The head of learning and development at a 3,000-person company has two proposals on her desk. The first is for “a modern, fully custom e-learning platform”: eight months of work, a budget in the high six figures in euros, and mock-ups that look excellent. The second comes from the agency that has hosted their Moodle for years: four more plugins, a new theme, about forty thousand euros. Both promise to fix the problem the team has wrestled with for two years. Employees do not log in. Line managers have no idea who is overdue on mandatory training. The compliance report for the auditors is assembled by hand in a spreadsheet every quarter.
Neither proposal starts by asking why people do not log in, and the answer decides which proposal makes sense. If the reason is that training means a separate login, leaving the tool you work in, and hunting for a course in a catalogue of four hundred items, a new theme changes nothing. But you do not necessarily have to build everything from scratch either.
This article is for HR and L&D leaders facing that decision, and for the IT colleagues who will help them make it. It defends one position many vendors will dispute: Moodle rarely stops being enough because of missing features, and a fully custom LMS pays off in only two situations. For most organisations the best answer sits in between: a standard core that tracks progress and produces compliance reports, plus your own experience and integration layer, connected to it through open standards. We show how to tell which option fits, which standards and regulations to plan for, and what each option is likely to cost.
When does Moodle really stop being enough?
Moodle is a mature open-source system. Its code is licensed under the GNU GPL version 3 or later, and the project ships two major releases a year. According to the schedule published on moodledev.io, version 4.5 from October 2024 is a long-term support release with security fixes until 4 October 2027, version 5.0 shipped on 14 April 2025 and 5.1 in October 2025. Around it sits a large ecosystem of plugins, themes and hosting partners. For a typical corporate training team that needs a course catalogue, learning paths, quizzes, certificates and reports, features are rarely what is missing.
The real signals that Moodle has stopped being enough look different, and it is worth recognising them before anyone proposes a rebuild.
Learning has to happen where work happens. A salesperson should get a three-minute lesson about a new product inside the CRM, at the moment they open a quote. A warehouse operator should confirm a safety instruction on the terminal at their station. Moodle is a standalone application with its own interface, and every attempt to squeeze it into someone else’s interface ends in compromise.
The platform is a product, not an internal tool. You sell courses to customers, train a partner or reseller network, or run a certification academy for users of your software. Payments, multiple brands, separate spaces for each customer organisation, content licensing and an experience that has to compete with consumer learning apps all start to matter.
Core modifications are eating the upgrade budget. If someone spent years changing core code instead of writing plugins, every version upgrade becomes a separate project. The organisation then stays on an old version, and the cost of deferred upgrades grows every six months.
Scale or data demand a different architecture. Tens of thousands of concurrent users in short peaks, behavioural analytics at the level of single interactions, a requirement to keep data in a specific region. These are still rarely reasons to abandon Moodle, but they can be reasons to change how it is hosted and who owns which part.
If none of these signals is present and the problem is low engagement and manual reporting, you probably do not need a new system. You need better configuration, integration and content.
Why is “we are missing features” usually the wrong diagnosis?
Because the list of missing features is usually written by comparing your platform with someone else’s demo, not by analysing why your platform does not work. The training team sees a polished dashboard with recommendations, gamification and a mobile app, and concludes that this is what they lack. Most of those elements can be added to the existing system or replaced with something simpler.
Try a simple test. Take the three most common complaints from employees and the three most common complaints from managers, and for each one write down whether its root cause is a missing feature, poor configuration, a missing integration or weak content. Very often “I cannot find the course” is a catalogue structure problem, “I have to log in separately” is missing single sign-on, “I do not know who is overdue” is missing synchronisation with the org structure in the HR system, and “the courses are boring” is a content problem no platform can solve.
The second source of misdiagnosis is confusing the system with the content. A short, well-designed course in an unattractive LMS will do better than a weak course in a beautiful one. Budget moved from building a platform to producing and refreshing content often delivers more than a new system would.
