Types of LIMS: There Are Types of Labs, Not Types of LIMS
Every vendor list of LIMS types is a list of the vendor's own products. The useful cut is by what the lab does with a sample, and it changes which questions matter in a demo.
In brief
The types of LIMS are listed by vendor packaging, which tells a buyer nothing. The useful cut is by what the laboratory tracks: testing labs organised around a sample, clinical labs around a patient, research groups around an experiment, manufacturing QC around a batch, and oil labs around a machine. Each needs different words, forms and reports.
Key takeaways
- Ask what the unit of record is in your lab before asking what type of LIMS you need. The answer places you in seconds.
- A clinical lab is organised around a patient, not a sample, which is why an LIS is not a type of LIMS.
- Deployment is an administration decision made after the type is settled. Cloud or on-premises is not a type.
- A mixed lab needs one platform that uses different words per group, not one product per group and a spreadsheet.
- The demo question that separates types: show me the report my customer receives, from your system, in my words.
Search for the types of LIMS and you will find lists of four, five, six and seven, and no two agree. One divides by deployment: cloud, hosted, on-premises. Another by industry: pharmaceutical, environmental, clinical, food. A third by vendor packaging: enterprise, mid-market, starter. Every one of those lists is written by someone selling one of the entries.
This article offers a different cut, and it is the one that predicts which product will fit. Laboratories differ in what they track. A testing lab tracks a sample. A clinical lab tracks a patient. A research group tracks an experiment. A manufacturing QC lab tracks a batch. A used oil lab tracks a machine. That thing, the unit of record, decides what the lab calls its work, what it reports and what it must never mix up. Choose a LIMS by that, and the rest of the list stops mattering.
Why do lists of LIMS types disagree with each other?#
Because they are describing how software is sold rather than how laboratories work. A vendor with three editions publishes a list of three types. A vendor with an industry-specific product line publishes a list of industries. A cloud vendor publishes a list that makes deployment the first question, because that is the question they win. None of the lists is dishonest. Each is a catalogue.
The result is a buyer’s problem. A lab manager who reads four lists comes away believing there are a dozen types and that the lab must be one of them, when the real question is much smaller. What does your lab track, and what does it call that thing? A metrology lab runs calibrations. A clinical lab has samples and patients. A training team has courses and people. A system that uses those words back, on screen and in reports, is one the lab will use. A system that makes a calibration lab call its work a sample is one the lab will work around.
So the cut below is by the thing the lab tracks. It produces five types, and the number is not the point. The point is that a lab can place itself in about thirty seconds, and that once placed, it knows which questions to ask in a demonstration and which to ignore.
What is the unit of record in your lab?#
The unit of record is the thing every other record attaches to. In a testing lab it is the sample: results, chain of custody, holding times and reports all hang off a sample identifier, and the customer receives a report about samples. In a clinical lab it is the patient: the same test on the same sample means something different depending on whose it is, and the obligations follow the person. That is why a laboratory information system is not a type of LIMS, and one sentence is all that argument needs here.
Five units of record cover nearly every laboratory, and the table shows what follows from each.
| Unit of record | Labs organised this way | What the lab calls its work | What it reports | What it must never mix |
|---|---|---|---|---|
| The sample | Environmental, food, water, analytical and contract testing | Samples, batches of samples, methods | A certificate of analysis or test report per sample or job | Two customers’ samples, or a result with the wrong holding-time clock |
| The patient | Clinical and diagnostic laboratories | Patients, specimens, orders | A result to a clinician, against a reference range for that patient | Two patients’ specimens, ever |
| The experiment | Research, biotech, method development | Experiments, protocols, materials | A write-up, a figure, a conclusion | Which material lot was used in which experiment |
| The batch | Manufacturing QC, stability, release testing | Batches, lots, specifications, timepoints | A release decision and a stability summary against a specification | Release limits with shelf-life limits, or one batch’s trend with another’s |
| The machine | Used oil, transformer oil, condition monitoring | Compartments, components, sampling history | A diagnosis and a recommended action for the machine | Readings from different compartments of the same machine |
Read down the last column and the buying question writes itself. A product built around one unit of record makes the mistakes in the other rows easy, because its structure does not know they are mistakes.
Which labs are organised around the sample, and which are not?#
Most of what is sold as a LIMS was built for the first row. Environmental, food, water and contract testing labs receive samples from outside, run methods on them, and send a report back to whoever sent the sample. The sample is the natural spine, and a sample-centred system does this well: chain of custody, holding times, batching for the instrument, results against limits, a report per job.
The other four rows are the ones where a sample-centred system fits badly, and each fails in its own way. In a clinical lab, a patient has many specimens over years, and the meaning of a result depends on the person’s history, so a system that starts from the specimen makes the patient an afterthought. In research, the experiment is the record and the sample is one of many materials that went into it; forcing the work through sample login makes scientists keep a second notebook. In manufacturing QC, a stability programme is a calendar of timepoints across batches, packs and conditions, known in full the day a study opens, and a sample-centred system discovers each timepoint one at a time.
