LMS Comparison 2026: 15 Platforms by 110 Criteria

This part describes how the research data was collected, verified and scored. The results themselves are in the overview and in the category-by-category analysis.
How we chose the 15 systems
The sample is not a “top 15” from a single ranking. Rankings are built on different methodologies and select products by the number of reviews, while we needed to compare different approaches to corporate learning: classic SaaS LMS, enterprise platforms, open source, LMS with authoring, collaborative learning, mobile-first and HCM platforms. Rankings and directories (G2, Capterra, Gartner) were used as one signal, but not as the only selection mechanism.
Five criteria were assessed for each candidate:
Market significance
Independent confirmation
Model representativeness
Corporate use
Verifiability
The final rule: verifiability and corporate applicability are mandatory, and on top of them we need either market significance together with independent confirmation, or the role of an important representative of a distinct product model. This is not a mathematical ranking but a rule for forming the sample.
Systems that did not make the first sample (LearnWorlds, Continu, Trainual, Skill Cup, Kampus, Knomary, “ТренингСпэйс”) remain in reserve: either they focus on selling external courses, SOPs or separate models such as LXP, or their approach is already represented in the sample. A product’s absence from the list does not mean it is worse.
How the research was conducted
Criteria. A single Data Dictionary v1.0 was used: 110 criteria in 21 categories — product and deployment, users and org structure, authentication, content and standards, learning and assignments, assessment, knowledge management, user experience, communication, reporting, automation, integrations, security, accessibility, languages, plans and pricing, scale, support, administration, learning methodology, AI. Each criterion has a definition and allowed values; the wording was not changed for individual products. Of the 110 criteria, 96 are scored (they produce points) and another 14 are informational: product type, deployment model, prices, limits, interface languages — their values are given as they are, without points.
Sources. Primary sources take priority: official documentation, help centers, feature, pricing and security pages, API documentation, release notes (Tier 1). Next in descending order: review and analyst sites — G2, Capterra, Gartner (Tier 2), independent reviews (Tier 3) and communities (Tier 4). Independent sources were used as additional confirmation, not as the basis for technical statements.
Verification. Every product went through the first pass — collecting evidence for all criteria. For 14 of 15 products a second pass was also performed: a separate check of 16 critical criteria (pricing, SSO, MFA, LDAP, SCORM, xAPI, mobile app, data region, ISO 27001, SOC 2, AI assistant, API, webhooks) without access to the results of the first pass. That is 16 × 14 = 224 checks; Mirapolis has no second pass. Disagreements were resolved manually: the source date, plan, version and edition of the product were checked.
The Unknown rule. A missing mention is not considered proof of absence. The value “No” was assigned only where the documentation directly states that the feature is absent or where the provider officially announced its retirement (for example, Docebo’s LDAP integration was officially retired on 20 August 2025).
The Partial rule. “Partial” was assigned when the feature exists but is limited: available only on the top plan, requires a separate module, works only in the mobile app, is implemented through a third-party connector instead of a native integration or relies on an outdated architecture. The limitation is always described.
Freshness. The verification date was recorded for every claim. Data on pricing, AI features and limits is the most volatile; it was checked first.
Scores. Functional values are converted to points: Yes = 1, Partial = 0.5, No = 0, n/a = NULL. A category score is the average over the verified criteria of that category, so next to every score the number of confirmed criteria is shown. For example, “83% (5/6)” means: 83% of the maximum over the criteria for which there is evidence — and that is five criteria out of six.
What the research does not have. We do not build a single overall ranking and do not name a winner. First, category weights are not defined in advance as an industry standard: any single number would reflect our priority model, not the properties of a product. Second, the products belong to different classes, and the sum of scores between them is only partly comparable. Third, for some systems public documentation is incomplete — a single score would punish them for being closed, not for functionality.
Equal conditions. Brusnika.LMS is one of the 15 systems in the sample and was assessed against the same criteria dictionary and by the same rules as the rest. Where there is no public confirmation (SSO, MFA, certificates), the tables show “n/a”, as for other systems.
The path of a single fact
The unit of research is “product × criterion × point in time”, for example Moodle × SCORM × 24.10.2026. The researcher does not collect “everything about Moodle” but checks one claim and first finds the evidence, then assigns the value — this lowers the risk of confirming what one wanted to see.
Search by criterion
The product’s official documentation, website and help center, then independent sources and a separate search for refutations.
Evidence
A short excerpt of the primary source is saved: one piece of evidence confirms one claim, neighboring criteria are checked separately.
Verification
Confirmation and refutation: whether the feature works only on a plan, in a separate module or in an outdated version.
Conflicts
A contradiction is not resolved by the researcher’s choice: the fresher official source wins, and a disputed value stays “n/a”.
Audit and publication
The value gets a confidence level and a verification date and enters the matrix, while scores are calculated by rules published in advance.
Source independence. The number of sources does not by itself mean the quality of evidence: ten sites retelling a vendor’s page are one source, not ten confirmations.
