How to Evaluate NDT Software Without Being Sold To

Scope the decision before you look at a single vendor

Two numbers eliminate most of the market. How many technicians you dispatch, and how many methods you run. A five-person shop running MT and PT for two fabricators has a different problem from a sixty-person contractor running RT, UT, phased array and rope access across four states with union rates. Write down technicians, methods, client count, states worked, whether you subcontract, and whether any client contract restricts where their data may be stored. That single page of facts does more filtering than three weeks of demos.

Then name the decision you are actually making, because three different projects get answered with the same demo. Replacing spreadsheets is one. Replacing a working system that is supported or priced badly is another. Adding a field capture layer to an office system that already works is a third. They carry different risks, different rollout orders and different definitions of success, and a vendor cannot tailor anything if you have not said which one you are running.

Finally decide what stays. Your accountant's ledger usually stays. Your report formats usually stay, because clients approved them. Your instrument software always stays, because it drives the hardware. Everything you keep becomes either an integration or a boundary, and both are work that belongs in the plan rather than in month two. The NDT software category page separates the four products that share the name, so your shortlist covers one category instead of three.

The eight criteria that decide the purchase

Offline field capture, certification enforcement, calibration enforcement, report templates by code, accounting integration, client portal, deployment model, and full data export. Every other feature on a comparison grid is a preference. These eight are the ones that, when missing, force a workaround, and the workaround grows into a second system — which is how contractors end up running a purpose-built platform alongside the spreadsheets it was bought to retire.

Weight them against your own facts rather than a generic template. A contractor working Texas refineries with three field crews weights offline capture and calibration enforcement at the top. A contractor doing shop inspection for fabricators weights report templates and the client portal. A contractor whose controller closes the month in four days weights accounting integration above both. The eight criteria are constant across the industry; the order is specific to your operation and nobody else's.

Score them on one sheet, a column per vendor, filled in from the trial rather than from the demo. The RFP requirements checklist is the long-form version, written so it goes out unchanged to every vendor on your list. Sending an identical document to everyone is what makes the responses comparable at all, and a vendor who answers a different question from the one asked has told you something about how the implementation will run.

Offline capture is the criterion nobody tests properly

Almost every vendor says the app works offline, and the claim is true and incomplete. What matters is reconnection behavior: two technicians edited the same job on two tablets, one attached forty photographs, one lost signal for six hours, and the sync ran while a coordinator was editing the same work order in the office. The product is what happens in that scenario, and no demo will ever show it to you unprompted.

So test it directly. Airplane mode, complete a full job with photographs and sketches, keep the device offline for four hours, have a second device edit the same job, then reconnect both. Acceptable outcomes are one merged record, or an explicit conflict the user resolves. Unacceptable outcomes are a silently overwritten record, a duplicate job, or photographs that quietly vanished. Run this before anything else, because a failure here is not fixable by configuration.

Ask separately about payload size. A tablet that syncs text fine but stalls on a hundred-megabyte scan file or a dozen high-resolution photographs will be used for text while the images get emailed, which puts your evidence back outside the system where it started. Confirm that an interrupted upload resumes rather than restarting. A crew on hotel wifi at the end of a shift is the real test environment, not your office.

Certification and calibration: enforcement, not reporting

Every vendor will show you a dashboard of expiring certifications, and dashboards are not the requirement. The requirement is that the system refuses an action. Ask the demonstrator, live, to assign a technician whose method certification expires two days into a five-day job, and watch what the software does. A warning that can be clicked through is a warning your coordinator will click through at six in the morning during a turnaround, every single time.

Run the same test on equipment. Allocate an instrument whose calibration certificate expires before the job ends, then one whose wedge certificate has lapsed while the detector's is still current — the second test catches products that model the instrument and ignore its accessories. Ask whether an override exists, who is permitted to use it, and whether it is recorded: a documented, role-restricted override is a legitimate operational tool, an undocumented one is a hole in your audit trail. Inspection management software sets out both constraints and the points they must fire at.

The maintenance load is the honest half of this conversation. Enforcement is only as good as the registers behind it, and somebody in your company will keep certifications, vision tests, calibration certificates and procedure revisions current in the system every month. Ask each vendor exactly how that data gets in and stays in — bulk import, scanned certificate with an extracted date, supplier feed, manual entry — and convert the answer into hours per month before you compare anything else.

Report templates, acceptance criteria, and who can change them

The report format is the part of the purchase your clients notice. Bring your most awkward client's format to the trial — the one with the unusual sketch, the client logo in the wrong corner, the acceptance table nobody else uses — and rebuild it yourself. Not the vendor, and not their implementation consultant. If your own staff cannot reproduce it in a day, every future client format becomes a support ticket with a lead time attached to it.

Acceptance criteria are the second half. A report is defensible because a thickness reading was compared against t-min, or an indication was measured against the applicable ASME, API or AWS criterion, and the comparison was made by a rule rather than by a technician's memory at the end of a twelve-hour shift. Ask who may edit those rules. The correct answer names your Level III, with a change record. The reporting software comparison covers the template and criteria models across products.

Then break it deliberately. Issue a report, correct a value, reissue, and ask for both versions, the difference between them, the authorizer and the timestamps. This is the test that most often separates products and it takes four minutes. A system that cannot hand you the superseded document has quietly made your quality manager personally responsible for a filing system. Repeat the test for a report that was delivered to a client and then withdrawn.