The third is the absence of an owner. A platform that “IT maintains” and “HR uses” often has nobody accountable for whether employees actually use it. A new system does not solve an organisational problem. It moves it to a new URL.
Only when the needs left after this test cannot be met with configuration, plugins and integration does it make sense to discuss building. Even then there is a middle path worth weighing before committing to a system from scratch.
What are the three options: extend, build a layer, or build an LMS from scratch?
Reduce the decision to three options rather than two extremes. We covered the general logic of this choice in our build vs buy software decision framework; here we apply it to learning platforms.
Option A: extend the existing LMS. You stay on Moodle (or another standard system), clean up the configuration, add plugins, write missing modules as plugins rather than core modifications, and implement single sign-on and synchronisation with the HR system. You own only the add-ons, and core upgrades come from the community.
Option B: standard core plus your own layer. The LMS becomes an engine: it holds course structures, launches packages, tracks progress, issues certificates and produces compliance reports. On top of it you build your own layer: a portal designed for your organisation, widgets embedded in the CRM, intranet or mobile app, role-based recommendations, a dashboard for line managers. The layer talks to the core through APIs and open standards (covered below). The build scope is markedly smaller than a full system, and the most tedious and most regulated parts, namely tracking, reporting and package handling, stay in proven software.
Option C: your own LMS from scratch. You build the whole system: the data model for courses and users, SCORM package handling and xAPI statements, reporting, certificates, permissions, admin console, localisation, notifications and integrations. You get full control and full responsibility, including for everything the Moodle community does for you today.
Most proposals on the market frame this as A versus C, because vendors specialise in one or the other. A Moodle hosting partner will propose more plugins; a software house will propose a build. Option B asks more architectural awareness of the buyer, but in many cases delivers the most value per euro, because the money goes into what employees see and use, not into rewriting mechanisms that already exist.
When does a custom LMS built from scratch make sense?
The thesis of this article does not mean option C is always a mistake. There are two situations where it is justified and a third where it is worth considering.
First: the platform is your product. If you sell courses, subscriptions or certifications and the platform generates revenue, its user experience, speed of change and data model are a competitive advantage. The same rules then apply as for any SaaS product: multi-tenancy, billing, isolation of customer data. We covered those decisions in our article on SaaS architecture, multi-tenancy and billing cost. Bending Moodle into a consumer product usually becomes a fight with its assumptions.
Second: learning is embedded in a regulated operational process. For example, a company where an employee may only perform a task (operate equipment, work at a station, sign off a document) if their training is current, and the operational system has to check that in real time. The “LMS” is then really a module of a competence and authorisation system, with a different data model than a course catalogue. Building from scratch can be simpler than bending a system that thinks in courses and enrolments.
Third, worth considering: an unusual data or interaction model. Simulations, practical exercises assessed by a supervisor, project-based learning where the “course” is not the basic unit at all. Even here, check first whether a separate tool connected to the LMS through LTI 1.3 is enough, rather than replacing the whole system.
In all three cases ask one control question: in three years, will you be willing to fund a team that keeps developing this system, the way you fund the team behind your core product? If the answer is no, option C is too big.
Which e-learning standards must you support: SCORM, xAPI, cmi5 or LTI 1.3?
Standards are where decisions made at the start weigh on a project longest. They cover three different things: packaging and tracking courses, recording learning experiences, and plugging in external tools.