Shelf life of data. Data is divided into current (pricing, integrations, features, AI, certificates — they change) and historical. Each category has its own recommended re-verification period; an expired value gets the status “needs re-verification” rather than being automatically considered wrong.
| Data category | Time to re-verification |
|---|---|
| Pricing, free plan and limits, AI features | 30 days |
| Integrations | 90 days |
| Product features, certificates, deployment, interface languages | 180 days |
| General product description | 365 days |
Besides the calendar period, data is re-verified after significant events: a major release, a price change, an acquisition or rebranding, a new AI module, a change of deployment model.
What a finished record with a quote, a link, a date and a confidence level looks like is shown in the part “Criteria and evidence”.
Data quality
The indicators below describe the whole research matrix: 15 products × 110 criteria.
| Metric | Value | Comment |
|---|---|---|
| Total cells in the matrix | 1650 | 15 products × 110 criteria |
| Cells with a confirmed value (Yes/Partial/No) | 1134 | across the 96 scored criteria |
| Evidence coverage (scored criteria) | 78,8% | share of verified values out of 96×15 |
| Cells with Unknown (all 110 criteria) | 346 | including informational criteria |
| Products with independent verification (2nd pass) | 14 | out of 15; Mirapolis has no second pass |
| Critical criteria checked in the second pass | 16 × 14 = 224 | pricing, SSO, MFA, LDAP, SCORM, xAPI, mobile, residency, ISO, SOC 2, AI, API, webhooks |
| Disagreements between passes | 54 | resolved manually |
| Conflict rate (of verified cells) | 4,8% | disagreements between passes not resolved automatically |
| Stale data rate | 0% | all checks were done on 24 Oct 2026 |
| Evidence records | 1304 | rows in the evidence base |
| Tier 1 (official) sources | 1300 of 1301 | the rest are independent reviews |
| Claims with high confidence | 1063 | direct primary source |
| Claims with low confidence | 6 | used only as a supplement |
Coverage is calculated only over the 96 scored criteria: 1,134 confirmed values out of 1,440 possible (15 × 96) — that is the 78.8%. The value “n/a” (Unknown) appears in 346 cells of the matrix, including informational criteria. It means “no public evidence found”, not “the feature does not exist”.
Where the two passes disagreed
There were 54 discrepancies between the passes; the discrepancy rate (Conflict Rate) is 4.8% of verified cells. All discrepancies were reviewed manually, and in all 54 cases the final value remained what the first pass gave. Most often (22 cases) the first pass found no evidence (“n/a”) while the second stated the feature is absent:
| First pass | Second pass | Final value | Cases |
|---|---|---|---|
| n/a | No | n/a | 22 |
| Partial | Yes | Partial | 14 |
| n/a | Partial | n/a | 8 |
| Yes | Partial | Yes | 4 |
| No | Partial | No | 2 |
| Partial | No | Partial | 3 |
| n/a | Yes | n/a | 1 |
Illustrative cases:
- Brusnika.LMS — SSO. The first pass found no evidence (“n/a”), the second gave “Partial”. Result: “n/a” — there is no confirmation in public sources.
- Docebo — LDAP / Active Directory. First pass: “No”, second: “Partial”. Result: “No” — the LDAP integration was retired on 20 August 2025, and Active Directory connects through SAML SSO.
- Brusnika.LMS — mobile app. First pass: “Partial”, second: “No”. Result: “Partial” — there is no native app in the App Store and Google Play, but the LMS installs as a PWA.
- Moodle — AI assistant. First pass: “Partial”, second: “Yes”. Result: “Partial” — the core has a course helper but not a free question-answering chat.
Most discrepancies fall on MFA (6), xAPI (6), webhooks (6), data storage region (6), ISO 27001 (5), SOC 2 (5) and the AI assistant (5): these are the criteria on which the two passes disagreed most often.
Limitations
- G2 and Capterra are closed to automated access (HTTP 403), so user reviews were not used in the research.
- The help centers of some vendors are also unavailable for automated access.
- Coverage differs between systems: from 93.8% for Absorb LMS to 64.6% for Websoft HCM. This reflects the openness of documentation, not product quality.
- Data on pricing, AI features and limits changes faster than the rest; the snapshot was taken on 24 October 2026, and these sections become outdated first.
- There is no single ranking on purpose: category weights depend on a company’s tasks, not on product properties.
How to verify the result
Every value in the matrix has a source URL, a short evidence excerpt, a source type (official documentation, help center, pricing page, security page, etc.), a confidence level and a verification date — 24 October 2026. In total the research database holds 1,304 evidence records, 1,303 of them from official sources; 1,063 are marked as direct primary sources (High confidence), and 6 are Low confidence and used only as a supplement. There is no stale data (Stale Data Rate): all checks were done on a single day.
What such records look like is shown in the “Criteria and evidence” part: it gives examples with a quote, a link and a date. Links to primary sources for each vendor are collected in the “Appendix and sources” part.




