Buying guide, Platform
How to choose a LIMS: questions to ask before you buy
- Author
- PureField
- Published
- Updated
When choosing a LIMS, first check whether the software enforces the ISO/IEC 17025 record rules itself: can the person who entered a result approve it, is the e-signature bound to the record, and is every change logged with its reason? Then look at measurement uncertainty, internal QC, instrument connections, deployment and the AI assistant's permissions.
When does a laboratory need a LIMS?
The need usually shows up once records are spread across too many places. Sample receipt adds a row to a spreadsheet, results are copied by hand from instrument printouts, reports are built from a Word template. Before an assessment, the quality manager gathers calibration certificates, training records and CAPA forms from folders.
For a laboratory preparing for accreditation, extending its scope or opening a second site, that spread becomes visible at assessment. ISO/IEC 17025 also requires the laboratory to validate the functionality of its LIMS before introducing it (§7.11.2). So selection starts with asking vendors the right questions.
Which questions cover record integrity?
The first questions are about whether the software enforces record rules itself rather than leaving them to users. A rule written only in a procedure may not be enough evidence at assessment; when the rule is locked in the software, you can show it on screen.
| Ask the vendor | Why it matters | ISO/IEC 17025 |
|---|---|---|
| Can the person who entered a result also approve it? | Results must be reviewed and authorised before release | §7.8.1.1 |
| Is the e-signature bound to the record as it stood? What happens to it if the record changes later? | It must be clear who checked which data | §7.5.1 |
| Are changes logged with old and new value, reason and server time? | Amendments to technical records must be traceable to the original | §7.5.2, §8.4 |
| When an approved result is corrected, is the original kept? | Both original and amended data must be retained | §7.5.2 |
| Is approval authority granted per method, to named people? | Authorisation is given for specific activities | §6.2.6 |
| Can an instrument past its calibration date be added to an analysis? | Equipment must be verified before use | §6.4 |
What should you check in measurement and quality control?
The second group asks whether the calculations behind a reliable result run inside the software. If they live in a separate spreadsheet, the laboratory has to verify those formulas itself.
- Which approach is used for measurement uncertainty (Nordtest, GUM), and does it carry through to the report?
- Does the decision rule for conformity statements (for example ILAC-G8) take uncertainty into account?
- Are control charts drawn automatically, with Westgard rules, EWMA or CUSUM alerts?
- What happens to result approval while a method is out of control?
- Are proficiency testing scores (z, z′, ζ, En) calculated? Can results be exported in the format your accreditation body asks for?
- Are your accreditation body's reference and mark rules applied to reports? Can a customer verify a report is valid?
What should you check in instrument, field and customer connections?
The third group is about how data enters and leaves the laboratory. Every value typed by hand is a chance for a transcription error.
- Which balances and instruments connect directly? Is the raw instrument output stored unchanged?
- How are values entered from an instrument with no connection?
- Can a sample collected in the field be recorded without an internet connection?
- Can customers follow their samples and reports in a portal, with logins and downloads recorded?
- Is there an API for passing data to your own systems?
- Does the software run in the cloud or on your own server? If cloud, where is the data centre?
- In a multi-site setup, can permissions and sample numbering be managed per site?
Which questions should you ask about the AI assistant?
The first question for an AI assistant is whose permissions it reads records with. If the assistant can see records the signed-in person cannot, details such as customer identity or staff files can reach the wrong person.
- Does the assistant read records with the signed-in person's permissions?
- Who creates records, approves and signs? How does an assistant proposal become a record?
- Are results such as audit rehearsal grades and scores calculated by code or by the model?
- Does each answer show which records and which document page it relies on?
- Which model providers are supported? Can the laboratory connect its own key?
- Does the assistant run in the cloud, on your own server, or both?
What should you try in a demo?
Run your own work through the demo instead of watching the vendor's prepared scenario. These five tests show the answers to the questions above on screen:
- Take a sample close to your own sample types from receipt to a signed report.
- Try to approve a result with the same user who entered it.
- Correct an approved result and look at what the audit trail records.
- Try to add an instrument past its calibration date to an analysis.
- Ask the assistant for a management review agenda and an audit rehearsal, then check which records the answer relies on.
How does PureField LIMS answer these questions?
PureField LIMS is a laboratory information management system for testing, calibration and medical laboratories that works in line with ISO/IEC 17025 and ISO 15189. Rules are enforced in the database, so the same rule applies whichever screen a change comes from.
| Question | PureField LIMS |
|---|---|
| Four eyes | The person who approves a result is always different from the person who entered it; the rule is locked |
| E-signature | Review, approval and report release ask for the password again; the signature is bound to the record as it stood and becomes invalid if the record changes |
| Audit trail | Old and new value, reason and server time; entries are linked in a SHA-256 chain that you can verify on screen |
| Corrections | An approved result is superseded with a reason and a signature; the original record is kept |
| Instruments | Only instruments in service with valid calibration can be added to an analysis |
| Uncertainty and decision rule | Nordtest and GUM; conformity decisions account for uncertainty under ILAC-G8 |
| Internal QC | Westgard, EWMA, CUSUM; when a method is out of control, approval opens once the violation has been assessed |
| Proficiency testing | z, z′, ζ and En scores; the F704-001 list of TÜRKAK (Türkiye's accreditation body) as CSV |
| Reports | Built from approved results and e-signed; TÜRKAK reference and mark rules applied automatically; a QR code opens a verification page |
| Instrument connections | An edge agent connects Mettler Toledo (MT-SICS) and Sartorius (SBI) balances and instruments with CSV output, and stores the raw output unchanged; values from an unconnected instrument are read from a photo of its display and the analyst checks each against the photo before saving |
| Field and customers | Android and iOS field app records offline; in the customer portal, logins and downloads go to the audit trail; read-only REST API (OpenAPI 3.1) |
| Deployment | Cloud: Google Cloud, Netherlands (europe-west4). Self-hosted: Linux (Docker) or Windows Server, offline |
| AI assistant | Reads with the signed-in person's permissions; records, approval and signature stay with authorised staff; grades and scores are calculated by code; runs in the cloud edition, with your own key as an option (Anthropic, OpenAI, Google Gemini, Azure OpenAI, OpenAI-compatible) |
To run this list with your own sample types, request a PureField LIMS account.