SCORM is the oldest and still the most widespread way to package courses. Version 1.2 dates from 2001 and SCORM 2004 4th Edition from 2009. The course runs in a browser and talks to the LMS through a JavaScript API, reporting mainly completion, score and time. Rustici Software, a company specialising in these standards, still calls SCORM 1.2 “the industry workhorse”. In practice, if you buy off-the-shelf courses from content vendors or have a library built up over the years, your platform must run SCORM packages. For option C this is far from trivial: correct SCORM support across all the quirks of different versions is a module in its own right that has to be built, tested against packages from many authoring tools and maintained.
xAPI (the Experience API) addresses SCORM’s limits. Instead of asking “did they complete the course”, it records any event as an “actor, verb, object” statement in a repository called an LRS (Learning Record Store). An event can come from a course, a simulator, a mobile app working offline, or an operational system. xAPI 2.0 was approved as IEEE standard 9274.1.1-2023 on 30 March 2023, and the ADL Initiative described it as the first IEEE standard released as open source. For L&D teams xAPI is especially valuable in option B, because it lets you record learning that happens outside the LMS, for instance in a CRM widget, and still report it in one place.
cmi5 is an xAPI profile, released in 2016, that adds what xAPI deliberately leaves undefined: rules for launch, authorisation, reporting and course structure between an LMS and the content. In practice cmi5 combines SCORM’s predictability (the LMS knows what it launched and when the course counts as passed) with xAPI’s flexibility (content can be hosted elsewhere and work on mobile devices). If you commission new content, ask producers to support cmi5 alongside SCORM.
LTI 1.3 (Learning Tools Interoperability) is the 1EdTech standard for connecting external tools to a learning platform without separate logins. The core specification for version 1.3 is dated 16 April 2019. A launch uses the OpenID Connect third-party initiated login flow, messages are signed JSON Web Tokens, and services use OAuth 2.0. The LTI Advantage services add grade passback, membership lists and deep linking of content. LTI 1.3 is the natural way in option A or B to add a specialist tool (a simulator, a lab environment, a practice platform) without replacing the LMS.
A simple rule for the RFP: the platform must run SCORM 1.2 and 2004 (your existing content library), send and receive xAPI statements with an external LRS (learning outside the LMS), support cmi5 (new content) and act as an LTI 1.3 platform (external tools). Any gap should be a deliberate decision with a reason, not an oversight discovered after go-live.
How do you integrate the platform with your HRIS and SSO so data stays consistent?
Integrations are where learning platform projects most often overrun. Not because they are technically hard, but because they require decisions nobody has taken before. We cover general approaches in our guide to software integration patterns, and the glossary entry on systems integration explains the concept. Here we focus on what is specific to an LMS.
Sign-in. Employees should sign in with the same account they use for email and the intranet, through the company’s identity provider. The standard protocols are SAML 2.0 and OpenID Connect. Single sign-on alone raises engagement, because it removes the most common barrier: a forgotten password for a system used once a quarter. External users (partners, customers) need a separate mechanism and a separate decision about who manages their accounts.
Account lifecycle. A new hire should get an account and onboarding assignments on day one, someone changing roles should get the new mandatory courses, and a leaver should lose access. Automated provisioning, updating and deprovisioning is what the SCIM protocol (RFC 7643 and RFC 7644, September 2015) is for, and many identity providers support it. The alternative is synchronisation through the HRIS API or a scheduled file import, but each needs error handling and monitoring.
Source of truth. This is the most important decision, and it is usually never written down. The HRIS should be the source of truth for employment, job title, department, manager and location. The LMS should be the source of truth for assigned and completed training. Trouble starts when both sides edit the same data: an LMS administrator corrects a department because “HR has it wrong”, and the nightly sync overwrites the fix. Document which system wins on conflict for each field, and lock editing of that field on the losing side.
The return flow. Should completion data flow back to the HRIS, for example to the employee record or the performance review module? If so, that is a separate integration with its own requirements for format, timing and error handling.
Personal data. The platform processes employee data, and in option B also data about how employees behave in other tools. Agree the scope of that data, the retention period for training history and the legal basis for processing with your data protection officer before the project, not after it. Click-level analytics may be tempting, but under the GDPR every additional category of data is an additional obligation, and in several EU countries works councils have a say in systems that can monitor employee behaviour.
In the estimate, price each of these integrations separately and add time for testing with real HR data, because only then do the edge cases surface: people with two contracts, transfers between group companies, long-term leave.
