician and machine capacity, escalation before a break date is missed, and jurisdiction specific special inspection output. Those are operational rather than laboratory problems, which is why they sit in the gap.

Build when two or more of these hold. You operate multiple labs or offices and samples move between them. You work across enough jurisdictions that special inspection forms are a maintained library rather than a template. Dispatch is one person's knowledge and the business stops when they take leave. You have missed a break date or issued a report late enough that a client escalated. Or you want per project and per client profitability including standby, cancellation and travel, and cannot currently produce it.

How do hidden costs get into the quote?

The bands are $65,000 to $145,000 over 12 to 16 weeks for a first release covering offline field capture with chain of custody, lab intake and curing management, the automatic break calendar with escalation, result entry with specification checking and templated reports, and $170,000 to $420,000 over 7 to 12 months for a full platform. Overruns come from four places and all four are foreseeable.

Jurisdiction forms are the largest and the least respected. Each building department's special inspection format is content work, some are fussy about layout and wording, and the final report of special inspections at project close is its own document. A quote that says jurisdiction forms as one line without counting them is guessing. Count them, name them, and price them individually.

Equipment integration is the second, for the reasons above. Multi lab operation is the third, because a sample moving between sites needs its custody chain, curing conditions and scheduling to move with it. The fourth is launching every discipline at once. Firms that insist on concrete, soils, masonry and steel in the first release typically take twice as long and end up with a system technicians distrust in all four.

The cost nobody itemises is your own quality manager's time defining what the system must enforce: which results hold, who reviews, which certifications apply to which tests. That is not developer work and it cannot be skipped.

What separates a build that works from one that fails here?

The builds that work produce one screen every morning that the lab manager actually opens: today's breaks, tomorrow's breaks, anything overdue with the project and client attached, and any sample whose custody chain is incomplete. Missing a break becomes a prevented failure rather than a reported status. The second thing they get right is separating the data from the deliverable completely, so results, observations and non conformances are recorded once and every output is a template rendering. Late reports in this industry are usually not a testing problem. They are a transcription bottleneck at the end of a correct process, and removing it is where the visible improvement comes from.

The builds that fail were chosen without testing the developer on the domain. Ask how a cylinder cast on a jobsite with no mobile signal gets into the lab system, and if the answer does not start with offline capture and a physical identifier created on site, your technicians will not use what they deliver. Ask what happens when a break comes in below the specified strength, and expect an automatic flag, a hold, a qualified reviewer and a linked retest rather than a line in a report. Ask which testing equipment they have integrated with by name and interface, not in general.

Then settle ownership in writing before kickoff. You should own the repository, the cloud accounts and the right to hire anyone else. Your test records are evidence in disputes about structures that will stand for decades, and they should live somewhere you control rather than inside a vendor's environment.