Accounting integration is a boundary decision

Decide first where the ledger lives. Most US inspection contractors keep QuickBooks or Sage because the bookkeeper and the CPA already work there, which makes the inspection system the system of record for jobs and the accounting system the system of record for money. The integration then has exactly one job: push invoices and their line detail out, pull payment status back, and never let the two systems disagree about what was billed to whom.

Test the round trip on real data rather than accepting a feature list. Complete one job, generate the invoice from the work order, push it, and confirm it lands under the right customer with the right tax treatment and the right revenue account in your own chart of accounts. Then mark it paid on the accounting side and confirm the job reflects as paid. QuickBooks integration covers the field mapping and both directions of that flow.

The alternative is moving the ledger into the same system, which removes the integration entirely and adds a migration plus a retrained bookkeeper. That trade is worth making when the operational and financial records are already tangled: job costing that requires a monthly export, revenue recognition on partially delivered campaigns, subcontractor costs that never reach the job they belong to. The ERP overview describes what the combined model looks like in practice.

Deployment, data ownership and getting out

Cloud, on-premise or private hosting is not a preference question in this industry. It is a contract question. Nuclear, defense and some refinery and pipeline clients write data residency and network restrictions into their terms, and a cloud-only vendor removes you from those bids without anybody in the sales process mentioning it. Ask for the on-premise option in writing during evaluation, not after a client's procurement team raises it during a bid.

Data ownership resolves at exit, and exit terms get negotiated at signing or never. The clause you want says your data is yours, exportable in full, in an open format, with relationships intact, on request, at no charge, within a defined number of days, including after termination for any reason. Confirm it in the contract language and confirm it actually works by requesting a complete export, unannounced, in the middle of the trial.

Ask what happens to your data during a billing dispute, where backups live, and how a restore is requested and how long it takes. These are unglamorous questions with short answers, and a vendor who has answered them before answers immediately. Atlantis deploys on Odoo, in the cloud or on your own infrastructure, so a residency requirement in a client contract is a deployment choice rather than a roadmap item.

How to run a trial that reveals problems

A demo is a performance and a trial is an experiment, so design the experiment before it starts. Pick one real job, ideally an awkward one with a subcontractor, a reschedule and a report revision, and run it end to end in the candidate system in parallel with your normal process. One complete job beats fifty partial ones, because only a complete job exercises the joins between scheduling, field capture, reporting and invoicing — and the joins are where products fail.

Give three people three separate tests rather than one shared walkthrough. Your Level III edits an acceptance criterion and reissues a report. Your coordinator reschedules a crew into a certification conflict and confirms the system refuses rather than warns. Your controller runs the job through to an invoice and into the ledger. Each reports independently, in writing. A product that satisfies one of the three and not the others tells you precisely which department will abandon it in month four.

Set a fixed two-week window with a written pass list agreed in advance, and refuse to extend it. Extensions get requested when the answer is already no and nobody wants to say so. Record how long the vendor takes to answer support questions during the trial, because that is the fastest their support will ever be — you are the most interesting prospect they have this month. The tools section collects the calculators and trackers we publish for this kind of comparison.

The questions vendors dislike

Six questions change the temperature of a sales conversation. Can I have a sandbox with my own data for two weeks? Can you export everything, right now, in an open format? Which NDT contractors use this, and may I call two of them without you on the line? What is on the roadmap that is not built yet, and what did you actually ship in the last twelve months? Who performs the implementation, your staff or a partner? What does support cost after year one?

None of these are hostile and all of them are answerable. The reference call is the highest-yield of the six: ask another contractor what broke in month three, what they configured themselves versus paid for, and what they still keep in a spreadsheet. That last answer maps directly onto the gaps you will inherit. Two calls take an hour and tell you more than the rest of the evaluation combined.

Ask about the implementation team specifically. Software configured to a written practice — certification rules, numbering schemes, acceptance criteria, rate schedules — depends on somebody who understands both the product and the industry it serves. Find out who that person is, whether they are available in your timeline, and how many inspection implementations they have run. Talk to us about a demo, a sandbox, or a quote on request.

Build versus buy

Building looks attractive when every product on the shortlist misses something specific to how you work. The honest comparison is not build cost against license cost. It is the ten-year cost of owning a codebase against the ten-year cost of a vendor relationship. Building means you own the mobile app, the offline sync engine, the permissions model, the report renderer, the accounting integration, the upgrade treadmill, and the developer who eventually takes another job.

Offline sync is where most in-house builds stop. Conflict resolution across multiple devices, with large binary attachments, over intermittent connectivity, is a genuinely hard distributed-systems problem — and it is the one feature an inspection business cannot operate without. A team that has not built it before will spend considerably longer on it than the estimate, and the failure mode is not a delayed release. It is a shift of field data that no longer exists.

The middle path is what most contractors actually want: a configurable platform they can extend themselves. That is the argument for an open, modular base — you configure registers, numbering schemes and rate schedules yourself, and write a module when something is genuinely specific to your operation, without owning the sync engine or the framework underneath it. Build versus buy works the comparison through in detail.

Request a consultation