Does your platform fall under the European Accessibility Act?
Many L&D teams are asking this now, and the answer is more nuanced than some proposals suggest. The European Accessibility Act (Directive (EU) 2019/882) applies from 28 June 2025 through national transposition laws, for example Poland’s Act of 26 April 2024 on ensuring compliance with accessibility requirements for certain products and services by economic operators, in force since that date.
Two elements of the scope are decisive. First, the services covered are those offered or provided to consumers, meaning individuals acting outside their trade, business or profession. Among the covered services are e-commerce services, that is services provided at a distance through websites and mobile apps with a view to concluding a consumer contract. Second, services provided by microenterprises are exempt.
What does that mean in practice? An internal platform used only by employees is usually not a consumer service and does not fall under the Act. A platform where individuals buy courses can be an e-commerce service, and then accessibility requirements apply at least to the purchase journey, which in practice is hard to separate from the rest of the site. A platform for business customers or partners sits in between and needs analysis of the specific model. When in doubt, take legal advice, because the answer depends on how you sell.
None of this means an internal platform can be built without regard for accessibility. Article 5 of Directive 2000/78/EC requires employers to provide reasonable accommodation for people with disabilities, and a mandatory course a blind employee cannot complete is a short road to a dispute. Public sector bodies in the EU have separate obligations under the Web Accessibility Directive, and in the United States Section 508 binds federal agencies and, through procurement, their suppliers. The practical reference point everywhere is WCAG 2.2, a W3C Recommendation since 5 October 2023, which added nine new success criteria, including focus not obscured, alternatives to dragging movements, minimum target size and accessible authentication. For an introduction, see our guide to getting started with WCAG accessibility testing.
Accessibility looks different in each option. In option A it depends on the theme and plugins, each of which must be checked separately. In options B and C you control the interface fully and can design for accessibility from the start, which is far cheaper than retrofitting. Remember too that an accessible platform does not guarantee accessible courses: content from external producers has to be verified separately, and the requirement written into their contracts.
What does each option cost?
The honest answer is always “it depends on scope”, but an order of magnitude can be shown. The figures below are an illustrative estimate, not a quote: we assume example team compositions and durations, and senior day rates from ARDURA’s pricing page (bottom and top of each range) at 21 working days a month. They cover delivery team effort only, and exclude licences, hosting, content production, discovery and your own staff time.
| Option | Example team | Duration (assumption) | Team cost per month | Total estimate (net) |
|---|---|---|---|---|
| A. Extend Moodle | backend developer (1 FTE), frontend developer (0.5), QA engineer (0.25) | 3 months | approx. EUR 12,100–17,300 | approx. EUR 36,000–52,000 |
| B. LMS core + own layer | backend (1), frontend (1), QA (0.5), project manager (0.5) | 5 months | approx. EUR 20,900–29,500 | approx. EUR 105,000–148,000 |
| C. Custom LMS from scratch | backend (2), frontend (2), QA (1), project manager (0.5), DevOps (0.5) | 10 months | approx. EUR 41,900–59,400 | approx. EUR 420,000–594,000 |
Where do the numbers come from? For option B, for example: a senior backend developer is EUR 335–485 a day on the pricing page, so EUR 7,035–10,185 a month at 21 days; a senior frontend developer EUR 315–445 a day, so EUR 6,615–9,345; half a senior QA engineer (EUR 335–465 a day) is EUR 3,518–4,883; half a senior project manager (EUR 355–485 a day) is EUR 3,728–5,093. That gives roughly EUR 20,900–29,500 a month, or EUR 105,000–148,000 over five months. For comparison, the ARDURA software development services page puts a dedicated team at from about EUR 14,500 a month.
The factors that move these figures matter more than the figures themselves. In option A cost rises with the number of past core modifications, because they first have to be moved into plugins or removed. In option B integrations are the most variable item: one HRIS with a good API is different work from three systems across group companies, one of which only offers file exports. In option C cost is driven mostly by the scope of admin and reporting functions, which are easy to leave off mock-ups because they are not glamorous, and which took years to mature in off-the-shelf systems.