The fifth row is the one most lists leave out. Used oil analysis labs do not analyse samples, they analyse machines. The same iron reading is normal in an engine and critical in a gearbox on the same vehicle, and the value the lab sells is the diagnosis at the end of the report, read across the machine’s sampling history. A system that treats each sample as a fresh event cannot produce that report.
Does deployment make a different type?#
No. Cloud, hosted and on-premises describe who runs the servers and who administers the system. They say nothing about what the lab tracks, what it calls its work or what it reports, so they cannot make a sample-centred product fit a batch-centred lab. A cloud LIMS built for testing labs is a testing LIMS that happens to be in the cloud.
Deployment is a real decision, and it deserves its own analysis: who patches the system, who holds the backups, whether a validated tenancy is available, whether an air-gapped installation is required by policy. But it is an administration decision, made after the type is settled, and the cloud or on-premises question has its own article on this site. Treating it as a type puts the least important question first.
The same is true of “industry-specific” as a type. A food LIMS and an environmental LIMS are both sample-centred testing systems with different method libraries, limit tables and report templates. The industry is a configuration of the type, not a type of its own. A vendor that cannot show you a food lab and a water lab on the same product, with different words and different reports, is selling two products with one name.
How should a mixed lab choose?#
Most laboratories of any size are mixed. A contract lab has a sample-centred testing operation and a batch-centred stability service. A hospital has a patient-centred diagnostic lab and an experiment-centred research group two floors up. A mining company has a sample-centred assay lab and a machine-centred oil lab in the same building. Buying one product per row ends in five systems, five logins and a reconciling spreadsheet.
The alternative is one platform that uses different words, forms and reports for different groups, on the same installation. The stability group sees batches and timepoints. The oil lab sees compartments and diagnoses. The research group sees experiments and materials. Each group’s applications are switched on for that group and not for the others, so nobody is made to call their work by someone else’s name, and the numbers that management wants across all of them come from one place.
The demonstration question that separates types, and separates a real platform from a relabelled one, is short: show me the report my customer receives. A testing lab’s customer receives a certificate of analysis. A clinician receives a result against a reference range. A plant manager receives a diagnosis for a machine. If the vendor can show all three from one installation, with the right words on each, the lab has found a platform. If every answer is the same report with a different logo, it has found a testing LIMS.
- Name the unit of record for each group in the lab before reading any vendor’s list
- Ask the vendor to use your words on screen: calibration, patient, batch, compartment, not sample for everything
- Ask to see the report each group’s customer receives, produced from the demo system
- Ask whether two groups on one installation can have different forms, fields and applications
- Settle deployment last, as an administration decision, once the type fits
What one platform carries across these types#
LabLynx has shipped sample-centred LIMS to testing labs for decades, and the clinical research and used oil analysis work taught us where a sample-centred system stops fitting. The five rows above are the lesson, and the platform is the answer to it.
Frequently Asked Questions #
How many types of LIMS are there?
Cut by what the laboratory tracks, five: systems organised around the sample, the patient, the experiment, the batch and the machine. Vendor lists give other numbers because they count editions, industries or deployment options, which describe how software is sold rather than how a lab works. Most laboratories can place themselves in one or two of the five rows in under a minute.
Is an LIS simply a type of LIMS?
No. A laboratory information system is organised around the patient, and the regulations that govern clinical laboratories attach obligations to the person whose specimen is tested. A LIMS is organised around the sample. The two share instruments and vocabulary but differ in structure, reporting and regulation. A lab that needs one rarely does well with the other.
What type of LIMS does a research lab need?
Usually an experiment-centred system, which most people would call an electronic notebook, with sample and material tracking attached to it rather than the other way round. Research work changes method with every run and the write-up is the product. A sample-centred LIMS forces scientists through sample login for work that is not about samples, and they respond by keeping a second record elsewhere.
Is a cloud LIMS a different type?
No. Cloud, hosted and on-premises describe who runs and administers the servers, not what the system tracks or reports. A cloud LIMS built for testing labs is a testing LIMS. Deployment is a real decision about patching, backups, validation and data sovereignty, and it deserves analysis, but it comes after the type has been settled and does not change it.
Can one LIMS serve two different kinds of lab?
It can if it is a platform rather than a product: one installation on which different working groups have their own vocabulary, forms, fields and applications, switched on per group. It cannot if it is a single sample-centred product with a configurable label. The test is whether two groups can see different screens and produce different customer reports from one system.
Sources and references
- ASTM E1578 Standard Guide for Laboratory Informatics ASTM International
- 42 CFR Part 493 Laboratory Requirements (CLIA) Electronic Code of Federal Regulations
