How to choose an LMS: criteria and checklist
Short answer
You find the right learning platform by first defining goals, audiences and must-have requirements, deriving weighted criteria from them and testing no more than three vendors in a real pilot with your own content. What matters most is usability for learners, effort for authors, data protection, integrations and total cost over several years.
Key points
- Requirements first, vendors second: starting with demos means buying features instead of solutions.
- Split criteria into must, should and could, and weight them.
- A pilot with your own content and real learners tells you more than any demo.
- Check data protection, export options and open standards early, not at contract stage.
- Compare total cost over three to five years, not just the licence fee.
Why selection so often goes wrong
Many organisations choose their LMS on the strength of the most impressive demo. Six months later it turns out that authors struggle with the tool, the HR integration is missing or learners get stuck on their phones. The reason is almost always the same: the requirements were not clear before the vendors arrived.
Good selection therefore starts with three questions. What should have measurably improved in a year? Who learns, who creates content, who administers? And which conditions are non-negotiable, such as data location, accessibility or a specific integration?
The criteria catalogue
The following table sorts the most important criteria. Adjust the weighting to your situation: a university weights exams and campus integration more heavily, a company weights assignment and proof.
| Criterion | Question to ask | Typical weight |
|---|---|---|
| Learner experience | Can learners find their way without training, including on a phone? | high |
| Author effort | How long does it take to turn existing material into a course? | high |
| Administration | Can groups, deadlines, reminders and records be managed without manual work? | high |
| Data protection | Where is data stored, who processes it, is there a data processing agreement? | must |
| Integrations | SCORM, LTI, xAPI, single sign-on, API, export? | medium to high |
| Languages | Interface and content in the languages of your staff or students? | depends |
| Accessibility | Which WCAG checks have been done, and what is still open? | must for public bodies |
| Credentials | Are certificates verifiable and tamper-proof? | medium to high |
| Cost | What does the system cost over three to five years including rollout and operation? | high |
Must, should, could: weighting properly
Must criteria are knock-out criteria. If a vendor fails one, it is out, however good the rest is. Typical must criteria are processing of personal data within the EU, a data processing agreement under Article 28 GDPR or an integration you cannot do without.
Should criteria are weighted and scored, for instance on a scale from zero to three. Multiply score by weight and add up. Could criteria are nice extras that only decide a tie. Keep the list short: twenty well-chosen criteria say more than two hundred that nobody seriously checks.
The pilot: the most important phase
No paper comparison replaces a test with your own content. Take two or three real courses, including one that currently exists as a presentation or PDF, and let your own authors build courses from them. Let a small group of real learners take them, on a computer and on a phone.
Measure soberly: How long did authoring take? How many learners completed the course? Which questions reached support? What do the authors say after the first week? Such observations are more reliable than any feature list.
Data protection, exit and total cost
Check early which services the vendor uses and where they operate, especially for AI features such as language models, speech synthesis and speech recognition. Ask for the list of sub-processors and what is sent to each of them.
Exit matters just as much. Ask to be shown how courses, learner data and records are exported and in which formats. A system you cannot get out of becomes more expensive every year. Finally, calculate total cost: licence, rollout, integrations, content creation, consumption of AI features and the internal effort for operation and support.
How Universe does it
Universe can be tried without a contract: a trial course runs without an account, and a trial author account lets you upload your own presentation and have a course built from it; the account disappears after one day. Modules can be booked individually, and exports as JSON, PPTX, PDF and CSV as well as SCORM packages are available.
Step by step
Set goals
Describe what should have measurably improved within a year, such as completion rates for compliance training or time to productivity.
Map roles and volumes
Who learns, who authors, who administers? How many people, languages and sites?
Weight criteria
Split requirements into must, should and could and assign weights.
Shortlist
Check five to eight vendors against the must criteria and invite at most three to a pilot.
Run the pilot
Let your own authors build your own courses and real learners take them; measure time, completion and questions.
Check total cost and exit
Compare costs over several years and ask to see a complete data export.
Frequently asked questions
How many vendors should you compare?
Five to eight for a paper shortlist, at most three for the pilot. More pilots overwhelm your own authors and learners and blur the result.
Who should be involved in the selection?
At least HR or academic leadership as sponsor, authors, a few learners, IT, data protection and, in companies with a works council, the works council. A decision made by IT alone or by procurement alone usually misses how the system will actually be used.
Is a formal tender required?
Public bodies such as state universities are subject to procurement law and must tender above certain thresholds. Private companies are free to choose, but benefit from the same discipline: clear requirements and transparent scoring.
What role does AI play in the selection?
AI can greatly speed up authoring and learner support. Check, however, whether AI output appears as a suggestion, whether content is checked against sources and where the models run.
How it works in Universe
Related articles
- What is an LMS? Learning management systems explainedAn LMS delivers courses, manages learners, tracks progress and records completions. Its structure, functions, types and limits at a glance.
- What does an LMS cost? Pricing models and total costWhich pricing models learning platforms use, which costs come on top of the licence and how to compare total cost over several years honestly.
- Rolling out an LMS: a project plan in seven phasesHow to introduce a learning platform successfully: goals, team, pilot, content, integrations, launch and operation — with project plan, roles and common mistakes.
- Learning platforms and data protection: checking EU hostingWhat data an LMS processes, what the GDPR requires, what EU hosting really means and how to check AI services, data processing and deletion.
- SCORM, LTI, xAPI and cmi5 explained simplyThe key e-learning standards compared: what SCORM, xAPI, cmi5, LTI, QTI and Common Cartridge do and which standard fits which job.