Weigh the size of option C against its risk. The ten-month estimate for a full LMS assumes a well-described scope. In practice such projects tend to grow, because every stakeholder adds features that “Moodle already has”. With option C in particular, start with a minimum version for one user group; we explain the reasoning in our guide to budgeting a first product version, and the glossary entry on the MVP defines the term.
What does the platform cost to run after launch?
Build cost is usually the smaller part of total cost of ownership over three to five years, and each option has a different profile after go-live.
Option A depends on whether you stay on long-term support releases or upgrade every six months. Under Moodle’s schedule, non-LTS versions get about a year of general support and eighteen months of security support: security support for 5.0 ends on 5 October 2026, and general support for 5.1 ends on the same day. Every plugin you added has to be tested and adapted with each upgrade. Add hosting, backups, monitoring and administration. If you rely on third-party plugins, check that they are actively maintained, because an abandoned plugin blocks the upgrade of the whole system.
Option B adds maintenance of your own layer to option A’s costs: fixes, adaptation to API changes in the core and in the systems it integrates with, and development driven by usage data. As a working assumption, budget part of a developer’s time, though the real figure depends on how fast the business expects change. The advantage is that the layer can be released independently of the core.
Option C means the whole system is yours: security, compatibility with new browser versions, support for new content formats, changes in data protection and accessibility law. A realistic assumption is a standing team, not a fraction of one person. If the organisation does not plan to fund that team, within two or three years the system becomes legacy that has to be rescued or replaced.
Some running costs are the same whatever the option: authoring tool licences, content production and refresh, the time of subject matter administrators and, if you use xAPI, running the LRS. When you present the budget, show build cost and five-year cost side by side. In option C the second number is often a surprise, and it is better that it lands before the decision than after.
How do you write an RFP that produces comparable proposals?
Proposals for e-learning platforms are unusually hard to compare, because vendors answer different questions. One prices a deployment of an off-the-shelf system, another a build from scratch, a third a subscription. A few elements in the RFP make this much easier.
Describe the problem, not the solution. Instead of “we need a platform with gamification and a mobile app”, write: “our 2,000 retail staff must complete three mandatory courses every quarter; engagement today is X and the audit report takes two days to compile by hand; we want training available from in-store terminals and line managers to see their team’s overdue items”. A description like this lets the vendor propose option A, B or C and justify the choice.
Ask for an answer for each option, or a reason why not. Ask every vendor to describe how they would solve the problem in options A, B and C, or explain why one of them makes no sense. This quickly reveals who is analysing your problem and who is selling their standard product.
List the standards and integrations. SCORM 1.2 and 2004, xAPI with an external LRS, cmi5, LTI 1.3, single sign-on through SAML 2.0 or OpenID Connect, account provisioning through SCIM or the HRIS API. Name the systems the platform must connect to and the interfaces they expose.
Set the accessibility requirements. State the WCAG 2.2 conformance level (usually AA) and how it will be verified at acceptance: an audit, testing with assistive technologies, an accessibility statement.
Ask about ownership and exit. Who owns the code of the layer or the system? In what format will you receive your data if you change vendors: training history, certificates, xAPI statements? For option A: does the vendor modify core, or only write plugins?
Ask for a five-year cost. Build, maintenance, hosting, licences, version upgrades. Only then can you compare an off-the-shelf deployment with a build.
If nobody internally can judge the architecture of the responses, consider starting with a short, paid discovery phase as a separate engagement, producing a document you can then use in the RFP.
Which risks most often derail a learning platform project?
Whatever the option, a handful of risks recur often enough to belong in the risk register from day one.
Migrating training history. An organisation that has run mandatory training for years holds completion history needed for audits and inspections. Moving it to a new system requires mapping courses (old “Health and Safety 2021” corresponds to which item in the new catalogue?), verifying completeness and deciding what to do with records of people who have left. This stage is often priced as a few days of “data import” and takes weeks.
Migrating content. SCORM packages from different authoring tools, from different years, sometimes without source files. Some will run in the new system unchanged, some need fixes, some have to be rebuilt. Before signing, test a representative sample of packages, including the oldest ones.
Compliance reporting. The report auditors receive from today’s spreadsheet has a specific format and logic: who, when, which course, valid until when. If the new system does not reproduce it exactly, the team will go back to the spreadsheet. Gather reporting requirements from auditors and legal at the start, not at acceptance. The list of obligations the platform is expected to help evidence keeps growing: one example is Article 4 of the EU AI Act, applicable since 2 February 2025, which concerns the AI literacy of staff at organisations deploying AI systems. Its wording has been the subject of proposed amendments, so check the current text before you design that report.
Adoption. A new platform nobody uses is worse than the old one, because it cost more. The rollout plan should include a pilot with one group, communication through line managers rather than only through HR, and engagement measured from the first week.
Admin scope. Mock-ups show the learner’s screens. They do not show the screens for an administrator managing a thousand courses, role-based assignments, exceptions, deadline extensions and bulk changes. In option C this is often half the work, and proposals sometimes leave it out.
Vendor dependency. In options A and C too much knowledge of the system can stay with the supplier. Require documentation, access to the code repository and knowledge transfer as part of the contract, not as an extra service.
What does the decision look like in one table?
The table below brings together the criteria that help choose an option. It does not replace analysis, but it gives structure to the conversation with the board and with vendors.
| Criterion | A. Extend Moodle | B. Core + own layer | C. Custom LMS from scratch |
|---|---|---|---|
| When it fits | The problem is configuration, sign-in, reporting or content | Learning must live in work tools; you need your own look and a manager dashboard | The platform is a product or part of an operational process |
| Estimated build cost | approx. EUR 36,000–52,000 | approx. EUR 105,000–148,000 | approx. EUR 420,000–594,000 |
| Time to launch (assumption) | 2–3 months | 4–6 months | 9–12 months or more |
| SCORM, xAPI, reporting | Ready in core and plugins | Ready in core | To be built and maintained |
| Control over experience | Limited to theme and plugins | Full in the learner layer | Full |
| Running cost | Core and plugin upgrades | As A, plus part of a developer for the layer | A standing team |
| Main risk | Past core modifications | Integration complexity | Scope creep, maintenance cost |
| Accessibility | Depends on theme and plugins | Designed in the layer, core to be checked | Designed from the start |
If your situation lands in column A on “when it fits”, do not let attractive mock-ups move you to column C. If it lands in column C, do not try to save money by bending Moodle into a role it was not designed for. Most decisions, though, sit in column B, and that is where to start the analysis.
What are the key takeaways for HR and L&D leaders?
A custom e-learning platform is a decision whose cost only becomes visible after a few years. A good decision starts with a diagnosis of why the current system does not work, not with a feature list from someone else’s demo.
Remember six things. First, “missing features” is rarely the real cause; check whether the problem is sign-in, integration with the org structure in your HRIS, the catalogue or the content. Second, consider three options, not two: extending, a standard core with your own layer, and a system from scratch. Third, a custom LMS from scratch makes sense when the platform is your product or part of an operational process, and only if you are prepared to fund a team that keeps developing it. Fourth, put SCORM, xAPI, cmi5 and LTI 1.3 into the requirements, along with single sign-on and automated account provisioning, and for every data field decide which system is the source of truth. Fifth, check whether your platform is a consumer service under accessibility law, but design for accessibility regardless of the answer. Sixth, compare proposals on five-year cost, not build cost.
If you want to check which option fits your organisation, or need a team to build your own layer or the whole system, explore our custom software development services, or book a call. Discovery takes one to two weeks and ends with a scope, an architecture and an estimate you can take to the board.